Michael Brown, director of the Defense Innovation Unit has
said "Artificial Intelligence is fundamentally redefining how we live,
work, and fight wars.� Within the Department of Defense, AI has the potential
to transform how the Department operates at all levels, from business to the
battlefield. In the face of competition from China and Russia, DoD aims to
maintain its technological edge through establishing a more decentralized,
experimental procurement approach: cultivating a leading AI workforce, engaging
academic, commercial, and international allies and partners, and developing
ethical and lawful guidelines for AI use." (Brown, 2019).
Businesses are investing billions of dollars on
artificial intelligence (AI), transforming their products and services.� The
government has begun to realize the potential, making it a priority in national
strategic guidance in recent years.� Many organizations are spending
considerable time and resources on artificial-intelligence research and
development.� However, its potential application to the Defense Acquisition
System (DAS) has yet to be examined.� This thesis examines this issue.� Procurement
is full of highly manual, laborious, and data-centric tasks.� Artificial
intelligence could automate these tasks, delivering procurement professionals
relevant real-time data.� It can pre-populate required forms, streamline market
research, and help procurement personnel manage contracts.� It can provide
management with better governance by giving them better insight into spending
among others.� It can aid procurement professionals, saving them time to focus
on more strategic tasks like negotiations or network management.
To better appreciate why artificial intelligence could be
important for U. S. procurement, we can look at the current national-security
picture.� The United States is entering a new era characterized by emerging
technologies, great-power competitions, and new battlespaces.� Future conflicts
will likely differ from past conflicts in that adapting and moving quickly will
be vital to victory.� America's military effectiveness in the new environment
is unclear.� Great powers like China and Russia are growing their military
capabilities and looking to take advantage of new technologies.� China has
detailed a strategy to become the world leader in it by 2030 (China State
Council, 2017).� Russia is also investing heavily in artificial intelligence
capabilities and believes "whoever becomes the leader in this sphere will
rule the world" (Simonite, 2017).
The National Security Strategy (NSS) directs agencies to
prioritize emerging technologies such as artificial intelligence (White House,
2017).� The National Defense Strategy calls for delivering performance at the
speed of relevance and to modernize key capabilities (DoD, 2018).� In 2018 the U.
S. Department of Defense (DoD) established a Joint Artificial Intelligence
Center (JAIC) and then released a Defense Artificial Intelligence Strategy
directing agencies to adopt artificial-intelligence capabilities to increase
the speed and agility of operations (DoD, 2019).� The Fiscal Year 2019 National
Defense Authorization Act (NDAA) established the National Security Commission
on Artificial Intelligence (The John S. McCain National Defense Authorization
Act for Fiscal Year 2019 [NDAA], 2018), and the Defense Innovation Board (DIB)
and Defense Advanced Research Projects Agency (DARPA) have studied the
implications of it for the DoD for many years.
Meanwhile, frequently Congress calls for the reform of
the Defense Acquisition System (DAS) (Barnett, 2019).� Since the 1990's, the DoD
Weapons Systems Acquisition and DoD Contract Management have both been
categorized as high-risk programs by the Government Accountability Office (GAO,
2019).� Additionally. the DoD Inspector General (DoDIG) has identified
Acquisition and Contract Management as a top-ten management challenge for
FY2020, stating "DoD programs still fall short of cost, schedule, and
performance expectations" with "395 open recommendations [for
improvement] related to acquisition and contracting� (DoD OIG, 2019).� Recently
DoD Instruction 5000.02 established a new Adaptive Acquisition Framework with
the objective of "delivering effective, suitable, survivable, sustainable,
and affordable solutions to the end user in a timely manner" (Office of
the Under Secretary of Defense for Acquisition and Sustainment, 2020).� A
rewrite of DoD Directive 5000.0, The Defense Acquisition System, is expected
before the end of this year. These changes are being touted as "the most
transformational acquisition policy change in decades" (Lord, 2019).
Time will tell if these reforms will satisfy decades of
criticism; however, historically the initiatives of the Office of the Under
Secretary of Defense for Acquisition and Sustainment (OUSD (A&S)) have been
unable to remove the GAO's categorization of high risk from DoD Acquisition and
Contract Management or satisfy DoD IG's assessment.� These current reforms do
not mention artificial intelligence or any new technology.� To the best of our
knowledge, there are no initiatives within OUSD (A&S) to implement any new
technologies in the acquisition process. �Moreover, recent reports from the DIB
and RAND Corporation suggest that DASs must be reorganized to support artificial-intelligence
systems.�
This thesis describes the potential impact artificial
intelligence could have on DoD acquisition and contracting.� Information was
collected through interviews, online forums, and reports. Organizations that
have integrated artificial intelligence into their procurement processes were
examined to identify successes and failures.� This thesis will close by
providing a potential roadmap for the DoD to adopt commercial-style artificial
intelligence for procurement.� More specifically, this thesis will address the
following questions:
�
How can the DoD adopt commercial-style artificial intelligence
for procurement?
�
Why did organizations adopt artificial intelligence into their
procurement processes?
�
What were the expected benefits of artificial intelligence for
procurement?
�
How well did the artificial intelligence improve speed and
agility?
�
What are the obstacles and risks?�
�
What are the lessoned learned?
Artificial intelligence is all around us, integrated and
woven into our daily lives in more ways than you are probably aware.� It is in
our homes, our cars, our phones, and our workplaces.� It flies planes and
drones.� It drives cars and trucks.� Artificial intelligence recommends what
music we should listen to or what movie we should watch next.� It suggests
products, and it personalizes our news and social media feeds.� It organizes
email inboxes and provides smart replies.� Artificial intelligence determines
the most efficient routes when you are driving, arranges economic matches for
us to share rides, and determines the rate charged when you use applications
like Uber or Lyft.� Artificial intelligence helps decide whether or not you
should receive a loan, detects fraud, and manages stock portfolios.� It
translates languages, provides customer service, and prepares taxes.� Virtual
assistants like Alexa and Siri allow you to check the weather, text or call
someone, answer general questions, set up meetings, and order products online
with only the sound of your voice.� Artificial intelligence has won Jeopardy
and beaten the world's best at Chess, Go, and Poker, tasks previously thought
to be impossible.� It even helped tennis player Novak Djokovic win Wimbledon by
helping him to not only improve his own game but to strategize on how to play
specific opponents (Loudin, 2019).� Artificial intelligence powers websites,
applications, and search engines.� The list goes on and on, and it will only
continue to grow since we are still in the early stages of harnessing its power.
Artificial intelligence can fundamentally transform
businesses.� It is already changing healthcare by helping professionals detect
and manage diseases, monitor health, perform surgery, personalize cancer
treatments, and identify people at risk of suicide (National Academy of
Medicine, 2019).� But more than this, it will change business strategy.� For
example, right now Amazon's artificial intelligence suggests items you may want
to buy.� It does an adequate job at this task.� However in the future, after it
has collected more data, it could become sufficiently accurate to change
Amazon's business model.� It will become "more profitable for Amazon to
ship you goods that it predicts you will want rather than wait for you to order
them" (Agrawal et al., 2018).� It will save the consumer from having to
shop, giving them items they want and need, and change "Amazon's business
model from shopping-then-shipping to shipping-then-shopping.�
Artificial intelligence is starting to be adequate in
preparing nontrivial legal documents.� A study compared it to twenty human
attorneys in examining non-disclosure agreements and identifying risks within
them.� Artificial intelligence had an average accuracy of 94 percent, while the
lawyers had 85 percent, as judged by independent auditors. �More importantly,
the lawyers took about 92 minutes on average for a contract while the
artificial-intelligence program needed only 26 seconds (Chaffin, 2018)."
In this thesis, I will use the term "artificial
intelligence" broadly, as a term for computers, software, or algorithms
that can perform intelligent tasks traditionally performed by human beings.� This
includes robotic process automation, machine learning, natural language
processing, and deep learning among others.
Such artificial intelligence has been around for over 60
years.� It has experienced a few booms and busts.� However, today's boom is very
strong for several reasons.� First is the amount of available data; data is
essential and the fuel for artificial intelligence.� In the past, it was costly
and challenging to acquire data. �Today we have oceans of data.� Cloud services
allow for a massive amount of storage and sharing.� Second, the required
computing power needed for important artificial-intelligence applications is
here.� Third, mobile connectivity improvements like the fifth generation give artificial
intelligence the necessary bandwidth and processing speed required to work remotely.�
Fourth, both industry and government support the current boom by having made
significant investments in artificial intelligence.� Last, advanced algorithms
are now possible due to improvements in computer hardware (Torres, 2018).
An example of the impact of artificial intelligence
occurred in 2016 when AlphaGo, an artificial intelligence algorithm, defeated
the world champion grandmaster in the ancient game of go. �The grandmaster has
since retired, convinced that machines "cannot be defeated" (Webb,
2019).� More impressive is the recent defeat of professional poker players at
six-player poker by the program Pluribus.� Poker presents incomplete
information, unlike chess and go, and involves multiple opponents who sometimes
bluff.� Yet Pluribus defeated them convincingly, by mixing strategies and
betting in ways the humans did not.� These milestones are noteworthy for two
reasons.� First is how the programs learned to play the games, by learning from
experience.� Only 13 days were needed to master go and 8 days to master poker (Carnegie
Mellon University, 2019).� Secondly, the programs made moves never before
considered.� In the second go game�s 37th move, AlphaGo made a move
never seen in thousands of years of gameplay; it was initially thought to be a
mistake but it turned out to be brilliant (Docherty & Fanning, 2019).
