Artificial Intelligence Will Continue to Transform Modern Credit Management:

Can We Handle It? Anthony J. Scriffignano, Ph.D. SVP/Chief Data Scientist Dun & Bradstreet @Scriffignano1 A little context before we get started…

Are we using our technology to make things better, or merely to make things different?

Is life getting easier, better, or just more complicated

Are we losing touch with our technology? 2 CURIOUS TIMES A REFLECTION ON WHAT IS HAPPENING IN THE WORLD OF ALL THINGS DATA AND AI TODAY

MAKING OLD DECISIONS IN NEW WAYS HOW WE ARE ABLE TO USE MODERN DATA AND TECHNOLOGY TO ANSWER NEW QUESTIONS IN THE TODAY CONTEXT OF BUSINESS DECISIONS

LOOKING AHEAD A LOOK AT SOME OF THE FUTURE DISRUPTIONS IN TECHNOLOGY AND WORKFORCE AND SOME ADVICE ON HOW TO GET READY

3 Getting Started: reflecting on credit decisions, now and then…

How much has the fundamental business problem changed? How is it likely to change with new data and capabilities?

4 It is possible that we no longer truly understand how information is created and consumed.

Not too long ago…

Reading about events that happened in the past, listening to people who are not present.

Reading about things in the “now” – can’t tell who is communicating with whom – Today… does anybody really know what is going on?

5 Why do we believe what we believe?

Too much data? Different objectives?

6 How data is being discovered and used is a constantly-changing environment with changing context.

Situational awareness… Unstructured • Far more data is being Data created than used • New types of business beget new types of data, which begets new methods • Commonly available tools and solutions only address a small part of this space

7 Myths / Inconvenient Truths:

More Data is better Data vs. noise

Data at rest vs. Put data in 1 place data in motion

AI can find answers AI methods have preconditions

Machine learning will Regression vs. find hidden truth unprecedented change

Natural language processing removes all Language is constantly language barriers changing

Real‐time analytics Data collected a second are best ago is not 1‐second old! Mature industries, such as financial services, and evolving industries, such as FinTech, are all disrupted by new technology

9 The Global Fintech Landscape Reaches Over 1000 Companies, $105B In Funding, $867B In Value Source: Forbes, 28-Sept 2016 report Successful FinTech Companies by industry

Personal Finance: Envestnet, Yodlee, Mint, Credit Karma, LearnVest, NerdWallet, Personal Capital, Motif, Wealthfront

Lending: Prosper, Lending Club, OnDeck, Kabbage, Funding Circle, CommonBond, SoFi, Affirm, Avant

Payments: Paypal, Klarna, Abyen, Braintree, Mozido, Square, Venmo, Stripe

Money Transfer/Currency: Xoom, Payoneer, TranferWise, WorldRemit, Coinbase, Circle

Crowdfunding: , , GoFundMe, CircleUp, Tilt

Source: The New York Times, "Ranking the Top Fintech Companies" 6‐April, 2016 Source: Nathan Pacer, VP at Venture Scanner At:https://www.nytimes.com/interactive/2016/04/07/business/dealbook/The‐ at:https://www.slideshare.net/NathanPacer/venture‐scanner‐fintech‐report‐q1‐2017 Fintech‐Power‐Grab.html Why would we assume it’s all for the good?

The Jetsons image is licensed under Creative Commons 11 11 Cost of Business & Occupational Fraud

The Association of Certified Fraud Examiners Report to Nations on Occupational Fraud and Abuses: “… with “Asset misappropriation by far the most common form of occupational fraud, occurring in more than 83% of cases…”

If applied to the estimated Gross World Product, this translates to a potential projected global fraud loss of nearly $3.7 trillion ACFE

More than 250,000 Micro & Small Business identity theft businesses were victims of unauthorized caused $268 million in transactions against their business damage in 2016, up from $122 million in 2015. account, with losses totaling $3.1Billion. IRS Javelin LLC

