This talk analyzes two case studies to illustrate the management and product development challenges companies face when applying machine learning.

1. No Silver Bullet
A Rational Framework to Evaluate the
Risks and Rewards of Enterprise AI
@integrateai | @humekathryn | quamproxime.com

2. @integrateai | @humekathryn
Osama bin Laden’s compound in Abbottabad, Pakistan (2011)

3. @integrateai | @humekathryn

4. @integrateai | @humekathryn

5. @integrateai | @humekathryn
Jeffrey Friedman & Richard Zeckhauser
95%
80%
60%
40%
30%
WTF?!?!

6. @integrateai | @humekathryn
Jeffrey Friedman & Richard Zeckhauser
President Obama reportedly complained that the discussion
offered ‘not more certainty but more confusion,’ and that his
advisors were offering ‘probabilities that disguised uncertainty as
opposed to actually providing you with useful information.’

7. @integrateai | @humekathryn
“If we did not take action, he might slip away. And it might be
years before he resurfaced.”

8. What the hell does this have to do with AI?
@integrateai | @humekathryn

9. Everything!
(Well, not everything but this is a talk and
hyperbole is a strong rhetorical device)
@integrateai | @humekathryn

10. “Over the past decades computers have broadly
automated tasks that programmers could describe with
clear rules and algorithms. Modern machine learning
techniques now allow us to do the same for tasks where
describing the precise rules is much harder.”
Bezos 2017 Letter to Shareholders
@integrateai | @humekathryn

11. AI?
Machine
Learning
Data Science
Analytics
“Big Data”
Hilary Mason
Fast Forward Labs
@integrateai | @humekathryn

12. @integrateai | @humekathryn
This is a cat. This is most probably a cat. This cat is dead and alive.

13. Developments in Language Processing
Traditional NLP N-grams Word Embeddings
@integrateai | @humekathryn

14. @integrateai | @humekathryn

15. @integrateai | @humekathryn
13% increase on auto insurance transactions
38% improvement in predictions during pilot

16. Chief Financial Officer Chief Scientist
@integrateai | @humekathryn

17. Sunk costs Angry customers Lost trust Unfair treatment
@integrateai | @humekathryn

18. @integrateai | @humekathryn
How can businesses reap the rewards
while managing the risks?

19. The Raw Ingredients of an AI product
Deep understanding of a business problem
Data, data, data
Algorithmic capability
Integrations and interfaces
@integrateai | @humekathryn

20. Machine Learning Product Lifecycle
Design
What problem
are we solving?
Data Exploration
What does the data
look like?
Production
Can we scale the model?
Which is best?
Maintenance
How to update
as data changes?
Data Processing
Is our data ready
for use?
Model Prototyping
Will this work?
Should we pivot?
Data Generation
How do tools capture
information?
Data Collection
Who is well represented?
Where are blind spots?
Evaluation/Interpretation
How do we know
it’s working?

21. @integrateai | @humekathryn
Case Study #1
Partially automate checklist completion
20,000 examples completed lists
First party keeps data & model

22. Design
Where does this fit into the entire business process?
@integrateai | @humekathryn

23. Design
What level of certainty does the problem demand?
@integrateai | @humekathryn
versus

24. Design
Do you have a motivated subject matter expert?
@integrateai | @humekathryn

25. Model Type Amount of data Structure of data
Subject Matter
Expertise
Explainability
Regression Small ok
Less Complex
Features
Selecting the
features
Medium-High
Bayesian Small ok
Hierarchical
Features
Defining the priors Medium-High
Deep Learning Large only
Hierarchical
Features
Little
(Labeling)
Low
Reinforcement
Learning
Explore/Exploit
Depends
(Deep vs.
Classical)
Simulate the
environment
Depends
(Deep vs.
Classical)

26. @integrateai | @humekathryn
Data Exploration
How much of this is actually a machine learning problem?

27. Model Prototyping
Will this work? Should we pivot?
@integrateai | @humekathryn

28. @integrateai | @humekathryn
Maintenance
How to update as data changes?

29. @integrateai | @humekathryn
Lessons learned
• Key risk is wasted cost & time
• Seek
• Clearly defined input-output with large labeled training set
• Motivated SME to collaborate with technical team
• Avoid
• Tasks that require certainty
• Tasks with little output variety
• Beware
• Hype leading to misunderstanding of what’s possible
• Tactics
• Scope PoC tightly
• Disprove hypothesis to minimize sunk cost

30. @integrateai | @humekathryn
Case Study #2
Increase conversion rates
Design infrastructure to measure outcomes
Third party gets data & owns model

31. Design
Where does this fit into the entire business process?
@integrateai | @humekathryn

32. “There is often a difference between the quantity you
want to measure and the one you can measure.
When these differ,
your model will become good at predicting the quantity you measured,
not the quantity for which it was meant to be a proxy. "
https://medium.com/@yonatanzunger/asking-the-right-questions-about-ai-7ed2d9820c48
@integrateai | @humekathryn

33. COMPAS risk assessment tool
• What we think we’re asking: “what is the
probability that this person will commit a
serious crime in the future, as a function of the
sentence you give them now?”
• What we’re actually asking: “who is more likely
to be convicted?”
@integrateai | @humekathryn
Design
What problem are we solving?

34. @integrateai | @humekathryn
Data Collection
How much data should we share with our vendor?
versus

35. https://visual.ly/community/infographic/
human-rights/taxonomy-transitions
@integrateai | @humekathryn
Data Collection
Removing attributes doesn’t always
remove bias

36. @integrateai | @humekathryn
Production
Can we scale the model?

37. @integrateai | @humekathryn
Evaluation/Interpretation
How do we know it’s working?

38. @integrateai | @humekathryn
Lessons learned
• Risks around sunk cost, data privacy & security, ethics
• Seek
• Problems where a good guess can provide measurable boost
• Collaborative and flexible partners
• Beware
• Understand what your proxies actually optimize for
• Be prepared to problem solve with imbalanced data sets
• Be Pragmatic but Principled
• Explainability is not an absolute - it’s context dependent
• Go the extra mile to evaluate fairness
• Tactics
• Performance-based fee structure to align incentives

39. @integrateai | @humekathryn
What should we do?

40. @integrateai | @humekathryn
Executive Leadership
Line Management
Project Leads
ML & Engineers
Vendors

41. There’s no silver bullet.
We are empowered to design
the future we want to live in.
@integrateai | @humekathryn

42. @integrateai | @humekathryn | quamproxime.com
Thank you!