Given these feats, one can see the potential utility for
the military.� "While the current rules of wargaming are complex and the
tactical moves subtle, it is not hard to imagine a version of (AlphaGo) that
could learn to play and win a war game. �The implications for future battle
management could prove enormous (Arquilla & Denning, 2019)."� These
examples illustrate the potential utility of artificial intelligence and its
relevance today: Artificial intelligence will not only improve business
processes, but it will win the future battlespace. �Application of artificial
intelligence to procurement could not only automate and streamline business
processes but could think of "moves" never before considered.� It can
merge and interpret a vast amount of data to predict an outcome.� "The
ability to fuse enormous amounts of data from across disparate data sets and
provide meaningful answers are exactly what artificial intelligence and machine
learning were designed to do" (Weinbaum & Shanahan, 2018).
Artificial intelligence is a major strategic concern for
the DoD and the U.S. Government as a key to maintaining a competitive advantage
with other countries.� An executive order established the American Artificial
Intelligence Initiative on Maintaining American Leadership on Artificial
Intelligence in February 2019, and directed federal agencies to act to support
several related objectives (Executive Order No. 13859, 2019).� Recently, the
Office of Science and Technology Policy released the first annual report on the
Initiative (Office of Science and Technology Policy [OSTP], 2020).� It concluded
that significant progress has been made. �Figure 1 was taken from this report
and represents a timeline of U.S. government actions to advance the initiative.
Figure 1.
Timeline of U. S. Government Actions to Advance the American Artificial
Intelligence Initiative. Source: (OSTP, 2020).
The figure shows some products produced in recent years as
a result of the Initiative, but the U.S. government does have a long history
with artificial intelligence.� For example, for nearly six decades, the Defense
Research Projects Project Agency (DARPA) has supported artificial-intelligence
research (DARPA, 2018).� Other parts of the U.S. government also have a long
history with artificial intelligence.�
More specific to the DoD is the National Security
Strategy (NSS) which states that "to maintain our competitive advantage,
the United States will prioritize emerging technologies critical to economic
growth and security", such as artificial intelligence (White House,
2017).� Modernization was identified as a critical objective and it specifically
listed acquisition as a priority.� "The United States will pursue new
approaches to acquisition to make better deals on behalf of the American people
that avoid cost overruns, eliminate bloated bureaucracies, and stop unnecessary
delays so that we can put the right equipment into the hands of our forces.� We
must harness innovative technologies outside of the traditional defense
industrial base".
The 2018 National Defense Strategy (NDS) was framed in
the same manner as the NSS and the American Artificial Intelligence initiative.�
The guidance acknowledges our eroding military competitive advantage, and it states
that new technologies like artificial intelligence are "the very
technologies that ensure we will be able to fight and win the wars of the
future" (DoD, 2018).� It recommends that we "must modernize our
military or the result would be a Joint Force that has legacy systems
irrelevant to the defense of our people" A specific defense objective
included is to "continuously delivering performance with affordability and
speed as we change the departmental mindset, culture, and management systems [and
that the U.S. must] organize for innovation" and "modernize key
capabilities".� Furthermore, it recognizes the "backlog of
procurement [that] has grown in the last decade and can no longer be ignored"
and that that "the department will invest broadly in military application
of autonomy, artificial intelligence, and machine learning, including rapid
application of commercial breakthroughs, to gain competitive military
advantages".� While the guidance did not call out the DAS by name, it
calls for action by stating to "reform the department for greater
performance and affordability" to be able to "deliver performance at
the speed of relevance".� Business processes in general were mentioned.
Current processes are not responsive to need; the
Department is over-optimized for exceptional performance at the expense of
providing timely decisions, policies, and capabilities to the warfighter. Our
response will be to prioritize speed of delivery, continuous adaptation, and
frequent modular upgrades.� We must not accept cumbersome approval chains,
wasteful applications of resources in uncompetitive space, or overly
risk-averse thinking that impedes change. Delivering performance means we will
shed outdated management practices and structures while integrating insights
from business innovation (DoD, 2018).
In 2018, the DoD Artificial Intelligence Strategy was
released, directing the adoption of artificial intelligence within the DoD
(DoD, 2019).� The Joint Artificial Intelligence Center (JAIC), established
earlier in the year, would serve as the principal agency to implement the
strategy and "to accelerate the delivery of artificial intelligence-enabled
capabilities, scale the Department-wide impact of artificial intelligence, and
synchronize DoD artificial intelligence activities to expand Joint Force
advantages".� The strategy reiterates much of the NSS and NDS, but it is
more specific on the artificial-intelligence tasks.� One focus area called for
streamlining business processes or tasks that are "highly manual,
repetitive, and frequent" and that "the ability of AI to reduce
inefficiencies from manual, laborious, data-centric tasks will be harnessed
across the Department with the objective of simplifying workflows and improving
the speed and accuracy of repetitive tasks".� The strategy goes on to say
the DoD "will prioritize the fielding of artificial intelligence systems
that augment the capabilities of our personnel by offloading tedious cognitive
or physical tasks and introducing new ways of working" and that
"there are numerous artificial intelligence applications that could
improve our day-to-day operations".� The strategy also offers a much
blunter assessment of what the cost would be if implementation fails. �"The
costs of not implementing this strategy are clear. Failure to adopt artificial
intelligence will result in legacy systems irrelevant to the defense of our
people."� New technologies like artificial intelligence not only present a
unique opportunity to help the DoD overcome challenges in acquisition and
contracting, but are essential to remaining relevant.
Other significant artificial-intelligence strategy was
the 2019 DoD Digital Modernization strategy which "provides a roadmap to
support implementation of the National Defense Strategy lines of effort through
(the priorities) of cloud, artificial intelligence, command, control and
communications and cybersecurity (DoD, 2019)."
An important event occurred in 2018 when DARPA invested $2
billion over several years in over 80 programs and called it the "Artificial
Intelligence Next" campaign, building on DARPA's decades of research
(DARPA, 2018).� Also, the Defense Innovation Unit was established in 2015 in
Silicon Valley to leverage better partnerships with the most innovative
American companies.� AI.gov was established recently and is an "All of
Government" portal for artificial-intelligence activities.� $1 billion is allocated
for DoD artificial-intelligence projects in 2020 and $268 million of this is allocated
to the JAIC.
So just how is the DoD spending their budget and
implementing this guidance?� Most initiatives are focusing on the battlefield.�
Weapons systems like Tomahawk missiles and the Phalanx close-in weapons system
are fully autonomous and have been in service for decades.� The DoD uses a
fleet of drones for a variety of missions; prominent in recent years are the Sea
Hunter and Project Maven.� DARPA developed Sea Hunter as an autonomous surface
ship that tracks submarines.� Project Maven was a collaboration with Silicon
Valley to develop computer-vision algorithms that could process and analyze
video footage from drones.� Other artificial-intelligence warfighting projects
include M1 tank engine analysis, summarizing sensor data during warfare,
automated pilots for combat simulation, detecting changes to buildings in
aerial photographs, cyberattack warning, predictive maintenance, battlefield
management and autonomy, and sensor-based target identification. �A central
project in 2020 for the JAIC is Artificial Intelligence for Maneuver and Fires.
It focuses on the warfighting operations like operations-intelligence fusion,
accelerated sensor-to-shooter timelines, autonomous and swarming systems,
target development, and operation-center workflows (Shanahan, 2019).� Other
projects help with decision making and risk analysis and help with joint
command and control.
While most of the artificial-intelligence initiatives
focus on the battlefield, many others focus elsewhere.� AI Next includes
initiatives like human-language technologies, biomedical advances, space-based
imagery analysis, analysis of microbiologic systems, and control of prosthetic
limbs (DARPA, 2018).� There are DoD initiatives for humanitarian assistance and
disaster relief.� Intelligent business automation for critical DoD processes
like background investigations and accrediting software systems is underway.�
Another recent major effort for the JAIC is the Joint Common Foundation (JCF)
which " will be a platform that will provide access to data, tools,
environments, and other certified platforms to enable software/artificial
intelligence engineers to rapidly develop, evaluate, test and deploy artificial
intelligence-enhanced solutions to warfighters." (Shanahan, 2019)� However,
the NSCAI interim report observed, "The United States is not translating
broad national artificial intelligence strengths and artificial intelligence
strategy statements into specific national security advantages.� Key departments
and agencies have not yet fully embraced high-level strategy pronouncements and
therefore critical national security missions have not incorporated artificial
intelligence (NSCAI, 2019)."
Within the U.S. Government, the agency of Health and
Human Services (HHS) and the General Services Administration (GSA) are putting artificial
intelligence into their procurement.� Health and Human Services developed and
recently deployed Accelerate, which "uses a neural network [to perform]
cluster analysis of 10 years of HHS contracts and provide visibility into price
discrepancies across its portfolio" (Malone, 2020).� They claim it will
save $30 million over five years.� This will help procurement professionals get
relevant pricing data when they perform market research.� It also prepopulates
the forms with applicable terms and conditions in real time, saving personnel time.
