MIT SMR CONNECTIONS

SURVEY RESEARCH REPORT

Generative AI for
Data and Analytics:
How Early Adopters Are Reaping the Rewards

ON BEHALF OF:
 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

Introduction ............................................................................................................... 1

C O N T E N TS
Digging Into the Data.................................................................................................. 1
Early Adopter Characteristic No. 1: Focusing on Better Decision-Making..................... 2
• Ecolab: Using Generative AI to Boost Sales........................................................... 3
Gaining a Competitive Edge....................................................................................... 5
Early Adopter Characteristic No. 2: Developing Strategies,
Solving Challenges, and Tracking Results.................................................................. 5
Early Adopter Characteristic No. 3: Establishing Bullish Expectations
for ROI and Revenue Generation.................................................................................7
• Verizon: Tracking Generative AI Results.................................................................7
Early Adopter Characteristic No. 4: Building Strong Business-Data Partnerships......... 8
Early Adopter Characteristic No. 5: Relying Primarily on Third-Party Tools.................10
• Data Snapshot: Different Company Sizes, Different Speeds..................................10
Early Adopter Characteristic No. 6: Obtaining the Right Skills for
Generative AI Deployment........................................................................................ 12
• Data Snapshot: Different Leadership Roles, Different Outlooks............................. 12
Keeping a Human in the Loop.................................................................................... 13
• Checklist: Getting Started With Generative AI for Analytics.................................. 14
• Data Snapshot: Different Places, Different Priorities............................................. 14
Conclusion: The Path to Generative AI Success......................................................... 14

Sponsor’s Viewpoint................................................................................................. 15
Methodology............................................................................................................ 17
Demographic Information......................................................................................... 17
Acknowledgments................................................................................................... 18

MIT SMR Connections develops sponsored content. It operates

independently of the MIT Sloan Management Review editorial group.

Copyright © Massachusetts Institute of Technology, 2024. All rights reserved.

MIT SMR CONNECTIONS i

 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

Introduction
Businesses are rapidly embracing generative AI, which has the potential to transform the way
people work and make decisions across nearly every facet of business. Use of the technology
nearly doubled in the first half of 2024, and while some organizations are still in the planning
stages, others are moving ahead to explore generative AI’s capabilities for everything from
boosting productivity and sales to enhancing the customer experience, designing better
products and services — and much more.

But for organizations currently deploying generative AI — the early adopters — the technology’s
greatest value appears to lie in its ability to analyze and improve strategic business decisions,
a new survey shows. This report unveils the survey’s key findings with an emphasis on what
those early adopters are doing to lead the pack.

The global survey, conducted by MIT SMR Connections and its This report covers early adopters’ strategies, progress, challeng-
research partner Kadence International with sponsorship from es, and accomplishments, supplemented with insights from busi-
ThoughtSpot, queried 1,000 data and business leaders at compa- ness and academic experts. Together, these components provide a
nies large and small, across a range of industries and geographies, snapshot of the state of generative AI in today’s enterprises and a
to identify prevailing trends in using generative AI for analytics. It glimpse of the transformative future it offers to those who under-
found that many early adopters are receiving encouraging results, stand how to wield its power effectively.
with a majority seeing significant payback from their investments.
These organizations have even higher expectations for ROI and Digging Into the Data
revenue for the future, as they expand both the number and scale The survey gathered information from data and business leaders in
of their deployments. technology, financial services, manufacturing, retail, construction,
health care, and professional services. It found that 67% of those sur-
Aside from their focus on analytics, organizations that are getting veyed are early adopters. Another 26% of respondents haven’t yet
ahead with generative AI share common characteristics in terms deployed the technology but expect to do so; we’re calling this group
of strategy, implementation, and collaboration. At the same time, “planners” in this report. Just 7% are still evaluating the technology.
they’re encountering challenges such as aligning business and data (Note: To gain a more relevant picture of why and how respondents
leaders’ priorities, dealing with data and model quality, and ad- are using generative AI for analytics, researchers screened out poten-
dressing security concerns. tial participants who indicated that their organizations aren’t using
the technology and have no plans to do so.)

