Emerging Corporate Learning Trends

With the Development of Generative AI

Dongshuo Li, UMU Founder & CEO

 Understanding the categories of AI through analytical AI and generative AI

Analyze, Find Patterns, Forecast Ideas, Innovation, Creative Work

Analytical AI Analyzing something that exists Generative AI Creating something new

Search Engine Images

e.g Google e.g Midjourney

Feeds Conversations

e.g Facebook e.g ChatGPT

Investment Videos

e.g Bridgewater e.g Sora

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Understanding the application boundaries of AI through general AI and
vertical AI

AI designed to learn any

AI designed to tackle the specific
General AI intellectual task that human Vertical AI challenges in a particular industry
beings can perform

Chatbots Sales Coach

E.g. UMU uShow

E.g. ChatGPT

Semiconductor Design
E.g. Google Gemini
E.g. Nvidia AutoDMP

Image Generation Sales Pitch Analysis

E.g. Midjourney E.g. Gong

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 As general-purpose AI rapidly develops, enterprises increasingly need to
leverage industry-specific AI to benefit from targeted scenarios

Generative AI

Midjourney Image Generation UMU AI Video Video Course Generation

ChatGPT Chatbot UMU uASK Knowledge Management

and Performance Support

GPT-4 Chatbot Nvidia AutoDMP Semiconductor Design

General AI Vertical AI

Google Search Search Engine UMU uShow AI Sales Enablement

Facebook Feeds Recommendation UMU Chatbot Coaching/Sales Pitch

Grammarly Writing Improvement UMU AI Gesture Exercise Sales Enablement

Analytical AI

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Two forms of AI: From AI “Co-pilot” to AI “Auto-pilot”
HIGH REPEATABILITY

Production Maintaining
Manufacturing Apps High-Value Opportunities for
Supply Chain Management
AI x Services
Auditing Pen
Financial Reporting Bookkeeping
Testing
Wealth Tax Planning Company Registration Legal Services
Management Services
Customer
Operations
Sales Transaction Services
Data Analytics
Marketing
Customer Support Financial, Accounting &
Patent/IP Services Auditing Services
Company
Strategy Litigation Cost Optimization
LOW REPEATABILITY

Market Entry Organizational Custom Software IT Development & Support

Strategy Restructuring Implementations

Financial Strategy & Mgmt. Consulting

M&A Advising Due Diligence
Capital
Raising Restructuring
Company business functions

Human + AI Copilot AI Autopilot

Source: The Death of the Big 4: AI-Enabled Services Are Opening a Whole New Market, 04/2024.
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 We expect AI to become even more impactful and autonomous

Accuracy

Copilots for skilled workers AI skilled labor

AI lawyer AI scientist AI exec. AI doctor

Human + AI Agent +
Today Future AI Autopilot
AI Copilot Human Copilot

Creative content generation AI content creators

AI marketers AI influencers AI artists

Creativity

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 The future relationship between human and AI

Cannot be replaced AI works Human and AI Human work AI will replace

小标题文字小
by AI 小标题文字小
for human 小标题文字小
collaboration 小标题文字小
for AI 小标题文字小
in the future

Nurse Siri Mathematical Weather forecast Translation

problem solving
Early childhood Image generation Big data Copywriting
educator Virtual interactive computation and
instructor processing
AI Q&A Basic customer
Caretaker service
…
AI Learning Copilot
Video generation
Craftsman …

…
Course cover
Key account sales
generation
manager

… uShow

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 We’re at Day 1 of AI… and riding on the top of past waves

% US Technology Adoption

100% Smartphones
Internet

75%
AI

50%

25%

1990 1995 2000 2005 2010 2015 2020 2025

Source: The AI Revolution Report from COATUE
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Adoption has been twice as fast with each platform shift

Halving of penetration time with new technology waves

# of years to reach 50% user penetration in the US

20 Years

12 Years

?
6 Years

3 Years

Pcs Internet Mobile Gen AI

Source: The AI Revolution Report from COATUE

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 AI accelerates productivity improvement significantly
Example: Designing Marketing Materials