Meanwhile, the General Services Administration is
implementing robots and other related technologies within their agency to
streamline processes and "establish a foundation for greater data-driven
decision-making through artificial intelligence" (Nakasone, 2018).� It is
also helping other government agencies adopt artificial intelligence into their
operations in three ways.� First, they have created an artificial-intelligence "community
of practice", which they hope will become the largest repository of use
cases in the government.� Second, they created the Center for Excellence for artificial
intelligence, performing consultant work with other government agencies and bringing
academic expertise to solve business problems.� Third, their Federal
Acquisition Service lets government agencies buy new artificial-intelligence
services quickly.
Other agencies like DARPA and the Department of Homeland
Security are considering process automation tools to increase efficiencies in
procurement.� Last year, both agencies solicitated for information on robotic
and intelligent machines but have not reported any results.
Businesses realize the potential of artificial
intelligence and are using it to transform their products and services.�
Investment in artificial intelligence is accelerating with projections that it
will reach $97.9 billion in 2023 (IDC, 2019).� American companies are competing
for artificial intelligence talent by acquiring relevant startups at an
increasing rate.� Since 2010, there have been 635 artificial intelligence
acquisitions that span every industry (CB Insights, 2019).� In 2019, the job most
recruited was an artificial-intelligence specialist (Hill, 2019).� It is the
third year in a row that artificial intelligence has claimed the top position,
growing 74% annually.� Another study found that 62% of North American companies
have one or more artificial-intelligence systems in operation, and 87% plan to
adopt some artificial intelligence in the next two years (Cognilytica, 2020).�
Despite these trends, businesses are still trying to
understand what artificial intelligence means for their future business.� A
study by PricewaterhouseCoopers concluded that "global GDP will be up to
14% higher in 2030 as a result of the accelerating development and take-up of artificial
intelligence � the equivalent of an additional $15.7 trillion"
(PricewaterhouseCoopers [PwC], 2017).� They claim that three factors will drive
the impact:
�
Productivity gains from businesses automating processes
(including use of robots).
�
Productivity gains from businesses making more efficient use of
their existing labor force with artificial-intelligence technologies.
�
Increased consumer demand resulting from the availability of
personalized and higher-quality artificial intelligence-enhanced products
and services (PwC, 2017).
Productivity gains will come mostly from automation of labor-intensive
manual tasks.� It will free up time for employees to perform more valuable or
strategic missions.� Businesses will have better insight into their customer's
preferences.� Shell's chief technology officer stated that �Artificial
intelligence enables us to process the vast quantity of data across our
businesses to generate new insights which can keep us ahead of the
competition" (Castellanos, 2020). While most companies are applying artificial
intelligence for narrow and well-defined tasks, the most progressive enterprises
want to use artificial intelligence to do things never done before.� Their goal
is to disrupt markets by creating entirely new services, products, or
businesses by harnessing the power of artificial intelligence.
Many companies are adopting new
technologies for procurement.� A recent survey found that 58% of businesses are
automating procurement (Deloitte, 2019).� Another survey of Chief Procurement
Officers (CPO) from Global Fortune 500 companies found that 67% of ranked artificial
intelligence was one of their top three priorities for the next ten years
(Marlinghaus, 2018).� Large technology companies are developing that capability
in-house; others are looking for contracts.� Figure 2 lists artificial-intelligence
and procurement startups that could help companies apply artificial
intelligence to procurement.� For instance, Prevedere helps a company forecast
future business from global data from hundreds of sources (Prevedere, n.d.); one
can then correlate that with internal data for insights.� Rapid RPA automates a
customer's business processes to enhance productivity (Echelon/RPA, n.d.).�
Other companies like Icertis, Sievo, and Seal are specific to procurement.�
Icertis is a bigger startup with a valuation of $1 billion (D'Onfro, 2019).� It
is a platform to manage the life cycle of contracting, from offers to negotiation,
acceptance, and management.� More specifically, it automates contract tasks and
analyzes past negotiations (D'Onfro, 2019).� Seal helps with contract
management by helping their customers find the contracts quickly, extract critical
data, and perform analytics (Seal, n.d.).� Sievo provides contract management, spending
analysis, and forecasting (Sievo, n.d.). �Similar startups for artificial
intelligence and procurement are growing.
Figure 2.
List of Artificial Intelligence and Procurement Startups
An early study examined artificial intelligence�s utility
in assisting claims analysts within the Differing Site Condition Analysis
System (Kruppenbacher, 1984).� The study found that a ruled-based expert system
can apply successfully to contract management, and can do accurate analysis.� Another
study examined how artificial intelligence can take over daily tasks in the
future workplace (Subramanian et al., 1999).� It concluded that the technology
can be helpful for sophisticated real-time negotiations, and that machines could
be programmed to do business tasks "so that decisions can be made in a
faster and accurate way").
Another study focused on aiding complex negotiation
scenarios with artificial intelligence (Rodriguez-Aguilar et al., 2003).� It
concluded their decision-support agent could help in the real world and that it
was time to deploy such agents.� Additionally, their agent could cooperatively
solve complex problems with other intelligent programs.� Another paper proposed
a formal reputation model to evaluate bids from suppliers and manage the trade-off
between price and a firm's reputation (Klabi, Mellouli, & Rekik, 2013).�
This provides the tools a government needed to evaluate bids and guarantee that
the most trusted supplier was identified while minimizing the total cost.
A more recent thesis considered the problems that artificial-intelligence
systems would encounter within the Joint Capabilities Integration and
Development System, and how well warfighter requirements would be addressed
through the DAS through the software-acquisition pathway (Ehn, 2017).� It
concluded that such systems face problems becoming an executable DoD program.� Lack
of knowledge about artificial intelligence within the DoD could mean poorly
defined requirements, causing cost, schedule, and performance issues.� This issue
was recently studied (Defense Innovation Board [DIB], 2019) (RAND, 2019) and refactoring
the DAS to support the development of artificial-intelligence systems was
suggested.
One study examined artificial intelligence in procurement
and concluded it "will change the face of procurement" (Chopra,
2019).� Automation and robots could help automate some tasks so people can
focus on higher-priority tasks, and new startups will offer cost-effective
procurement solutions within the next five years:
Bottom line, procurement leaders must become AI-conversant
so they can onboard and direct professionals who possess the skills to leverage
the tools.� Without competent staff, it's widely believed it will be nearly
impossible for companies to build next-generation strategic sourcing and procurement
organizations (Chopra, 2019).
To better answer the research questions, we sought people
with experience integrating artificial intelligence into procurement.� We sent
out ninety-five requests but received only ten responses.� Some respondents could
not discuss their knowledge due to restrictions on sharing work-related
information.� However, we did connect with and interview five people from
government and industry.� Some respondents invited us to webinars where we could
gather more information. �Additionally, we gathered some data through podcasts
and online interviews.� We focused on why their organization implemented artificial
intelligence in their procurement processes, how it improved speed and agility,
its benefits and risks, its challenges overcome, and its lessons learned.
The organizations of the interviewees implemented
artificial intelligence for different reasons.� One organization was exploring
how "digitization" could help their procurement, since there is
always pressure to deliver savings and value.� They had already implemented some
automation to help save time and deliver value.� However, they lacked a digital
strategy and wanted to "leverage technology to help us with capacity and
to get us from transactional to more strategic, high-impact activities, like using
data to make smarter decisions, or to gain insights to negotiate better deals.
We also started to see a lot of companies providing AI and digital products and
services, so these capabilities were maturing and improving the user
experience" (J. Thong, PowerPoint
presentation, March 31, 2020).� They also hoped artificial intelligence would
help with spending analytics. They wanted to know for a given category how much
money they spend with which suppliers.� For example, if they were spending $10
million in one category with four suppliers, they might achieve a better cost
by using only two suppliers.� Finding these opportunities is difficult for
employees alone.� Artificial intelligence could provide this insight and analysis
automatically.�
One company hoped to improve insight and accessibility for
their contracts.� Because it was a global company, it had difficulty comparing
the contracts in different regions.� It was difficult and time-consuming to
compare and reconcile differences in the terms they were agreeing to.� They hoped
to solve this problem by creating a database to maintain all of their
contracts, and use artificial intelligence to help them identify significant
terms and conditions.� The U.S. government organization Health and Human
Services (HHS) was one of the first U.S. agencies to exploit artificial
intelligence.� They had three main problems that they wanted to address.� One
was that they had no insight into the data at the headquarters level and had
little ability to see what people were buying.� Another was the procurement
process itself was too convoluted and taking too long to buy something because
of the regulations and clauses.� Another was that they have much old technology
and systems and they wanted to see if there was overlap or redundancy so they
could retire some systems (O. Mek, interview with author, March 24, 2020).
The General Services Administration (GSA) wanted to use
machine learning to improve enterprise-user licensing agreements, which usually
contain thirty to fifty pages of terms and conditions.� They thought they could
use software to help them perform preliminary analysis and then have humans
look at the details (K. Nakasone, interview with author, March 23, 2020).� They
also wanted to automate repetitive low-level tasks that are well-defined such
as determining a contractor's financial responsibility (Walch & Schmelzer,
2019).� The U.S. Department of Veteran Affairs (VA) believes tasks should be
automated when possible.