MIT SMR CONNECTIONS 1

 SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

Early Adopter Characteristic No. 1:

Focusing on Better Decision-Making
Analysis of the survey data revealed that the early adopters, many In the short term, early adopters are keen on using generative AI to
of whom are already achieving demonstrable results, share several improve data-driven insights; longer term, they’re looking to use
characteristics. Specifically, they: those insights to solve business problems, such as increasing sales
and improving customer experience. (For an example, see: “Ecolab:
Using Generative AI to Boost Sales.”)
1. Focus on deploying generative AI to help executives and
managers make better-informed decisions. When it comes to recognizing the technology’s transformational
2. Recognize the importance of creating a clear strategy impact, the survey respondents are far from alone. Using genera-
for generative AI projects and closely tracking results tive AI for business decisions will soon become essential for suc-
and spending.
cess, according to the World Economic Forum. The organization
3. Have bullish expectations for solid ROI and notes that 40% of CEOs are already using generative AI to inform
revenue generation. their decision-making, and states: “We think that future compet-
4. Work to build strong partnerships between their business itiveness may not only be about who has the most advanced AI,
and data teams. but also how this technology is used for strategic decision-making.
5. Rely primarily on third-party tools, rather than attempting This will ultimately influence the competitiveness of companies
to develop generative AI technology in-house. and countries in the age of AI.”
6. Understand the key technical and organizational skills
ing they need for success. In the survey, early adopters specifically singled out the technolo-
gy’s ability to speed data-driven decision-making, simplify the pre-
sentation of complex data, improve business insights, and improve

We will examine each of these characteristics in detail. the quality of products and services as key benefits (see Fig. 2).

“Generative AI can support informed and strategic decision-mak-

ing in a way that couldn’t be done before,” says Beena Ammanath,

Figure 1: global head of the Deloitte AI Institute and a noted AI author and
Generative AI speaker. “It can help business leaders quickly understand complex
in Analytics: 67% data sets and use historical data to forecast future trends and sim-
Strong Adoption ulate business scenarios. That allows leaders to anticipate and plan
for market changes on a scale that was not previously possible.”

Executives and employees know that making data-based deci-

Early Adopters
sions is important, but traditional methods of incorporating it are
time-consuming and off-putting for many users, says Thomas H.
Davenport, the President’s Distinguished Professor of Information
Technology and Management at Babson College and coauthor of,
among other books, All Hands on Tech: The AI-Powered Citizen
Revolution (John Wiley & Sons, October 2024).
26% 7%

Planners Evaluating

2 MIT SMR CONNECTIONS

ata-driven decision making: 44%
products and services: 44%
business insights: 42%
on of complex data: 40%
productivity: 38% MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING
atisfaction: 38%
zation of data: 36%
g: 35%
35%
29%

llowed

Figure 2:
Early Adopters’ Improving speed of data-driven decision-making 44%
Top Reasons for Improving quality of products and services 44%
Using Generative
AI in Analytics* Improving quality of business insights 42%
Simplifying presentation of complex data 40%
Increasing employee productivity 38%
Increasing customer satisfaction 38%
Increasing democratization of data 36%
Improving forecasting 35%
Mitigating risk 35%
Competing for talent 29%
*Multiple selections allowed

Ecolab: Using Generative AI to Boost Sales

Ecolab, a multinational corporation that provides technology to The model uses predictive analytics to point managers and
help companies improve safety, health, and cleanliness, analyzes salespeople to potentially lucrative opportunities they wouldn’t
a lot of customer data. But the platforms it was using to do that have discovered otherwise.
were labor-intensive and time-consuming.
For example, the company provides several types of services
“We have business analysts, finance analysts, sales analysts to a global manufacturer client, including equipment cleaning,
— analysts across the spectrum — who spend a lot of time water recycling, pest control, and safety controls. “We likely
downloading data and wrangling it to try to come up with do business with all of their plants, but every plant doesn’t use
answers,” says Marc Labelle, global director of the enterprise services from all our divisions,” Labelle says.
data office for Ecolab. “We wanted to shrink the time from
question to answer.” Instead of manually reaching out to individual customers and
encouraging them to try Ecolab’s other services, they were able
Labelle and his team also wanted the company’s analytics to to train the generative AI model on data about the client’s oper-
yield more actionable results. “They tell you what happened ations and financial performance.
yesterday,” he says. “What people want to know is: ‘What op-
portunities do I have today?’” “We used advanced AI techniques to learn which plants are
performing the best,” Labelle explains. “Which one is producing
A pilot program using generative AI for analytics is helping products with the least pest issues? Which is saving the most on
the company achieve both of these goals. And the insights energy costs? Which is fulfilling the promises it made about wa-
the technology is delivering could supercharge sales and ter conservation? Then we determined what actions the other
customer service. plants need to take to bring themselves up to that level.”

