Sönke Rohde gave a presentation at UX STRAT USA 2021 on the topic: "Deep Learning UX: AI-Powered, Personalized Design at Scale"

1. Deep Learning UX
AI-powered, Personalized Design at Scale
Twitter: @soenkerohde #DLUX #UXSTRAT
Sönke Rohde (he/him/his), VP UX R&D

3. This presentation contains forward-looking statements about the Company’s financial and operating results, which may include expected GAAP and non-GAAP
financial and other operating and non-operating results, including revenue, net income, earnings per share, operating cash flow growth, operating margin
improvement, expected revenue growth, expected current remaining performance obligation growth, expected tax rates, stock-based compensation expenses,
amortization of purchased intangibles, shares outstanding, market growth, environmental, social and governance goals, expected capital allocation, including
mergers and acquisitions (such as the proposed acquisition of Slack Technologies, Inc.), capital expenditures and other investments, expectations regarding closing
contemplated acquisitions and contributions from acquired companies. The achievement or success of the matters covered by such forward-looking statements
involves risks, uncertainties and assumptions. If any such risks or uncertainties materialize or if any of the assumptions prove incorrect, the Company’s results could
differ materially from the results expressed or implied by the forward-looking statements it makes. The risks and uncertainties referred to above include those
factors discussed in Salesforce’s reports filed from time to time with the Securities and Exchange Commission, including, but not limited to: risks associated with
our ability to consummate the proposed Slack Technologies, Inc. transaction on a timely basis or at all; our ability to successfully integrate Slack Technologies, Inc.’s
operations; our ability to realize the anticipated benefits of the proposed transaction; the impact of Slack Technologies, Inc.’s business model on our ability to
forecast revenue results; disruption from the transaction making it more difficult to maintain business and operational relationships; the impact of, and actions we
may take in response to, the COVID-19 pandemic, related public health measures and resulting economic downturn and market volatility; our ability to maintain
service performance and security levels meeting the expectations of our customers, and the resources and costs required to avoid unanticipated downtime and
prevent, detect and remediate performance degradation and security breaches; our ability to secure and costs related to data center capacity and other
infrastructure provided by third parties; our reliance on third-party hardware, software and platform providers; the effect of evolving domestic and foreign
government regulations, including those related to the provision of services on the Internet, those related to accessing the Internet, and those addressing data
privacy; current and potential litigation involving us or our industry, including litigation involving acquired entities such as Tableau; regulatory developments and
regulatory investigations involving us or affecting our industry; our ability to successfully introduce new services and product features, including any efforts to
expand our services beyond the CRM market; the success of our strategy of acquiring or making investments in complementary businesses and strategic
partnerships; our ability to compete in the market in which we participate; the success of our business strategy and our plan to build our business; our ability to
execute our business plans; our ability to continue to grow unearned revenue and remaining performance obligation; the pace of change and innovation in
enterprise cloud computing services; the seasonal nature of our sales cycles; our ability to limit customer attrition and costs related to those efforts; the success of
our international expansion strategy; the demands on our personnel and infrastructure resulting from significant growth in our customer base and operations; our
dependency on the development and maintenance of the infrastructure of the Internet; our real estate and office facilities strategy and related costs and
uncertainties; fluctuations in, and our ability to predict, our operating results and cash flows; the variability in our results arising from the accounting for term
license revenue products; the performance and fair value of our investments in complementary businesses through our strategic investment portfolio; our ability to
protect our intellectual property rights; our ability to develop our brands; the valuation of our deferred tax assets and the release of related valuation allowances;
uncertainties regarding our tax obligations in connection with potential jurisdictional transfers of intellectual property; uncertainties regarding the effect of general
economic conditions; and risks related to our debt and lease obligations.
Forward-Looking Statements
031621

4. Personalization
Is everywhere

5. Personalization
What about Design?

6. Personalization Today
Personalize the channel &
time to deliver the content
Personalize which
content to show
Content Messaging

7. Personalization Today
Personalize the channel &
time to deliver the content
Personalize which
content to show
Template
Content Messaging

8. Shopper Experience
Paula Beth

9. Paula Beth
Shopper Experience

10. Paula Beth
Shopper Experience

11. One size does not fit all
Personalized Design

12. One size does not fit all
Personalized Design
1
3
2
4

13. Meaningful, bespoke experiences tailored to each individual
Relationships by Design

14. Why don’t we see Personalized Design today?
It’d take an army

15. “Pick two”

16. Personalized Design
Requires a New Approach (hint: AI)
Personalization of the Experience Layer:
1. variations covering the spectrum of preferences
2. continuously explore and validate learnings
Variations (at Scale)
• Programmatic/Declarative UI
• Generative UI
Experimentation (1:1)
• A/B & Multivariate testing
• Contextual Bandit

17. Deep Learning UX
AI-Powered, Generative Design

18. ● ~1K web pages of the Fortune 500
and Unicorns
● Extracted screenshots, normalized
DOM and styles
Data Analysis
Color
Full Height Website Color Distribution / Content Block
Data Viz by Moritz Stefaner for Salesforce

19. Color Split by Industry

20. ● Typography is intrinsic to design
and gives content structure and
hierarchy.
● Style relationships are complex
and brand specific
Data Analysis
Typography
Font Size Distribution
Style Relationships
Human bias towards even over odd # Font Face Distribution

21. Latent Style Space
Data Exploration Takeaways
The latent space is basically a representation of
compressed data which in this case enables the
featurization of design rules and heuristics.
The key takeaway from our Data Science research is that
there is a meaningful latent space when it comes to style
encapsulating aspects around color, typography, layout
and other design aspects.