Automating things that computers should be doing. So,
there's a lot of things that computers are good at that we don't really need
people to be doing. And so, whenever those situations arise, it's best to
automate them, if we can, so that people can focus on the things that only
people can do. So, one type of use for this type of technology is just to
automate business processes (Fedscoop, 2019).
Most interviewees hoped that streamlined processes and
the automated routine work would be the primary benefits of artificial
intelligence.� "We're doing a lot of robotic tasks.� It's the same thing
over and over again.� Now, we're seeing some processes improve.� [Something]
that could take four hours is really now down to maybe less than an hour or
even minutes.� So there's an efficiency in terms of time" (J. Thong, PowerPoint presentation, March
31, 2020).� Another interviewee made a similar comment.
It gives us the ability to access the terms and conditions
in contracts that we need much more quickly.� Also, we no longer have to spend
a week to review hundreds or thousands of pages in the contract.� We can now
find what we need within a day. So it's a time saver. It's automatic. And it
can be customized and tailored to the terms that
we focus on for our industry. So, that alone is already a major deal for us (S.
Negrete, interview with author, March 26, 2020).
Another anticipated benefit of artificial intelligence is
the gathering of information about a supplier to better a negotiating position.�
Software tools can continuously gather and analyze millions of data indicators,
then use machine learning and predictive analytics to tell you if a specific company
is doing well based on positive or negative views.� They can use that data to
analyze how external data correlates to a company's internal data.� When an
external data value changes, it affects their internal data.� For example, one could
then predict future sales based on past performance or a demographic change, saving
money in a negotiation (J. Thong,
interview with author, April 9, 2020).� The company found that artificial
intelligence helped for spend analytics, predictive analysis, strategy
development, forecasting, digital assistants, requests for proposals, supply
demand matching, contract management, supplier management, and robotic process
automation. (J. Thong, email to author, April 9, 2020).�
Another company found it improved communication and
transparency between regions around the world in a product line.� It let them
quickly run reports to give them better insight into how many contracts were in
a region. �Machine learning models quickly integrated data from enterprise
systems, operational networks, and external sources to generate predictive
insights and find critical issues like start dates, end dates, renewal dates,
force majeure, escalation clause dates, and third party charges like storage or
standby fees.� Their contracts are now readable and extractable, letting them
see the status of the contract and find opportunities for better pricing or
terms while negotiating (S. Negrete, interview with author, March 26, 2020).
HHS also hopes to realize significant savings in money, obtain
insight for negotiations, and do process optimization.� They believe they will
save $2 billion over five years due to these benefits.� For example, better
data access allowed their neural network to perform cluster analysis of over 10
years of contracts, allowing them to find price differences.� This information
will let them buy in bulk and negotiate prices.� Better data also lets them
manage contracts better since they can better measure performance.� For instance,
to ensure compliance with a requirement, they can pull the data from
headquarters, rather than having to ask for a report from each agency.�
Artificial intelligence alerts them when the terms and conditions of a contract
are violated.� Last, a centralized database helps them find steps that can be
combined or eliminated, and enables them to identify and retire overlapping
technology and systems.� (O. Mek, interview with author, March 24, 2020).
HHS found that artificial intelligence improved both
processes and quality.� They use chatbot, a virtual customer service agent, to answer
questions about relevant regulations and clauses and help guide their employees
through the acquisition process. �This improves the quality of an acquisition
plan before it gets to a contracting officer.� It also shortens the process
since it helps to eliminate many questions and discussions between the
requirements generator and procurement organization.� Faster and better insight
was also cited by multiple interviewees.
The other (benefit) is just good insights, so, to help us
go source and save money on some things. Often, it would take us days to
complete an analysis that a robot could do in a mere two minutes now, providing
the same conclusion.� So it gives you more opportunity, more insight to then
generate savings (J. Thong,
interview with author, April 9, 2020).
Combining machine learning with linear regression helped HHS
predict which types of contracts would have problems by looking at past
performance.� Software then gave them recommendations, such as updating a
policy or changing personnel (O. Mek, interview with author, March 24, 2020).
Predication could be an additional benefit of artificial
intelligence.� The software could alert the organization that an event may occur
in the future, such as a demographic change in their future customers, which
gives them time to react and be proactive about it rather than reactive (J. Thong, interview with author, April 9, 2020).� It also
can provide more relevant data to help humans make better decisions.
The other, probably more interesting, use of artificial
intelligence is sort of expanding the boundaries of what a computer can
actually help with. And so, especially in the VA, we have a lot of decisions
that are made all the time by really talented professionals, and if we are able
to increase the amount of support that computers can give to those
professionals as they make these critical decisions -- whether it's a doctor,
you know, treating a patient or somebody processing a disability claim -- we
see a lot of potential to help provide those professionals the exact right
information they need to make the right decision (Fedscoop, 2019)
Last, interviewees mentioned how artificial intelligence
allowed them to focus more on strategy.� It helped them complete transactions more
quickly, with relevant insight and data that forces them to be more strategic (J. Thong, 2020).� "Humans can
then focus on human to human connections" (Fedscoop, 2019).
Streamlining processes and providing better insights should
save both time and money and let employees work on higher-level tasks.� HHS
found that once they had better data and insights, they could optimize and even
eliminate steps, improving speed and agility.� They realized that much data was
just being cut and pasted from one form to another.� Using machine learning, they
simplified a ten-step process into three steps, greatly improving the process
(O. Mek, interview with author, March 24, 2020).� Another interviewee said:
I would say proactive and faster decision-making�at the end
of the day, we're looking to make decisions fast and confidently. And that's
where the speed and agility comes, too, is quicker decision-making. Because we
have data that's telling us to do certain things, and then we're confident with
it, then we can�become more proactive than reactive, helping us make decisions
quicker that are fact based, so that when we look at it, we can all agree that,
hey, this is what it's saying and here's the data backing it up, so let's do it
(J. Thong, interview with author, March 31, 2020).
The General Services Administration is seeing immediate
benefits from being able to pull data from an enterprise data lake and put into
the acquisition forms they build.� "We are already are starting to see
results where you can save a whole lot of time that used to be done by humans,
all the extraction and the import of data, importing it again into another
document" (K. Nakasone, interview with author, March 23, 2020).� He also
said that instead of "spending weeks and weeks and churn over determining
financial responsibility, we know we can use AI and ML to help us with
that".
An organization from industry is seeing a similar benefit.�
"For us, it's the ability to review and organize documents.� Something
that used to take a week can be done in one day.� We can find and extract the
applicable clauses, terms, and conditions quickly.� So, that alone is gold"
(S. Negrete, interview with author, March 26, 2020).
Lastly, some interviewees noted that artificial
intelligence works well with an agile methodology for implementation, the idea
of starting small and developing specific processes side by side with users.� Agile
methods often use microservices, which allow for continuous delivery and
improvement during deployment.� Microservices can be built to represent each of
the planning aspects of the acquisition life cycle (GovernmentCIO Media &
Research, 2020).� Microservices can be designed with artificial-intelligence
methods to make them smarter.
Every interviewee said change management is one of the
biggest challenges with artificial intelligence just as much as with other
innovations.� "There's no innovation without adoption -- if the workforce
doesn't adopt it, you don't have anything� (O. Mek, interview with author,
March 24, 2020). Building and proving that a solution works is not really the
hard part.� Rather, the biggest challenge is getting people to use it and
change the way they conduct business.� Another interviewee said:
It's the change management, because here's the deal.�
You're finding out that these companies are really good at implementing for you
and getting it up and running.� The robot is a robot. �But the change
management really is -- it's not worth anything if the everyday user is not going
in there and interacting with it and using it for its power.� People still want
to do their own thing and may not want to believe that the robot is returning the
right insights (J. Thong, interview with author, April 9, 2020). �
To overcome this challenge, HHS identified human-centered
design as critical.� Users identify bottlenecks or manual processes that can be
automated.� Developers then build a proof of concept and bring it back to the
user for feedback. �This feedback loop occurs every two or three weeks,
refining the solution each time. �Human-centered design gets the user to support
the automation since they were a part of the process throughout.� "If you
win the hearts and minds of the people, you could change, you could lead
change, you could influence others to come aboard -- because in modernization
and innovation, it's not the technology challenge. It's the culture change"
(O. Mek, interview with author, March 24, 2020).� Most people do not yet understand
artificial intelligence and what it can do.� Explaining how artificial
intelligence can help them will not convince them until you can prove it.� It's
essential to give users tangible results with improvements every few weeks.�
Just explaining it won't mean anything.� That's why "human-centered design
is so critical.� That's probably the biggest part".
This is consistent with an industry practice identified
by one of the interviewees called design thinking.� The company uses design
thinking to creatively solve problems and sell the innovative mindset
concurrently.� This not only creates solutions and improves processes but helps
the change management problem.� Moreover, it helped change people's mindsets, improving
ideation, creativity, and collaboration (J. Thong, PowerPoint presentation, March 31, 2020).� Another
industry organization observed that support from leadership is key for
adoption.
Nobody wants to change, right?� People don't like new programs
or new ways of doing things.� You have to have support from above to push it to
more people.� When they see that the boss's boss and everybody is pushing for
it, they'll start to make the effort.� It's a slow process to get people to
jump on board at first, but if you have the support from above to help push it
and be part of the process to improve and customize the program, they start to
see the benefit (S. Negrete, interview with author, March 26, 2020).