The pilot program uses generative AI and natural language By providing valuable information that can help its customers
processing, which allow analysts and salespeople to ask ques- meet their goals, Ecolab stands to substantially improve its own
tions of their business data rather than manually manipulating financial performance — selling its services naturally, with less
and interpreting it in spreadsheets. effort. “This is not something you can do with a traditional busi-
ness intelligence platform,” Labelle says.
“You don’t have to wrangle anymore. You just ask questions,
and when you get a response, drill down,” Labelle says. Ecolab is also exploring generative AI for other use cases, he
adds: “We’re just getting started, but we think it could offer ben-
efits in supply chain, finance, and entering new markets.”

MIT SMR CONNECTIONS 3

SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

“Evidence shows that data-based decisions are more likely to be Planners also desire such advanced business-intelligence capabil-
accurate and successful, but many businesspeople — especially ities, but to a lesser extent. Of the planners, 38% are seeking to
those without technology backgrounds — don’t always make use of improve the speed of data-driven decisions or improve products
it,” says Davenport, who is also a fellow of the MIT Initiative on the and services, and 39% are seeking to improve business insights.
Digital Economy. “Generative AI can change that by making data
insights accessible to everyone.”

Generative AI can comb through internal and external databas-

es and retrieve relevant information much faster than executives
or knowledge workers could ever do on their own. And it enables
Early adopters believe that
people to find the answers they need by asking questions in natural
language and exploring results in a conversation, instead of down-
using generative AI for analytics
loading information created by data experts, who may have lacked
has already moved them ahead
the business knowledge to make it helpful in practical situations. of their competitors, and they
“I believe the use of natural language prompts will soon expand expect that gap to widen as they
dramatically,” Davenport says. move forward.

Figure 3:
Generative AI and 51%
Competitive Advantage 46%

How respondents view 37%

their generative AI use in
comparison to the market
and their competitors:* 29%

13%
11%
6%
2% 2% 2%

Far ahead Moving fast; About even Slightly behind Significantly behind
of market and slightly ahead of with market and market and market and
competitors market/competitors competitors competitors competitors

Early adopters Planners

*Results reflect rounding.

4 MIT SMR CONNECTIONS

 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

Gaining a Competitive Edge

Making swift, data-backed decisions is an important competitive “Companies need to be very clear on why they want to use generative
advantage. Early adopters believe that using generative AI for an- AI and determine where it can add the most value,” Ammanath says.
alytics has already moved them ahead of their competitors, and
they expect that gap to widen as they move forward (see Fig. 3). “The possibilities of generative AI are very exciting, but creating
a strategy will yield a better outcome,” agrees Randy Bean, an au-
Thirty-seven percent of early adopters see themselves as far ahead thor, speaker, and advisor to Fortune 1000 companies on data and
of the market and competitors, compared to just 11% of planners. AI leadership. “Organizations should decide which processes they
And just 4% of early adopters consider themselves behind, com- need to change, how to change them, how to develop the skills to
pared with twice as many planners. These data points suggest that do that, and determine a realistic time frame. That doesn’t mean
the longer planners wait to implement generative AI, the harder it slowing down; it means taking a thoughtful approach.”
will be for them to catch up with their peers.
A thoughtful approach also includes addressing the challenges
In terms of industries, respondents most likely to describe them- generative AI introduces. For both early adopters and planners, se-
selves as being far ahead from a competitive standpoint work ei- curity considerations rank as a top concern, cited by 40% of each
ther in technology/telecom (34%) or financial services (33%), with group (see Fig. 4).
others lagging significantly behind.
“You must have safeguards and guardrails in place to ensure that
Early Adopter Characteristic No. 2: you have the right data governance and privacy controls in place
Developing Strategies, Solving Challenges, and deliver customers the right information,” Bean says. (For more
and Tracking Results practical advice, see “Checklist: Getting Started With Generative
Early adopters realize they must align their AI and data strategies to AI for Analytics.”)
their business strategies. That includes developing a coherent strat-
egy for integrating generative AI, with 37% citing strategic consid- Though the responses are similar, strategic challenges are a big-
erations as among the top challenges they must manage (see Fig. 3). ger priority for early adopters than for planners, perhaps because