Analog PCs Internet & Cloud Generative AI

1950s 1960s - 1980s 1990s - 2010s 2020s + Beyond

Months Weeks Days Minutes

Design on the paper Design on local software Design on web apps AI-created design
Iterate on physical copies Easy to iterate Instant collaboration Human prompts with text

Physical Digital Digital Digital

Individual Individual Collaborative Collaborative

Manual Manual Manual Automated

Source: The AI Revolution Report from COATUE

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 AI accelerates productivity improvement significantly
Example: Corporate Learning

Face to Face E-learning Online Learning Digital Learning

1950s 1960s-1980s 1990s - 2010s 2020s + Beyond

Months Weeks Days Minutes

Whiteboard and Learning in Learning on AI-driven learning
face-to-face a local area network an internet platform
Iterate on Physical Copies Easy to Iterate Instant Collaboration Prompts with Text

Physical Digital Digital Digital

Individual Individual Collaborative Collaborative

Manual Manual Manual Automated

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Last 50 years was about building faster & faster “calculators”...

CPU Software Computer

Serial processing Based on instruction by programmer
or
“1 instruction at a time”
Chips got smaller & more powerful
Follows sequential programming logic
Does not require data
“Calculator”

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 But next 50 years will be about building super-intelligent “brains”

GPU AI Model Brains

Neural networks trained by data
Parallel processing Much more than calculators!
Many calculations simultaneously Learns patterns from data
Could become connected:
System makes decisions based on model Brain-to-brain network = AI Internet?
rather than explicit instructions

"Reasoning" is opaque,
not driven by programming logic

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Generative AI’s biggest impacts on collaboration & expertise application
With Generative AI
Overall technical automation potential, comparison in midpoint scenarios, % in 2023 Without Generative AI
Activity groups

Decision making and Applying expertise 58.5

collaboration 24.5
Managing 49.0
15.5
Interfacing with 45.0
stakeholders 24.0

Data management Processing data 90.5

73.0
79.0
Collecting data
68.0

Physical Performancing 46.0

unpredictable physical work 45.5

Performancing predictable 73.0

physical work 72.5

Source: The economic potential of generative AI from McKinsey & Company

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 NBER research paper: Generative AI effectively increases productivity

Researchers study the staggered introduction of a generative AI-based conversational assistant using data from
5,179 customer support agents. By measuring factors such as resolutions per hour, chats per hour, and average
handle time, it is found that the use of generative AI tools is effective in increasing productivity. The average
improvement is 14%.

Evidence Based Analysis Based on 5,179 Customer Support Agents

Never AI Pre AI Post AI

.05
6 6
.04
4 4 .03
Density

Density
Density .02
2 2
.01

0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 20 40 60 80
Resolutions Per Hour Chats Per Hour Average Handle Time

Source: Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at work (No. w31161). National Bureau of Economic Research.
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 MIT research demonstration: AI application on writing tasks

After using ChatGPT, the time taken on task drops by 10 mins (37%), and the
average grades increase by 0.45 standard deviations.
GPT has narrowed the differences between individuals.
Whitney Zhang Shakked Noy
PhD PhD

(a) Time Taken Decreases (b) Average Grades Increase

35 5.2
Treatment Effect: -0.83 SDs Treated Treatment Effect: 0.45 SDs
95% CI: [-0.63. -1.03] Control 95%CI: [0.27, 0.63]
Self-Reported Time Spent (Minutes)

30 4.8

Mean Grade
25 4.4

20 4

Treated
15 Control 3.6

Pre-Treatment Post-Treatment Pre-Treatment Post-Treatment

Source: Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187-192.
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Harvard Business School & BCG research: AI improves work efficiency

Performance Distribution - Inside Frontier Bottom-Half Skills and Top-Half Skills - Inside the Frontier
Baseline Task Experimental Task
0.8 GPT+Overview
GPT Only
0.7 Control