12 So what kind of business fraud are we talking about?

BUSINESS-TO-HIGH RISK & FRAUD

Trade –(AR & AP) Bust out Mortgage Business Id Theft Healthcare – Financial Statement Medical Payment / Trade-Ring Embezzlement Business Credit Coach Tax Evasion Money Laundering Ponzi Wire Fraud Mail Fraud Bribery Forgery Conspiracy Shell / Shelf Corp Public Corruption

EMERGING Cyber Cryptoeconometrics Convergence (e.g. Iot and Crypto) Unexplainable AI Fake news Dark Web ACFE Silent Data Breach 13 Understanding traditional types of malfeasance does not guarantee understanding of new behaviors

Business Identity Theft

Place Orders as an agent Representing Various Delivery Points Impersonating Large VICTIM COMPANIES Impersonated Business - Identity Theft Victim Commercial Business Entities

Walk-In Freezer

Television

Two Men Charged With $900,000 Computers Interstate Theft Scheme NEWARK, N.J. - Two New Jersey men Microwave were arrested today for their roles in a scheme to fraudulently obtain more than $900,000 in commercial and residential Plasma Cutters merchandise from various companies, Acting U.S. Attorney William E. Fitzpatrick Commercial Ice Maker announced. Roy Depack, a/k/a “Ray Depack,” a/k/a “Roy Soriano,” a/k/a “John Soriano,” 42, of Digital Scales Elizabeth, New Jersey, and Louis J. Pobutkiewicz Sr., 39, of Newark, are 100+ Different Power Tools charged by complaint with conspiracy to Phone commit mail and wire fraud. Numbers ETC. Single Family House, Apartments Retail Stores, Restaurants

14 The future of credit management is in part up to us to write…

Geo/Device Identification Bot/Automated attacks Behavior/Activity analysis Smarter human attacks Advanced anomaly detection Data Breach Control vs. Audit/Trust IP theft Federation of AI Adversarial interventions CURIOUS TIMES A REFLECTION ON WHAT IS HAPPENING IN THE WORLD OF ALL THINGS DATA AND AI TODAY

MAKING OLD DECISIONS IN NEW WAYS HOW WE ARE ABLE TO USE MODERN DATA AND TECHNOLOGY TO ANSWER NEW QUESTIONS IN THE TODAY CONTEXT OF BUSINESS DECISIONS

LOOKING AHEAD A LOOK AT SOME OF THE FUTURE DISRUPTIONS IN TECHNOLOGY AND WORKFORCE AND SOME ADVICE ON HOW TO GET READY

16 17 With machine learning anything is possible. That is sometimes part of the problem…

18 The definitions, objectives and methods of “Artificial Intelligence” continue to evolve

• Mimic human behavior ……or loosely approximate the mental models • Mimic thinking • Project human intent • Advise humans • Behave intelligently • Behave rationally • Behave empathetically

…explicitly model the human brain Multistate Markovian Survival Model

Bayes’ Network of perfume characteristics

Architecture of the “projectome” Network model of ‘target’ and ‘source’ nodes via anatomical tracing

19 Learning about learning

Best Practices: • Don’t lead with a method • Understand preconditions • Understand character of data • Mixed methods • Controls for stewardship • Controls for stability of approach • Dispositive threshold Thinking about Artificial Intelligence

What do you have to believe?

Can you vs. may you use information

“Unlearning”

Veracity adjudication

Provenance / decision synthesis

Recreating prior conditions for forensics Thinking about Partner 1 Customer 1 Cloud relationships Cloud Some examples • Supply Chain / Integrated Value Chain Ink Blot, City 1 • Understanding Signals derived from changes to business information Public data • Discovering and investigating clusters of D1.1, D1.2, Cloud DUNS unusual behavior D1.3,…D1.n • Exploring the impact of new regulation (e.g. privacy) Customer 2 Cloud • Understanding intra‐regional opportunities (e.g. cross‐border) • Exploring the impact of new market forces (e.g. Brexit)