Figure 4:
Top 5 Challenges
With Generative AI

Early adopters Planners

Security considerations 40% Security considerations 40%

Strategic challenges 37% Data challenges 32%
Model usage/quality concerns 35% Implementation challenges 32%
Data challenges 35% Lack of talent 31%
Implementation challenges 33% Strategic challenges 30%

MIT SMR CONNECTIONS 5

SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

Figure 5:
Positive Outlook for ROI 25%
24%
21%
How respondents
expect their generative 18%
AI investments to pay 16%
15%
off over three years:
12% 12%
11%

2%

ROI of ROI of ROI of ROI of ROI of

300%+ 200-299% 100-199% 75-99% 50-74%

13%

8% 8%

4% 4%
2% 2% 2%
1% 1%

ROI of ROI of ROI No ROI Don’t know/NA

25-49% 10-24% below 10%

Early adopters* Planners

*Percentages reflect rounding.

planners have yet to adopt generative AI strategies. Early adopters Tracking spending and ROI is essential for making progress with
are also more concerned about model usage and quality, with just generative AI, and that’s an area where early adopters are clearly
25% of planners citing that issue as a challenge — and it’s not even ahead. Forty-one percent say they have very robust capabilities in
the planners’ “top 5 challenges” list. place for measuring generative AI’s financial performance, com-
pared with just 15% of planners.
More than a third of both groups are concerned about data chal-
lenges — and rightly so, Davenport says: “Data integration, data Early adopters are also doing more tracking of nonfinancial met-
preparation, and data management are the most important factors rics, including measuring generative AI’s success in increasing the
for being successful with generative AI. Organizations should work speed and accuracy of decisions, improving customer experience
on getting their data in order before getting started.” and satisfaction, and identifying areas in the company that need
improvement. (For an example, see: “Verizon: Tracking Generative
Data modeling and data skills are discussed further under “Early AI Results”)
Adopter Characteristic No. 6,” below.

6 MIT SMR CONNECTIONS

 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

Early Adopter Characteristic No. 3: Verizon: Tracking Generative AI Results

Establishing Bullish Expectations for Tracking results is essential for success with generative
ROI and Revenue Generation AI, but what does that look like in practice? Verizon’s
Many early adopters are seeing impressive results from generative AI experience provides an instructive example.

deployments and have high expectations for ROI and revenue. Nearly
Verizon is especially meticulous about tracking results.
half of early adopters (48%) expect investment returns of 100% or
The telecom giant created a center of excellence to
more over three years, compared to 38% of planners (see Fig. 5).
evaluate potential ROI for all generative AI proposals and
measure the success of projects that get the green light.
Twelve percent of early adopters expect to see an ROI of 300% or more,
compared with just 2% of planners. Almost all the survey respondents “Every use case must go through the center of excel-
expect investments to at least pay for themselves, with fewer than lence to make sure it will produce ROI,” says Anil Kumar
1% of early adopters and just 2% of planners expecting to see no ROI. of Verizon.

Both groups also said they expect revenue to increase as a result For starters, the telecom company, uses a variety of
of generative AI deployments. But compared to planners, more metrics to measure results. “In some cases, generative
than twice as many early adopters expect that increase to be ma- AI creates operational savings or improves productivity.
jor (see Fig. 6). In others, it may generate revenue from a new product
idea,” Kumar says.