7 7 +17%
0.6 +43%

0.5 6 6
Density

5 5
0.4

Score Task

Score Task
4 4
0.3 5.79 6.06
3 3
5.20
0.2
2 4.05 2
0.1 1 1
0.0 0 0
1 2 3 4 5 6 7 8
Bottom-Half Top-Half
Quality Skilled Participants Skilled Participants

Participants using AI were significantly more productive There were performance enhancements in the
(they completed 12.2% more tasks on average, and experimental task for both groups when leveraging
completed tasks 25.1% more quickly), and produced GPT-4. The bottom-half-skill performers exhibited the
significantly higher quality results (more than 40% most substantial surge in performance, 43%, compared
higher quality compared to a control group). to the top-half-skill subjects, 17%.

Source: Dell'Acqua, F., McFowland, E., Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., ... & Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the
effects of AI on knowledge worker productivity and quality. Harvard Business School Technology & Operations Mgt. Unit Working Paper, (24-013).
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Insights from March: Exploration and exploitation in organizational learning

Organizations should have greater patience in exploring

the application of AI to the development of individual and organizational productivity.

Exploitation - Exploration

Keep putting effort into areas where the Traits Invest in areas where organization currently lacks
Stanford Professor organization excels. expertise but may have significant market potential.
James G. March
(Jan. 1928 – Sep. 2018) The certainty, speed, proximity, and clarity of Returns are systematically less certain, more remote
feedback ties exploitation to its consequences more in time, and organizationally more distant from the
quickly and more precisely. locus of action and adaptation.

Competence Trap Trap Short-Term Failure Trap

They suffer the costs of experimentation without They are likely to find themselves trapped in suboptimal
gaining many of its benefits. They exhibit too many stable equilibria.
undeveloped new ideas and too little distinctive
competence.

Source: March, J. G. (1991). Exploration and exploitation in organizational learning. Organization science, 2(1), 71-87.
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Generative AI's impact on freelance professionals’ performance

A Group 2-Month Performance

Control: No AI Observation
The Impact of Generative AI
Tools and Training Enablement
on the Productivity of
Freelance Professionals
B Group 2-Month Performance
AI Only Observation
Control

AI Only Group

AI and Training

C Group 2-Month
Training Performance
AI and Training Observation

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Generative AI's impact on freelance professionals’ performance

Boxplot of Active Days by Group Mean Income by Group

50
51.69
50

40
40

Mean Income
Active Days

30
30

20
20

10 10.41
8.59
10

0
0
Control AI Only Group AI and Training Control AI Only Group AI and Training

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Training + LLM AI boosts performance significantly

Two-sample T-test - Mean LLM AI Only Training + LLM AI MeanDiff

Sample Size N=201 N=215 -

Interactive Round 6.2 7.8 1.6 / 25.8%

Total Prompt Length 347.4 451.2 103.8 / 30.0%

First Prompt Length 95.2 120.9 25.7 / 26.2%

Question Diversity 5.3 5.7 0.4 / 7.5%

Question Depth 5.9 6.5 0.6 / 10.2%

Question Clarity 6.6 7.1 0.5 / 7.6%

Feedback 3.4 4.9 1.5 / 44.1%

Source: Li Ning et al. (2024). Human Factors as Drivers of Success in Generative AI-Human Collaborations
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Introducing “AI Literacy” for organization & the Competency Model

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 AI has become an integral part of organizational structures

Organization Organization

Talents Tasks Talents AI

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 "AI Literacy" is a new key competency for organizations and talent

Why do I feel somewhat clumsy

What is “AI Literacy” ?
using LLM ?

IQ
“AI Literacy” is the ability of an individual to understand and apply AI
EQ to solve key issues in the workplace.
Increasing the “AI Literacy” level of talent allows them to perform
Teamwork Skills
better at work, and increase organizational productivity by
Leadership leveraging AI.

AI Literacy

Organization Organizations need to keep up with

the training of talent to ensure that
they interact and collaborate better
with AI, thereby enhancing the
Talent AI individual and organizational
productivity.