Another key practice worth noting is to use training as a
tool for change management.� One interviewee found that tailoring the training
to very specific user groups and teams helped improve feedback and responses.
They also incorporated their artificial intelligence platform into review
processes and meetings, making it a requirement for managers to display
specific information and read their spend analysis (J. Thong, PowerPoint presentation, March 31, 2020).
The director of the Joint Artificial Intelligence Center
believes the three biggest challenges the DoD faces are culture, talent, and
data management (Fedscoop, 2019).� He noted the importance of having
experienced people in the DoD who have worked with artificial intelligence and "have
the breadth and depth of understanding of what AI can do and, more importantly,
[know] how you move it from research and development or pilots, prototypes,
into true operational production that is scaled across the Department of
Defense."� He said that the DoD must have "a culture that begins to
understand the art of the possible in artificial intelligence".
Most of the interviewees cited data acquisition was cited
as another big challenge to using artificial intelligence.� This includes
access to the data, assuring quality of the data, labeling the data, and classifying
the data (M. Arendt, interview with author, March 12, 2020).� Acquiring the
data and getting it into usable format for artificial intelligence will take
time.� One interviewee said the following about data management.
If you accept, as almost everybody that I've talked to
does, that data is that strategic asset, that fuel that really allows you to do
AI�we have a lot of work to do on the data management.� [For instance, for] Project
Maven, we found we were working much longer and much harder on the data
management side of the AI delivery cycle or life cycle than we ever expected
when we first started. �No surprise to anybody in industry, who warned us
that's where you're going to spend more time than you would ever expect�One of
the impediments we face is that, often, to create these new types of models
requires a lot of training data, and the data to train these models is in a lot
of disparate systems. And so trying to figure out how do we get the data into a
place where it can be easily analyzed and then used to train models is a big
challenge. You know, we've got databases that are operational databases. We
have different databases that are more sort of focused on analytic use cases,
and that's spread across a lot of different systems, so we need to find ways to
get that data in a place where the people building these models can actually
use it (Fedscoop, 2019).
Often, organizations want to start using artificial
intelligence or robots without solving the data problem first.� "If you
want to do [artificial intelligence], 80% of it the work will be cleaning the
data, ingesting the data, normalizing the data, and trying to structure and
format the data" (O. Mek, interview with author, March 24, 2020).� Artificial
intelligence can then help organizations understand and organize the data.�
Machine learning, supervised and unsupervised, can look at the data and give an
understanding of what it is.� A model or algorithm is selected and subsequently
trained.� Subject matter experts, like acquisition experts, look at the outcome
and help refine it, which teaches the model or algorithm.� The reinforcement
process occurs over and over, allowing it to become smarter and smarter, which
ultimately normalizes the data.
Several interviewees noted that the success of artificial-intelligence
software depended on the quality of the data.� "If you don't have the
right data in, then you won't have the right data out -- trash in, trash out"
(S. Negrete, interview with author, March 26, 2020).� Therefore, it is
important to have the right processes in the beginning to capture and structure
the data.� A data standard and data governance strategy will save time.� Several
interviewees said they did not have a good process to capture the important data
and had to spend significant time cleaning the data to correct the problem.� Also
a good governance framework for the data is needed that addresses other
considerations like data privacy and data rights (Walch & Schmelzer, 2019).
Security concerns were brought up by one interviewee, in
the form of adversaries manipulating artificial-intelligence software or
training data.� This challenge could be addressed in part by the right language
written into the acquisition (K. Nakasone, interview with author, March 23,
2020). Another interviewee noted also that machine learning could help dissect
and label sensitive data.� Other technologies like blockchains could encrypt
the data and maintain immutable transactional records so that there is a full
record of who did what (O. Mek, interview with author, March 24, 2020).
Another challenge in using artificial intelligence is
finding a contracting process with the necessary flexibility and speed to
develop software that uses artificial intelligence. �Traditional federal
acquisition regulation-based procurement processes do not have these
characteristics.� However, industry can write software, test, and deploy it
with a few weeks, so we need to use its methods.� A second contracting problem
is that the companies that are best at developing artificial intelligence do
not want to work with the government.� Alternative
contracting methods like other transaction authorities or indefinite-delivery/indefinite-quantity
contracts offer a solution, but incentivizing and lowering the barriers to
entry for new companies to do business with the government must be considered (M.
Arendt, interview with author, March 12, 2020).� Third-party contract-management
software could incorporate necessary artificial-intelligence methods, but it
usually needs to be tailored to that industry.� It has taken one company nearly
a year to create a dashboard that is easy to read (S. Negrete, interview with
author, March 26, 2020).� Another potential risk is developer bias in the
program.� A governance program must have good rigor and transparency to certify
ethical behavior without exposing their intellectual property (K. Nakasone,
PowerPoint presentation, March 25, 2020).
Two interviewees thought it was important to start with
innovative and technology-oriented people within the organization to help
improve adoption of artificial intelligence.
The first people we brought in were people excited about
doing it. �Lessons learned is that when you implement something, you always
want to tag certain people that you know that will be good proponents of it.�
Train them first, the superusers -- get them
excited, and then they can train others (J. Thong, interview with author, April 9, 2020).
Another interviewee found that providing the artificial-intelligence
assistant to innovative workers allowed them to individually train the robot
and find new ways to perform their work more efficiently (Fedscoop, 2019).� This
helped encourage adoption and led to more successful outcomes.� Another
interviewee noted that focused training that is specific and relevant to that's
group's procurement function made group members feel like they were being
catered to and encouraged their adoption (J. Thong, interview with author, April 9, 2020).
Another important insight is the importance of the change
agent.� "Find someone who's willing to disrupt the system, that is not willing
to take no for an answer" (Fedscoop, 2019).� The change agent must have support
from above, empowered to make decisions and move things forward very quickly. �Again,
an innovator mindset was cited as critically important.� They need to be able
to see and do things differently.� They have to be "excited and curious
about innovation , technology, and digitization, and not comfortable with the
status quo, but very comfortable with the unknown, working in a loose, chaotic
environment (J. Thong, PowerPoint
presentation, March 31, 2020).� Additionally, the makeup of the team and
rules they operate by was cited as critically important.�
We treated this special project like a startup. We wanted
to be small, nimble, and fast�different years of experience and skill sets, a
blend of millennials and more seasoned members. �However, what was common among
them was that innovator mindset. �Also�none of them had straight-up technical
skills. They represented the 'user community', with a strong desire to simplify
things via technology�(we) wanted to model and exercise one tenet, which is
move fast, make decisions fast, like a startup. �So I want to stress that this
was probably the most important key to our success: choosing the team and to
model it like a startup�we called the team the Digital Lab. �So, why so
important? �I mean what's in a name?� The importance is the perception that is
conveyed. We wanted to convey a picture of a team experimenting and producing
new and great products for the rest of team to use (J. Thong, PowerPoint presentation, March 31, 2020).
One interviewee discussed the makeup of the acquisition
team and thought it important to include with the subject-matter experts on the
team like a data scientist, a business analyst, and an information-technology
specialist.� He believed that having technical people was key to managing the
project and determining whether and where the process needs artificial intelligence
(K. Nakasone, PowerPoint presentation, March 25, 2020).� A related issue is
that the skills required to use some of the artificial intelligence techniques
are relatively new skills that the traditional information-technology staff in
government does not have.� The government has to make sure that they hire some people
with those skills (Fedscoop, 2019).
Several interviewees suggested to start with the problem
and not necessarily the artificial intelligence:
You don't want to start out with artificial intelligence or
blockchain or robotic process automation.� You want to start out with the
business need.� What problem are you trying to solve?� Is it aligning with the
mission of the agency?� You have to look at the mission.� You have to start
with what are you trying to solve before you just start looking at the
technology (O. Mek, interview with author, March 24, 2020).
One interviewee said to work with the users in their
workplace and not develop solutions in a vacuum.� Start by asking users what
their pain points are, develop a solution quickly, and iterate.� The work
cannot be done separate from the users (Fedscoop, 2019).
A big lessoned learned noted by one interviewee was to start
small.� It is important not to chase every single idea that the organization
has.� In fact, it may be better to only try and improve one idea.�
Additionally, the organization has to determine if the process being improved
is feasible, sustainable, and scalable.� Important questions include; will the
budget allow for this in five years and is it marketable?� Again, if people cannot
be convinced that this is a good idea and to buy in, then the organization
should stop development (O. Mek, interview with author, March 24, 2020).
Multiple interviewees stated that the best way to start
is simply to start.� One interviewee made the following observation.
There is no time better than the present to just dive into
AI and get a project started.� Admiring a problem from afar won't do anything
to accelerate the adoption and integration of AI to achieve mission impact at
scale, and that's our mission statement.� So you just have to get your sleeves
rolled up, your hands dirty , and just dive into it.� And you'll quickly find
it's quite hard to get going, but that starting point is let's just figure out
how to do this (Fedscoop, 2019).
Additionally, the solution does not need to be perfect.�
Rather, it needs to be in the hands of the warfighter or user as soon as
possible.� What is more important is the feedback loop.� It has to be easy for
the user to tell you what they need and how think it could be better.� The
solution must have a continuous development and feedback loop to improve it.�
The interviewee identified this as
"deployment, adoption, and refinement".� He believes it to be critical
for success that, a problem that not everyone is thinking about (M.