Figure 6: Once a project is launched, the company tracks its

Strong Financial Forecast progress over time. For example, a code-generating
application took two or three months for users to get
How respondents expect generative AI accustomed to, but after six months, it is creating a
to impact financial performance:* 10% to 15% improvement in productivity.

42% A content-generating application for marketing is also

39%
getting strong results. “It used to take three to four
35%
weeks going back and forth with an agency to create
29% marketing emails. Now it takes a couple of days,” Kumar
says. The company makes sure that a human reviews

16% the content before sending it to customers — an im-

15%
portant safeguard for the evolving technology.

In one of its largest deployments, the company is mea-

suring how much time a chatbot can shave off the current
Major revenue Moderate revenue Small revenue three-to-six-month training period for its 40,000 cus-
increase increase increase tomer service agents. The next step may be implementing
and tracking a voice recognition system that deciphers
10% customer questions and sends answers to agents, who
3% will no longer need to spend time typing in queries.
2% 2% 1% 1%
“More use cases are popping up in customer service, cus-
No change Small revenue Moderate revenue
decrease decrease tomer relations, sales, network management, and other
areas all the time,” Kumar says. “We are definitely going to
increase our generative AI investments. But on the other
1% 1%
Early adopters side, we are going to see considerable cost savings.”
Significant revenue
Planners decrease

*Percentages reflect rounding.

MIT SMR CONNECTIONS 7

SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

Figure 7:
Companies measure revenue increases in different ways, depend-
Current Business-Data
ing on how it is achieved.
Team Alignment

“We have several calculations for ROI,” says Anil Kumar, vice pres-
How respondents view
ident of data and AI engineering for Verizon. “In some cases, it will their current business-data
be based on productivity gains or operational savings. In other cas- team partnerships:
es, it will be from incremental revenue. Other times, it may be from 75%
a new product idea.”

Generative AI can also boost revenue by improving existing prod- 47% 48%
ucts and services. For example, it can suggest design enhancements
to products, hyper-personalize marketing, improve customer ser-
vice, or suggest more effective sales strategies. All of these efforts 23%
can help the business attract and retain more customers.

2% 5%
When it comes to cutting cost, generative AI can improve produc-
tivity by automating routine tasks in supply chain management, Strong partnership Good partnership Poor partnership
customer support, and other areas, experts say. with centralized with some without centralized
strategy misalignment strategy

Early Adopter Characteristic No. 4: Building

Strong Business-Data Partnerships
To obtain results from generative AI, organizations must develop Early adopters Planners
clear and consistent lines of communication between their busi-
ness teams and data teams, which need to be aligned on a common
execution strategy.

“Generative AI isn’t just an IT or data project. It’s going to be driv-

ing strategic value for the business, so you need to establish an
environment of collaboration, continuous learning, and mutual re-
Early adopters have made much spect,” Ammanath says.
more progress in aligning their
business and data teams, with 75% It’s incumbent on the data team to forge relationships with the
business to drive business value — not the other way around, Bean
saying those teams have strong
says: “If data people don’t have a strong partnership with the busi-
partnerships and a centralized
ness, the organization won’t derive full value from data analytics.”
strategy. Fewer than half (47%) of And unless generative AI is tied to specific sources of business val-
planners said the same. ue, companies run the risk of creating backlash and frustration.

Early adopters have made much more progress in aligning their

business and data teams, with 75% saying those teams have strong
partnerships and a centralized strategy. Fewer than half (47%) of
planners said the same (see Fig. 7).

8 MIT SMR CONNECTIONS

 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

Nevertheless, nearly a quarter of early adopters (23%) admitted “Data teams need to do a better job of explaining their work in
to experiencing some misalignment — a problem faced by 48% terms business teams can understand. And business teams need
of planners. to gain a better grasp of technical complexity so that they can un-
derstand what the data teams are doing,” he says. “If you do both of
Both groups believe their partnerships will improve over the next those things, you’ll narrow the gap.”
year, as generative AI is further incorporated into their data strat-
egies. But while more than half of early adopters expect significant There are many ways of bringing business and data teams closer.
improvement, only about a quarter of planners (26%) shared their For example, Deloitte provides AI fluency training for its employ-
optimism (see Fig. 8). ees. “Generative AI is going to be in every nook and corner of your
business, so it’s important for everyone to understand the basics
Better alignment can’t happen unless both business and data teams and speak the same language,” Ammanath says.
extend themselves a bit beyond their current comfort zones, says
Sam Ransbotham, professor of business analytics at the Carroll Verizon created three different types of generative AI training for
School of Management at Boston College. its workforce: a high-level, daylong session for executives, a more