Partner 2 • Studying the real or potential impact of Cloud supply chain interruptions (e.g. disasters) • Investigating emerging capabilities (e.g. reputational risk) Ink Blot, City 2 • Malfeasance DUNS D1.1, D1.2, D1.3,…D1.n

22 Black Cat problems…

Dealing with “Black Cat” problems • Signals – internal / external • Nodes / edges • Systemic measures – quality, character • Anomaly detection / Graph inspection  Anisotropism  Betweenness centrality  Hierarchical simplification / Sparse graph analysis  Cliques  Clustering coefficients • Data sensing – new sources / uses • Triggers – events, observations

23 The use of: data science tools, machine learning/artificial intelligence are keys our success to mitigate B2B malfeasance

24 CURIOUS TIMES A REFLECTION ON WHAT IS HAPPENING IN THE WORLD OF ALL THINGS DATA AND AI TODAY

MAKING OLD DECISIONS IN NEW WAYS HOW WE ARE ABLE TO USE MODERN DATA AND TECHNOLOGY TO ANSWER NEW QUESTIONS IN THE TODAY CONTEXT OF BUSINESS DECISIONS

LOOKING AHEAD A LOOK AT SOME OF THE FUTURE DISRUPTIONS IN TECHNOLOGY AND WORKFORCE AND SOME ADVICE ON HOW TO GET READY

25 We live in an age of promise.

Big data and advanced methods are making things possible that were science fiction only a few short years ago. Unintended use is a serious consideration in the rush to market products and services Massive connectivity is also increasing the risk of being “wrong” The state of all things AI also continues to evolve

Source: https://www.forbes.com/sites/gilpress/2017/01/23/top‐10‐hot‐ artificial‐intelligence‐ai‐technologies/#10eb1f7c1928 Truth be told, it’s overwhelming and more than a bit unstable… It will never be simpler.

30 31 We must continue to challenge our perspectives and our beliefs New Behaviors Changing Perspectives Complex Issues

New Opportunity

32 And what about the next generation?

In the past, we looked for…

• Data Curator • Analyst Now, we need all of that • Modeler and more! • Statistician • Methodologist • Coder • Governance Expert • Problem Formulator • Detective • Visionary • Story‐Teller • Diplomat 33 Some Thoughts on The Future New questions to ask… What do we have to believe? (Stability, permissibility)

What happens if we do nothing? (New malfeasance, eroding relevance)

Who are the disruptors?

What do we know? How do we know it (explainability)? What has to remain true (reference frames)?

What is possible if we work together? 35 “While technology is important, it's what we do with it that truly matters.”

Muhammad Yunus, economist, Nobel Prize winner Thank you Anthony J. Scriffignano, Ph.D. SVP/Chief Data Scientist Dun & Bradstreet @Scriffignano1 Presentation Title: Artificial Intelligence Will Continue to Transform Modern Credit Management: Can We Handle It?

Presentation Abstract: Credit decisions have always been informed by data. Data about the counterparties in a relationship, the nature of a transition and historical risk are but a few examples. We are now operating in an era where virtually all of the underlying data is changing in curious and sometime alarming ways. Artificial intelligence (AI), a huge buzzword right now, is one of the driving disruptive forces. While AI is not a new notion, it has taken on increased focus with the explosion of Big Data over the past few years. Machine learning holds the potential to make AI even more effective in the commercial credit industry, and may be a catalyst for AI’s greater adoption. However, some very big questions arise regarding our ability to explain methods and to comply with regulations. As AI increases in complexity, we need to examine the pivotal role that data and the inferences drawn from it by our digital agents have in determining the success businesses will have—and some pitfalls to watch out for as organizations adopt these new technologies.

This session, both relevant and occasionally irreverent, will cover three main themes: Curious Times: An reflection on what is happening in the world of all things data and AI today Making Old Decision in New Ways: How we are able to use modern data and technology to answer new questions in the context of business decisions Looking Ahead: A look at some of the future disruptions in technology and workforce and some advice on how to get ready