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 The ASK of "AI Literacy": The Attitude, Skills, & Knowledge that talent need

Attitude

Open Mindset: Embrace AI, learn and explore new tech.

Self-Iteration: Adapt to evolving AI, embrace change.

A Innovation Awareness: Apply AI to solve problems,

explore new methods.

Skills

Prompt Optimization: Enhance AI interaction

Knowledge
with high-quality prompts. AI Fundamentals: Understand the basic concepts,
history, future trends, and applications of AI.
Critical Evaluation: Evaluate the information
and advice provided by AI, identifying possible
S K Prompt Engineering: Optimize LLM’s performance with
biases, errors or inaccurate outputs. prompt optimization.

Information Synthesis: Combine AI insights LLM Based Tools: Understand and apply various LLM
with other sources for comprehensive tools effectively.
decision-making.
Real-world Use Cases: Learn how AI is solving
real-world problems in various industries.

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 The training industry has strong ownership in “reskilling” in the AI wave

“ Respondents at these organizations are over

three times more likely than others to say their
organizations will re-skill more than 30 percent of
their workforces over the next three years as a
result of AI adoption.
”
Source: The state of AI in 2023: Generative AI’s breakout year - published by McKinsey in August 2023.

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 Organizations are increasingly focusing on AI literacy and its impacts

Organizational CEOs' outlook on generative AI in the next 3 years— US CEOs

To what extent do you agree with these statements about generative AI? CEOs Globally

"AI Literacy" - the premise of skill

Organizational competitiveness Organizational value creation
development

● In the next three years, generative Al will increase ● In the next three years, generative Al will ● In the next three years, generative Al will require
competitive intensity in my industry significantly change the way my company most of my workforce to develop new skills
creates, delivers and captures value

68% 70% 69%

68%
65%

61%

Source: PwC’s 27th Annual CEO Survey (01/2024), base of 4,702, US base of 231
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 High-performing AI leaders strategically focus on the advancement of AI
within their organizations

All other respondents Respondents at AI high performers

19 Reduce costs in core business

33

23 Create new businesses and/or sources of revenue

12
100%
21 27 Increase revenue from core business

33 30
Increase values of offerings
by integrating AI-based features or insights

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 High-performing AI leaders strategically focus on the advancement of AI
within their organizations

All other respondents Respondents at AI high performers

9 Don’t Know
Respondents at AI high performers
21 73 >30% expect their organizations to reskill
A larger part of the workforce than other
14 respondents do.

100%
18 Share of employees at respondent’s
organization expected to be reskilled over
the next 3 years as a result of AI adoption,
9 21-30% % of respondents.
38
10 11-20%
8 ≤10%

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 What are the attitudes of organizations and teams towards AI?

58% of the respondents think their teams are still learning about AI
and may not understand it yet

58%

10% 10% 10% 12%

My team is pessimistic My team is indifferent to My team is still learning My team is trusting of AI My team is overly
about the use of AI AI about AI and may not optimistic about the use
understand it yet of AI
Q: Which of the following best describes your teams’ attitude toward AI?
© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 What are the challenges of AI applications in enterprises?
65%
Need more knowledge and skills on how AI-enabled applications work

61%
Need a better understanding of how AI will positively impact our work and business outcomes

60%
Need more knowledge and skills on how to effectively implement AI-enabled applications in our company

55%
Need more benchmarking data and use cases on how other organizations are using AI enabled
applications for talent and learning

46%
Need to partner with internal stakeholders to get buy-in

31%
We are interested, but not sure where to start

3%
Other

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 ATD x UMU: Tapping the Potential of Sales Enablement Report

AI in Learning and Talent Development

● How prepared are Enterprises for AI?
● How do Enterprises use AI for learning, TD?
● What have Enterprises achieved with AI?
● What are the challenges of AI applications?

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
 ATD x UMU: Tapping the Potential of Sales Enablement Report

Download Report

© 2014-2024 UMU Technology Co. Ltd All Rights Reserved UMU Confidential
Thank You !
[email protected]