Arendt, interview with author, March 12, 2020).
Several interviewees noted that artificial intelligence
takes time to set up and time to refine.� "AI is neither a silver bullet nor an easy button ready to be
pushed.� So, in looking at whatever problem a supported unit, or customer,
wants us to help them with, the question is�is AI the best solution. �If it is,
then we dive down deep in the details" (Fedscoop, 2019).�
Table 1 summarizes
the findings from this chapter.
Table 1.
Summary of secondary
questions and learnings
|
Questions
|
Learnings
|
|
Why companies adopted artificial intelligence
|
�
Improve employee work experience
�
Smarter strategic decision-making
�
Better deals through insights and
analysis
�
Uncover insights and
inconsistencies in contracting
�
Eliminate inconsistencies in
contracting
�
Better dissemination of
information
�
Speed procurement process
�
Retire old systems and upgrade
technology
�
Automate low-level tasks
|
|
Benefits
|
�
Streamline processes and
automate routine work
�
Save staff time and eliminate
tedious work
�
Transparency between regions,
customers and product lines
�
Ensure compliance with contract
terms and conditions
�
Improve negotiating position
�
Save money through better
negotiation, process, insight, etc.
�
Improved procurement processes
�
Identify and predict contract
problems and suggest improvements
�
Support professionals to allow
their focus on important decisions
�
Allows for strategic focus
|
|
Speed and Agility
|
�
Faster and improved
decision-making
�
Allow employees to focus on
higher-level tasks
�
Greater speed and agility in
multiple applications
�
Flexible and independent
implementation
|
|
Risk and Challenges
|
�
Change management
�
Data acquisition and quality
�
Premature rollout before data is
ready
�
Government workforce culture and
skills
�
Cybersecurity
�
Contracting pathway for
artificial intelligence
�
Resistance of companies to
contract with government
�
Tailoring third-party software
�
Developer bias in software
�
Lack of buy-in
|
|
Lessons Learned
|
�
Choose best change agent and
like-minded team to lead artificial intelligence implementation
�
Provide specific & relevant
training to encourage artificial intelligence implementation
�
Include users throughout
development
�
Properly sequence implementation
� data first
�
Start with the mission and �pain
points� first
�
Start small, nimble, and fast
�
Accept �good enough� in the
beginning
�
Encourage a robust feedback loop
�
Start now
|
In this chapter, I will briefly discuss the results of
the interviews, explore areas for potential applications of artificial
intelligence in procurement for the DoD, and provide a potential roadmap for
implementation.
The benefits of artificial intelligence anticipated by
interviewees were significant.� They thought that it could save significant
time and money, improve speed and agility, streamline processes, and automate
tasks.� It could provide new insights from a large amount of data, improving
decision-making and policies.� It could save time to allow procurement professionals
to work on tasks not adequately addressed today like contract management, which
is often improperly performed due to time and capacity constraints.� Similar
tasks that could be improved are network management and proactively planning
for emergencies; more time could permit identifying contingency contractors and
putting advanced contracts in place.� Lastly, these improvements could save
significant money and improve the likelihood that programs meet schedule and
performance expectations, allowing the Defense Acquisition System (DAS) to
address some long-standing challenges and criticisms.
Table 2 lists potential opportunities for artificial
intelligence and related software concepts in procurement.� It is organized by
procurement phase or competency along with a corresponding procurement
activity.� While the list is fairly long, �improving just one of these could
save the DoD substantial time and money.� For instance, if artificial intelligence
could analyze large amounts of data about a company's contracts, the DoD could
negotiate better prices.� And if cost analytics could organize and classify
cost data to increase transparency and compatibility of spending across an
organization, that could lead to significant savings.
Table 2.
Potential applications of the artificial-intelligence complex to
procurement
|
PROCUREMENT LIFE CYCLE
PHASE OR COMPETENCY
|
PROCUREMENT ACTIVITY
|
POTENTIAL APPLICATIONS
|
|
MANAGEMENT COMPETENCY
|
�
Governance
� Planning and Strategy
� Cost Analytics
� Requirement Compliance
� Risk Management
� Root Cause Analysis
� FAR and Regulatory Compliance
|
� Provide data insight across the
enterprise
� Identify opportunities for new policy
� Automate classification by cost
categories
� Streamline or eliminate reporting
requirements
� Examine contracts and identify risks
� Identify redundancies to be eliminated
�
Generate applicable
clauses automatically
|
|
PRE-AWARD LIFE CYCLE
|
�
Acquisition Planning
� Market Research
� Contracting Strategy
� Offer Validation and Evaluation
� Communication
� Risk and Fraud Analysis
� Terms and Conditions Preparation
� Solicitation Documents Preparation
|
� Allow focus on strategic tasks
� Analyze pricing of potential sources
� Provide insight to develop strategy
� Determine if a bid meets the minimum
criteria
� Allow time for network management
� Determine contractor�s financial responsibility
� Pre-populate applicable terms and
conditions
�
Pre-populate forms
with applicable language
|
|
AWARD LIFE CYCLE
|
�
Cost or Pricing
Analysis
� Negotiation Preparation & Strategy
� Negotiation Terms and Conditions
� Negotiations
� Proposal Evaluation
� Evaluation Criteria Application
� Invoice Automation
� Spend Classification
� Contract Analytics
|
� Identify opportunities to negotiate
better rates
� Gather and analyze data, e.g. future
sales
� Provide insight into past performance
� Solve complex negotiation problems
� Evaluate bids and manage trade-offs
� Perform more accurate analysis without
bias
� Automate payment if a certain condition
is met
� Automate classification by cost category
�
Improve insight and
analysis
|
|
POST AWARD LIFE CYCLE
|
�
Contract Administration
� Contract Management
� Terms & Conditions Compliance
|
� Track and measure contractor performance
� Alert management if a condition/price
change
�
Alert management of
fee and renewal dates
|
Implementing artificial intelligence into procurement
activities depends on access to quality data that is structured and labeled
well, and solving the change management problem.� The technology to do this is not
difficult; as discussed earlier, a growing number of organizations can support these
tasks.� However, artificial intelligence will not provide sufficient utility if
the database problem is not solved too;.� �the more data, the better artificial
intelligence works.� The DoD possesses oceans of data, but it is fragmented on
legacy systems, local computer drives, and file cabinets.� The key is to collect
the data, clean it, label it, and put it into distributed database systems for
the future.� A data governance plan will ensure this is ongoing. to bring in
the right data in a timely way.
Even with the right technologies, another challenge will be
getting people to adopt the new solution.� "Human-centered design" or
"design thinking" should be used to ensure that users participate in the
development of the solution from the outset.� They must help identify the tasks
that must be improved.� Prototypes should be built quickly and taken back to
them for their feedback.� It should be improved based on their input and then
brought back to them.�
The issue of whether artificial intelligence will replace
people often discourages adoption of artificial intelligence by the workforce.�
However, often artificial intelligence is a tool or a virtual co-worker that
can help an employee complete their work more efficiently.� This could certainly
be true for procurement. At this time anyway, full procurement process
automation without human input is impossible.� Communicating this early and
often to the workforce will ease fears associated with emerging technologies.
Interviewees identified other risks and challenges to
implementing the artificial-intelligence complex like cybersecurity, developer
biases, and data rights; other problems, like the Procurement Integrity Act,
remain.� These challenges can be solved because� Health and Human Services has
solve similar problems..� �HHS is like the DoD in many ways; they have many of the
same rules and face the same challenges, and they have successfully awarded contracts
using artificial intelligence. Another problem is that organizations that are
best at developing artificial intelligence do not want to work with the DoD.�
The DoD must lower the barriers to entry and properly incentivize industry with
these technologies.
To implement new technologies like artificial
intelligence in the DoD, the Defense Acquisition System (DAS) should treat it
differently from other software.� Traditional procurement pathways, including
the newly released Adaptive Acquisition Framework, cannot address the speed at
which new technologies like artificial intelligence need to be implemented and
refined since algorithms and software can be programmed and improved within
weeks, if not days or hours.� New technologies like artificial intelligence typically
require a �spiral path� development involving constant refinement of the
solution with a continuous feedback loop.�
Figure 3 illustrates our ideas.�
It also addresses two significant challenges facing the DoD, change management
and data.� However, the champion identified should not simply be a
program manager and the team identified should not merely be an acquisition
team.�
Figure 3.
Potential pathway for artificial intelligence
Leadership must make this approach a top priority.� They
must allocate the necessary resources now to avoid delays. �Emerging technologies
like artificial intelligence are becoming more commonplace and essential to
organizations, and the government is falling behind.� Moreover, starting small can
work well; HHS began with a proof of concept that cost them $175,000 and took
them four weeks to build (Nyczepir, 2019).
A senior leaders council should develop a data strategy.�
It should standardize capturing and labeling data across the enterprise so it
can be exploited by emerging technologies like artificial intelligence.� �Next,
the council should designate a champion to lead the initiative who will be
innovation-minded.� They must be empowered to make decisions and allocate
resources. �They should select a team of innovation-minded people who think
differently and have a strong desire to simplify processes with technology They
should have varying backgrounds ; most should be
users, but some should have some technical ability to monitor the performance
of contractors.� They should make decisions quickly by learning, failing, and adapting.�
Last, the team should communicate their initiatives to their organization to
ease the fears associated with emerging technology and broadcast that an
innovative team is creating exciting new products for the sponsoring organization
to use.