22. Simulation

23. Placing Content on Image Assets
Templates don’t cut it

24. Computer Vision
Copyspace
● Identifies/predicts the best regions for
placing content on top of an image
● Avoids covering areas of interest
● Accessibility (color-contrast) built in

25. Design & Content Variations
Design
Variations
Content
Variations

26. Experimentation
Turning into Personalization

27. A/B & Multivariate (MVT) testing
● Tests consist of a randomized
experiment with a set of
variants
● Multivariate testing : testing
multiple variables to find the
ideal combination
Experimentation
Terminology
Contextual Bandit
● MAB + feature & context vector
(e.g. user and state)
● Experimentation turning into
Personalization
(Explore/Exploit)
Multi-armed Bandit (MAB)
● A/B/n Testing + Machine
Learning
● Dynamically allocate traffic to
variations based on
performance to mitigates
“regret”
A/B Testing Multi-armed Bandit Testing Contextual Bandit Testing
Variation A
Variation B
Variation D
Variation C
t
%

28. AI-Powered, Personalized Design at Scale
AI-generated Variations
e.g. DLUX
Learn
Feedback Loop
AI Personalization Engine
e.g. Contextual Bandit

29. Artificial yet Ethical
How might we mitigate
unintended consequences?

30. Conducted during initial conception, roadmap planning, and
feature creation. Answers 3 questions:
1. What are the intended and unintended consequences of this
product or feature?
2. What are the positive consequences we want to focus on?
3. What are the negative consequences we want to mitigate?
Consequence Scanning
An agile practice for responsible innovators
“Consequence Scanning is a lightweight methodology
that gets teams accustomed to flexing their ethical
foresight and building moral muscle memory.”
— Yoav Schlesinger, Director, Ethical AI Practice
● Answers are logged, responsibilities to act are assigned, actions
tracked in product backlog
More info: https://doteveryone.org.uk/project/consequence-scanning/

31. Source: NATIONAL HIGHWAY TRAFFIC SAFETY ADMINISTRATION https://www.nhtsa.gov/technology-innovation/automated-vehicles-safety
Levels of Driving Automation

32. Level 2: Automate Tedious, Repetitive Tasks
Design Automation Levels
Design Assistance Partial Design Automation Conditional Design
Automation
High Design Automation
No Design Automation Full Design Automation
Hand-drawn designs
(e.g. Photoshop)
using a style guide to
ensure consistency.
Hand coded UI.
Modern design tools
assist with design
and Design Systems
for consistency.
Component
development.
Computational
model of a brand’s
design generating
experience
fragments.
Automatic
composition of
experience
fragments and layout
variations.
TBD TBD
1 2 3 4
0 5
Goal
Today Tomorrow
Scope
Accelerate Designer &
Developers tasks
Automate tedious,
repetitive Junior
Designer/Developer
tasks
Automate tedious,
repetitive Senior
Designer/Developer
tasks
TBD TBD

33. Human in the loop
Augmented Intelligence
Design AI
Customer
User
AI
Variations
Personalization

34. Human in the loop
Augmented Intelligence
Design AI
Customer
User
AI
Variations
Selections Personalization

35. Human in the loop
Augmented Intelligence
Design AI
Customer
User
AI
Variations
Selections
Prediction
Personalization

36. Human in the loop
Augmented Intelligence
Design AI
Customer
User
AI
Variations
Selections
Prediction
Feedback Loop
Personalization

37. Human in the loop
Augmented Intelligence
Design AI
Customer
User
AI
Variations
Selections
Prediction
Feedback Loop
Personalization

38. Human in the loop
Augmented Intelligence
Design AI
Customer
User
AI
Variations
Selections
Prediction
Feedback Loop
Personalization
Implicit & Explicit

39. Einstein Designer
Powered by Deep Learning UX

40. Marketing Cloud
Humanize every moment with real-time personalization
Einstein Designer creates relevant content with the help of AI in seconds using
variations of available images, copy, and headlines.

41. Democratization of Design
Today
Marketing
Specialist
Creative
Designer
days
or
weeks
images
and
text
creativ
e
assets
Creative
Designer
Tomorrow
Marketing
Specialist
minutes
images
and
text
creativ
e
assets
Elevate designers to
think creatively
about the brand.
Designers
perform
repetitive
work
combining
images and
text
Einstein
Designer
enables
Marketing
Specialists
to operate
autonomou
sly
Creative
team can
update
DLUX to
control their
brand from
a single
point.
seasonally
Tune
Branding
Einstein
Designer

42. Personalized Design at Scale
AI-generative Design
• Automates tedious, repetitive tasks - Democratize Design
• Generates variations for experimentation/personalization
AI-powered Experimentation
• Enables efficient data-driven design - take the subjectiveness
out of the creative process
• Advanced experimentation frameworks like Contextual Bandit
enable 1:1 personalization of design
Personalized Design
• No more “one size fits all” for the experience layer
• Meaningful, bespoke experiences tailored to each individual
Building Relationships by Design

43. Thank You