A Case for Artificial Intelligence in Future Procurement CHRIS ROBEY, U.S. CUSTOMS AND BORDER PROTECTION, U.S. DHS The views expressed in this presentation are those of the author and do not reflect the policy or position of U.S. Customs and Border Protection, U.S. Department of Homeland Security, or the U.S. Government.
17
Embed
A Case for Artificial Intelligence in Future Procurement€¦ · A Case for Artificial Intelligence in Future Procurement ... business-to-business (B2B) ecommerce: ... “Big Data
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
A Case for Artificial Intelligence in
Future Procurement
CHRIS ROBEY, U.S. CUSTOMS AND BORDER PROTECTION, U.S. DHS
The views expressed in this presentation are those of the author and do not reflect the policy or position of U.S. Customs and Border Protection, U.S. Department of Homeland Security, or the U.S. Government.
Agenda
State of the art
What is a business agent?
AI Business Agent Typology – Welcome to the zoo!
AI Business Agents in Negotiations – Who do you trust?
AI Business Agents in Audits – Are you ready?
Conclusion
Sources
Q&A
2
State of the Art - Thresholds
Bots and algorithms using structured data?
Turing Test?
General intelligence AI using unstructured data?
Superhuman intelligence AI? 20 years out, per Bostrom
Opinion: Agnostic whether AI can achieve consciousness, will, or intent, as understood within human experience:
(h)owever, goal-directed and adaptive strategic behavior by decision support systems with domain expertise is very much within the current state of the art of cognitive computing.1
1 Sue Feldman and Hadley Reynolds, “Cognitive computing: A definition and some thoughts,” KW World, November/December 2014, ref. http://www.kmworld.com/Articles/News/News-Analysis/Cognitive-computing-A-definition-and-some-thoughts-99956.aspx
A software entity that acts autonomously, that is, makes its own decisions on behalf of the designer, typically in dynamic environments from which it learns and to which it adapts.
When applied to the development of new internet technologies, agents need also to show a social attitude.1
Which leads us to consider:
Multi-agent system (MAS)
A collection of autonomous agents that need to coordinate their activities in order to achieve their individual goals.
Coordination is achieved through negotiation or argumentation and, in most applications, requires that the agents learn to adapt to each other's strategies.2
1,2 Keith Frankish and William Ramsey, ed., The Cambridge Handbook of Artificial Intelligence, Cambridge University Press, 2014, Pages 335, 339
4
AI Business Agent Typology
Current SOA
In operation
SCM and logistics/
transportation/
scheduling algorithms
Retail and B2B
transactions
Can negotiate using
structured data
Will take off with ISO
standards and NIST
policy (both TBD)
BA advisor
Emerging SOA
Aid to human operators in negotiation and maintain institutional knowledge
Natural-language, real time query-and-response access to enterprise-wide
Finance
Procurement
Operations
databases delivered securely as a cloud service
As aid through differing Internet of Things (IoT) standards/protocols/ platforms for SCM
AI-enabled BA’s
Future SOA
General-purpose AI in
autonomous BA’s
Unstructured data
Capable of negotiation
within multi-agent
environment
Legal status of contracts
formed – uncertain (?)
In labs – 5 years out (?)
Competition will drive
technical progress
5
Bots & web crawlers
AI Business Agents in Negotiations Disruption to processes
Business capture
Execution/operations
Audit
Why?
Compelling economies
Competitive pressure
Scalability
Applicability to business processes
Unforeseen consequences
Evolving models of trust need for governance
6
Models of trust in negotiations
1.0 Human/legal
Counterparty:
Has authority as
agent to represent
his/her principal
Will provide
accurate/truthful
responses to factual
queries, i.e., no
fraudulent intent
Will carry out the
bargain upon
agreement
2.0 E-commerce
Both consumer and
business-to-business
(B2B) protocols:
Availability
Visibility
Security
Transaction
execution
New questions?
Emergence of cloud-
based, business-to-
business (B2B)
ecommerce:
Do you trust the data
structures of your
transaction partners?
How will certification/
accreditation processes
evolve with the threat
environment?
7
Models of trust in negotiations
E. Alonzo on social rationality among BA:
Sincerity: No agent will attempt to have another believe a proposition that it either knows or believes to be false or a proposition that it wants to be false
e.g., agents cannot commit themselves to execute actions that they are not able to perform.
Honesty: Agents have to act according to their beliefs.
Fair play: Agents must abide by the agreed deals.
Sociability: In case of indifference, agents must accept others' offers, and deals must always be individually rational.1
1 Eduardo Alonzo, “Actions and Agents,” from The Cambridge Handbook of Artificial Intelligence, Cambridge University Press, 2014, Page 240.
8
3.0 AI-enabled business agents (in multi-agent systems)
Who do you trust? 9
BA (Irene) and human (Brian) in mock
negotiations for IT hardware
Academic experiment to explore how agent
behaviors affect success/failure of negotiations
with human subjects segmented by personality
type
From IEEE Intelligent Systems, March/April 2014, pps. 36 - 43; Figure 1 used with permission of IEEE.
Social Engineering Big Data Analytics
Descriptive
Predictive
Prescriptive
Intention analysis vs. sentiment analysis in user interaction with
websites1
High-volume retail websites will test new consumer-facing algorithms in
competition against each other, while using metrics of user behavior to
determine which prototype to scale up across the website
B2B traffic will not be exempt from the lessons learned.
Watch for “Moneyball”-style metrics on individual human negotiators on
the other side, i.e., characteristics and negotiating style, to be gathered
by AI BA’s for sharing with other entities in the enterprise1Jeff Bertolucci, “Big Data Tool Analyzes Intentions: Cool or Creepy?” Information Week, 12/15/2014, ref. http://www.informationweek.com/big-data/big-data-analytics/big-data-tool-analyzes-intentions-cool-or-creepy/d/d-id/1318128??itc=edit_in_body_cross%20
To evaluate the integrity of internal controls which support
business decisions and governance
Types of audit
Financial statement – qualified vs. “clean”
IT – cybersecurity
Forensic – suspicion of criminal intent/activity
Audit standards are changing
Corporate – GAAP standard to COSO Framework
More structured evaluation process, driven by Sarbanes-Oxley
Government – Convergence to COSO Framework through
GAO endorsement
11
AI Business Agents in Audits COSO Framework
Five Components of Internal Control (on front)
Committee of Sponsoring Organizations of the Treadway Commission(1980’s)
Voluntary private sector initiative
Endorsed by U.S. Government for agencies
Many variations
Highly amenable to AI treatment
12
From U.S. Government Accountability Office, Standards for
Internal Control in the Federal Government, GAO-14-704G,
Published: Sep 10, 2014
AI Business Agents in Audits Beyond data mining
To recognize the difference between relevant and irrelevant data correlation
Phase 1 – BA as instrumentality to aid audit
Phase 2 – BA as intermediary/decision maker, in business process being audited
Can a BA be trained to deceive?
Asimov’s Laws. . . What then?
Liability issues and dispute resolution
The traceability of the machine recommendations (i.e., why a recommendation was made) will be important in fostering confidence and trust.1
1 Dr. Francesca Rossi, Professor of Computer Science, University of Padova and Harvard University, Your cognitive future: How next-gen computing changes the way we live and work, IBM Institute for Business Value, 2015; http://www-935.ibm.com/services/us/gbs/thoughtleadership/cognitivefuture/