The team should begin with a procurement framework that
is commonly used by the sponsoring organization to make it easier and familiar when
solutions are piloted, tested, and ultimately scaled.� Next, it should identify
�time-consuming manual processes that procurement professionals struggle to
complete.� It should work with the artificial-intelligence and software experts
and bring companies in to show their platforms and products.� If a platform of
product looks promising, let them to create a proof of concept with test data.�
If it works, scale it and keep testing; if it doesn't, try something else.� The
idea is to allow people to see immediate benefits and applications.� .
This thesis discussed applying artificial intelligence to
procurement within the Defense Acquisition System (DAS).� Our interviews
suggested that artificial intelligence, properly implemented, could help the
acquisition process in ways large and small.� It could help people complete
their work more quickly and accurately, help market research, assist in negotiations,
give insights, and aid contract management.� It could change the DoD�s business
strategies and thinking.
Our primary research question was how the DoD could adopt
commercial-style artificial intelligence for procurement.� The previous chapter
outlined a plan incorporating industry practices.� It addressed the two biggest
DoD challenges when implementing emerging technologies, data integrity and
change management.� It also emphasizes the importance of a quick feedback loop during
software development. We make the following
recommendations for implementation the artificial-intelligence complex of
concepts in procurement:
�
Make it a priority.� Procurement is full of manual and
repetitive tasks that can be helped by technologies like artificial
intelligence.� Further research on what to attack is not needed; a team can
learn as they work through each issue.� Rather, organizational commitment and the
will to innovate is more important.
�
Start with data.� Data is the fuel for artificial intelligence,
but working out processes to get the data will take much time.� Do not start
implementation until the data problem is solved, because otherwise cleaning and
labeling the data will take much more time.
�
Treat it differently than traditional software.� Artificial
intelligence requires procedures that the Defense Acquisition System has yet to
develop.� It must be fast and agile; it must be spiral and not linear; and the
user must be at the center of development and improvement.
�
Start small.� First identify small problems and develop small
solutions at the organizational level.� If successful, scale it.
�
Monitor similar projects at other government agencies.� Study
and follow the progress of Health and Human Services.� The Accelerate platform
should continue to be monitored since they are a potential model the DoD could
follow.� Also, use the General Services Administration as a consultant.
There is no better time to begin implementing the artificial-intelligence
complex into the Defense Acquisition System.� Agencies have been directed to begin
adoption, and industry investment and support for procurement is growing.�
Organizations that do not adopt these methods soon will face increased competition
for resources and talent, and intelligence will not achieve the speed necessary
to maintain their competitive advantage and relevance.� This means that the
Defense Acquisition System must modernize if it is to achieve its top priority of
"enabl[ing] innovative acquisition approaches that deliver warfighting
capability at the speed of relevance" (OUSD (A&S), n.d.).�
Former Secretary of Defense Jim Mattis captured the essence
of why action today is essential:
We will modernize key capabilities, recognizing we cannot
expect success fighting tomorrow's conflicts with yesterday's weapons or
equipment -- To keep pace with our times, the department will transition to a
culture of performance and affordability that operates at the speed of
relevance.� Success does not go to the country that develops a new technology
first, but rather, to the one that better integrates it and more swiftly adapts
its way of fighting.� Our current bureaucratic processes are insufficiently
responsive to the department's needs for new equipment. We will prioritize
speed of delivery, continuous adaptation and frequent modular upgrades.� We
must shed outdated management and acquisition practices, while adopting
American industry's best practices. Our management structure and process are
not engraved in stone.� They are a means to an end, empowering our warfighters
with the knowledge, equipment and support needed to fight and win.� Failure to
modernize our military risks leaving us with a force that could dominate the
last war, but be irrelevant to tomorrow's security (Mattis, 2018).
Agrawal, A., Gans, J., & Goldfarb, A. (2018). Prediction
Machines: The Simple Economics of Artificial Intelligence. Harvard Business
Review Press.
Arquilla, J. & Denning, P. (2019). Automation
will Change Sea Power. The Denning Institute. http://denninginstitute.com/pjd/PUBS/usni-seapower-2019.pdf
Barnett, Jackson (2019, June 13). The 2020 NDAA Takes
Up a Common Cause: Acquisition Reform. Fedscoop.
https://www.fedscoop.com/ndaa-acquisition-reforms-dod/
Brown, Michael (2019, March 12). Statement by Michael
Brown Director of the Defense Innovation Unit Before the Senate Armed Services
Committee Subcommittee on Emerging Threats and Capabilities on �Artificial
Intelligence Initiatives within the Defense Innovation Unit�.
https://www.armed-services.senate.gov/imo/media/doc/Brown_03-12-19.pdf
Carnegie Mellon University. (2019, July 11). AI beats
professionals in six-player poker. ScienceDaily. https://www.sciencedaily.com/releases/2019/07/190711141343.htm
Castellanos, Sara. (2020, February 13). Shell Aims to
Enroll Thousands in Online Artificial-Intelligence Training. The Wall
Street Journal.
https://www.wsj.com/articles/shell-enrolls-thousands-in-online-artificial-intelligence-training-11581622566
CB Insights. (2019, September 17). The Race For AI:
Here Are the Tech Giants Rushing to Snap Up Artificial Intelligence Startups.
https://www.cbinsights.com/research/top-acquirers-ai-startups-ma-timeline/
Chaffin, Eric. (2018, October 2). Will Artificial
Intelligence Eventually Nix Lawyer Jobs? The Legal Examiner.
https://newyork.legalexaminer.com/technology/latest-technology/will-artificial-intelligence-eventually-nix-lawyer-jobs/
China State Council (2017, July 20). A Next
Generation Artificial Intelligence Development Plan. Translated by New
America. https://www.newamerica.org/documents/1959/translation-fulltext-8.1.17.pdf
Chopra, Ashok. AI in Supply & Procurement. 2019
Amity International Conference on Artificial Intelligence (AICAI). Dubai,
United Arab Emirates. 2019. pp. 308-316, DOI: 10.1109/AICAI.2019.8701357
Cognilytica. (2020, January 22). Global AI Adoption
Trends & Forecast 2020: Patterns of Worldwide Adoption of AI.
https://www.cognilytica.com/2020/01/22/global-ai-adoption-trends-forecast-2020/
Daquila, Marianne. (2019, September 4). Worldwide
Spending on Artificial Intelligence Systems Will Be Nearly $98 Billion in 2023.
IDC. https://www.idc.com/getdoc.jsp?containerId=prUS45481219
Defense Innovation Board. (2019). Software Is Never
Done: Refactoring the Acquisition Code for Competitive Advantage. https://media.defense.gov/2019/Mar/26/2002105909/-1/-1/0/SWAP.REPORT_MAIN.BODY.3.21.19.PDF
Department of Defense. (2018).� National Defense
Strategy. https://dod.defense.gov/Portals/1/Documents/pubs/2018-National-Defense-Strategy-Summary.pdf
Department of Defense. (2019). Summary of the 2018
Department of Defense Artificial Intelligence Strategy: Harnessing AI to
Advance Our Security and Prosperity. https://media.defense.gov/2019/Feb/12/2002088963/-1/-1/1/SUMMARY-OF-DOD-AI-STRATEGY.PDF
Department of Defense Office of the Inspector General.
(2019). Top DoD Management Challenges � Fiscal Year 2020.
https://media.defense.gov/2020/Mar/11/2002263093/-1/-1/1/TOP%20DOD%20MANAGEMENT%20CHALLENGES%20FISCAL%20YEAR%202020.PDF
Defense Advanced Research Projects Agency. (2018). DARPA
Defense Advanced Research Projects Agency 1958�2018. https://www.darpa.mil/attachments/DARAPA60_publication-no-ads.pdf
Department of Defense. (2019). DoD Digital Modernization
Strategy.
https://media.defense.gov/2019/Jul/12/2002156622/-1/-1/1/DOD-DIGITAL-MODERNIZATION-STRATEGY-2019.PDF
Defense Advanced Research Projects Agency (2018). AI
Next Campaign. https://www.darpa.mil/work-with-us/ai-next-campaign
Docherty, N. & Fanning, D. (2019). In the Age of
AI. Frontline. https://www.youtube.com/watch?v=5dZ_lvDgevk
D'Onfro, Jillian.� (2019, September 17). AI 50:
America�s Most Promising Artificial Intelligence Companies. Forbes. https://www.forbes.com/sites/jilliandonfro/2019/09/17/ai-50-americas-most-promising-artificial-intelligence-companies/#588d8a92565c
Echelon/RPA. (n.d.). Empowering the Next Generation
Workforce Today. http://rapidrpa.com/company/
Ehn, Eric. (2017). Artificial Intelligence: The Bumpy
Path Through Defense Acquisition [Master�s thesis, Naval Postgraduate
School]. NPS Archive: Calhoun. https://calhoun.nps.edu/handle/10945/56910
Fedscoop. (2019, April 16). AI Innovators:
Accenture's Dominic Delmolino. [Video]. https://www.youtube.com/watch?v=z9BRXaGHfBs&feature=emb_title
Fedscoop. (2019, April 29). AI Innovators: HHS'
Kristen Honey. [Video]. https://www.youtube.com/watch?v=CImbhCMVK2w
Fedscoop. (2019, May 28). AI Innovators: DoD's Lt.
Gen. Jack Shanahan. [Video]. https://www.youtube.com/watch?v=0K-lvpRwjys&feature=emb_title
Fedscoop. (2019, May 20). AI Innovators: General
Service Administration's Keith Nakasone. [Video]. https://www.youtube.com/watch?v=xD4HPQ4S8Cw&feature=emb_title
Fedscoop. (2019, May 13). AI Innovators: Dept. of
Veterans Affairs' Charles Worthington. [Video]. https://www.youtube.com/watch?v=VGg8RziocQg&feature=emb_title
Government Accountability Office. (2019). High-Risk
Series: Substantial Efforts Needed to Achieve Greater Progress on High-Risk Areas
(GAO-19-157SP).
GovernmentCIO Media & Research, (2020, February 13).
CIO Highlight: Jose Arrieta, HHS. [Video].
https://www.youtube.com/watch?v=XdFt6zkeCPI&t=4s
Hill, Catey. (2019, December 28). The No. 1 job of
2019 pays $140,000 � and its hiring growth has exploded 74%. MarketWatch.
https://www.marketwatch.com/story/the-no-1-job-of-2019-pays-140000-and-its-hiring-growth-has-exploded-74-2019-12-10
Klabi, H., Mellouli, S., & Rekik, M. (2013). A
Reputation Based Electronic Government Procurement Model. ICEGOV '13:
Proceedings of the 7th International Conference on Theory and Practice of
Electronic Governance. October 2013 Pages 94�102. https://doi.org/10.1145/2591888.2591904
Kruppenbacher, T.A. (1984). The Application of
Artificial Intelligence to Contract Management. [Master�s thesis, U.S. Army
Construction Engineering Research Laboratory]. Defense Technical Information
Center.
Lord, Ellen (2019, December 10). Under Secretary of
Defense for Acquisition & Sustainment Ellen Lord Press Briefing on Defense
Acquisition. https://www.defense.gov/Newsroom/Transcripts/Transcript/Article/2037206/under-secretary-of-defense-for-acquisition-sustainment-ellen-lord-press-briefin/
Loudin, Amanda (2019, October 30). Novak Djokovic
Used A.I. to Train for Wimbledon. OneZero.
https://onezero.medium.com/novak-djokovic-used-a-i-to-train-for-wimbledon-e22b62f7504a
Malone, Katie. (2020, February 11). Blockchain Saving
HHS $30M on First Accelerate Contract. Meritalk.
https://www.meritalk.com/articles/blockchain-saving-hhs-30m-on-first-accelerate-contract/
Nakasone, Keith. (2018, March 7). Statement of Keith
Nakasone Deputy Assistant Commissioner, Acquisition Operations, Office of
Information Technology Category, U.S. General Services Administration Before
the Subcommittee on Information Technology of the Committee on Oversight and
Government Reform. Game Changers: Artificial Intelligence Part II; Artificial
Intelligence and the Federal Government. https://republicans-oversight.house.gov/wp-content/uploads/2018/03/Nakasone-GSA-Statement-AI-II-3-7.pdfNational
Academy of Medicine & Government Accountability Office. (2019). Artificial
Intelligence in Health Care (GAO-20-215SP).
National Security Commission on Artificial Intelligence.
(2019). Interim Report November 2019. https://drive.google.com/a/nscai.org/file/d/153OrxnuGEjsUvlxWsFYauslwNeCEkvUb/view?usp=sharing
Nyczepir, Dave. (2019, Sep 17). HHS
touts first recurrent neural network for making acquisition plans. Fedscoop.
https://www.fedscoop.com/hhs-recurrent-neural-network/
Office of Science and Technology Policy. (2020). American
Artificial Intelligence Initiative: Year One Annual Report. https://www.whitehouse.gov/wp-content/uploads/2020/02/American-AI-Initiative-One-Year-Annual-Report.pdf
PricewaterhouseCoopers. (2017). What�s the real value
of AI for your business and how can you capitalize?
https://www.pwc.com/gx/en/issues/analytics/assets/pwc-ai-analysis-sizing-the-prize-report.pdf
Mattis, James. (2018, January 19). Statement by Secretary
of Defense James N. Mattis on the National Defense Strategy. https://www.defense.gov/Newsroom/Transcripts/Transcript/Article/1420042/remarks-by-secretary-mattis-on-the-national-defense-strategy/
Marlinghaus, Sven. (2018). AI and the Future of
Procurement. Roland Berger. https://www.rolandberger.com/publications/publication_pdf/roland_berger_ai_rpa_in_procurement.pdf
Office of the Under Secretary of Defense for Acquisition
and Sustainment (2020, January 23). Operation of the Adaptive Acquisition
Framework (DoDI 5000.02). Department of Defense.
https://www.esd.whs.mil/Portals/54/Documents/DD/issuances/dodi/500002p.pdf?ver=2020-01-23-144114-093
Office of the Under Secretary of Defense for Acquisition
and Sustainment. (n.d.). https://www.acq.osd.mil/fo/acq/docs/as-road-map-new.pdf
Prevedere. (n.d.). Overview. https://www.prevedere.com/about/overview/
RAND Corporation. (2019). The Department of Defense
Posture for Artificial Intelligence: Assessment and Recommendations. https://www.rand.org/pubs/research_reports/RR4229.html
Rodriguez-Aguilar, J., Giovanucci A.,
Reyes-Moro A., Noria F. & Cerquides J.. Agent-based Decision Support
for Actual-World Procurement Scenarios. IEEE/WIC International Conference
on Intelligent Agent Technology. IAT 2003. Halifax, NS. �pp. 371-377. DOI:
10.1109/IAT.2003.1241102
Seal Software. (n.d.). About Seal: Quick Facts. https://www.seal-software.com/about/
Sievo. (n.d.). Intelligent Spend Management Software.
https://sievo.com/solutions/spend-management
Shanahan, Jack. (2019, September 11). A Game Changer
and Decisive Edge. https://www.ai.mil/blog_09_11_19.html
Simonite, Tom (2017, August 8). For Superpowers,
Artificial Intelligence Fuels New Global Arms Race. Wired.
https://www.wired.com/story/for-superpowers-artificial-intelligence-fuels-new-global-arms-race/
Subramanian K., Lee S., Shiang, T. & Sue, G. Intelligent
Agent Platform for Procurement. IEEE SMC'99 Conference Proceedings. 1999
IEEE International Conference on Systems, Man, and Cybernetics (Cat.
No.99CH37028). Tokyo, Japan. (1999). pp. 107-112 vol.3. DOI:
10.1109/ICSMC.1999.823163
Torres, Jordi. (2018). Why now this Artificial
Intelligence boom? Towards Data Science. https://towardsdatascience.com/why-now-this-artificial-intelligence-boom-b50a35713090
Umbenhauer, B., Flynn, R., & Mitchell, P. (2019). The
Deloitte Global CPO survey 2019. Deloitte Insights. https://www2.deloitte.com/content/dam/Deloitte/nl/Documents/strategy/deloitte-nl-so-cpo-survey-2019-pdf.pdf
Walch, K. & Schmelzer R. (Hosts). (2019, November
20). AI Today Podcast # 116: AI and Procurement in the Federal Government:
Keith Nakasone, GSA. [Audio podcast episode]. Cognilytica. https://www.cognilytica.com/2019/11/20/ai-today-podcast-116-ai-and-procurement-in-the-federal-government-keith-nakasone-gsa/
Walch, K. & Schmelzer R. (Hosts). (2018, July 11). AI
Today Podcast # 45: AI and Blockchain in Government: Interview with Jose
Arrieta, US Department of Health & Human Services . [Audio podcast
episode]. Cognilytica. https://www.cognilytica.com/2018/07/11/ai-today-podcast-45-ai-and-blockchain-in-government-interview-with-jose-arrieta-us-department-of-health-human-services-hhs/
Webb, Kevin. (2019, November 27). A former world
champion of the game Go says he's retiring because AI is so strong: 'Even if I
become the No. 1, there is an entity that cannot be defeated. Business
Insider. https://www.businessinsider.com/deep-mind-alphago-ai-lee-sedol-south-korea-go-2019-11
Weinbaum, C. & Shanahan, J. (2018). Intelligence in
a Data-Driven Age. Joint Force Quarterly 90, 3rd Quarter 2018, 4-9.
https://ndupress.ndu.edu/Portals/68/Documents/jfq/jfq-90/jfq-90_4-9_Weinbaum-Shanahan.pdf?ver=2018-04-11-125441-307
White House. (2017). National Security Strategy.
https://www.whitehouse.gov/wp-content/uploads/2017/12/NSS-Final-12-18-2017-0905.pdf
White House (2019). Exec. Order No. 13859, 3 C.F.R 3967.
https://www.govinfo.gov/content/pkg/FR-2019-02-14/pdf/2019-02544.pdf
United States Congress.� The John S. McCain National
Defense Authorization Act for Fiscal Year 2019, Pub. L. No. 115-232, 132 Stat.
1636, (2018). https://www.congress.gov/115/plaws/publ232/PLAW-115publ232.pdf