Figure 8:
Future Business-Data
Team Alignment

How respondents expect their

business-data team partnerships
to change in the future:

58%
52% 58%
52%

38%
38%
26%
26%

11%
8% 11%
8% 5%
2% 5%
2%
Improve Improve Remain Worsen/worsen
Improve
significantly Improve
somewhat Remain
the same Worsen/worsen
significantly
significantly somewhat the same significantly

Early adopters Planners

Early adopters Planners

MIT SMR CONNECTIONS 9

SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

in-depth workshop for mid-level managers, and a two-week deep

dive for developers, to make sure they are grounded in the tools
and platforms they will need to develop use cases. So far, the com-
Data Snapshot:
pany has trained thousands of its more than 100,000 employees,
Different Company Sizes, Different Speeds
and it plans to extend training to the rest over the next year or so.
Midsize-to-large organizations are adopting generative
AI for analytics faster than their larger and smaller counter-
“We realized early on that we needed to bring in the whole organi-
parts, according to the survey.
zation to be successful with generative AI,” Kumar says.

Early Adopter Characteristic No. 5:

Relying Primarily on Third-Party Tools
Company Size by 79%
Early adopters are the keenest on using third-party generative
Percentage of 75%
AI tools for analytics. More than half of that group are current-
Number of Employees
ly deploying and/or planning to adopt such tools, compared with 62% 63%

Percentage of Early Adopters

Early Adopters
32% of planners (see Fig. 9). Fewer than a third of early adopters 56%

100-499: 56%

5,000-9,999
1,000-4,999
500-999: 62%
1,000 to 4,999: 75%

500-999
100-499

10,000+
5,000 to 9,999: 79%
Figure 9:
10,000+ 63%
Generative AI Tools:
Buy or Build?
Company Size by Number of Employees
Tools that respondents are
using or expect to use:

Bigger companies may be out in front simply because

52% they have more resources. But some smaller companies
may catch up quickly, experts say: Typically, they’re work-
41% ing with smaller amounts of data and aren’t hampered by
32% legacy systems.
28% 27%
20% Meanwhile, the largest and smallest businesses clearly
have different reasons for adopting the technology. Early
adopters at the largest businesses identify their top three
priorities as improving business insights (cited by 59%),
improving products and services (54%), and improving
Third-party In-house Combination customer satisfaction (48%).
tools tools of the two

Meanwhile, early adopters at the smallest companies are

also interested in improving products and services (32%),
Early adopters Planners
but they cite as higher priorities improving the speed of
decision-making (36%) and improving productivity (34%).

10 MIT SMR CONNECTIONS

 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

Figure 10:
Generative AI Tools: Key
Capabilities*

Capabilities that respondents

40%
consider most important in
35% 37% 38% 37% 35%
generative AI tools: 34% 32%
31% 31%

Automated Ability to Integration and Developer Data-modeling

surfacing of insights analyze data on interactivity tools assistance
and patterns cloud platform

Early adopters Planners

*Multiple selections allowed

prefer developing in-house generative AI tools for analytics, with vendor, it’s critical to ask about that company’s controls for data
20% using or planning a combination of externally and internally privacy and security, according to experts interviewed for this re-
developed applications. port. Respondents understand the importance of those two issues,
having ranked them as high priorities.
“Buy-versus-build is a crucial decision in the generative AI world,”
Ammanath says. “Building in-house obviously takes more time and Early adopters and planners are looking for similar capabilities
diverts resources that could be going to the core business.” from generative AI tools, including the ability to analyze data to re-
veal patterns and insights and help developers with complex tasks,
That’s especially true for large language models, which require signif- such as data modeling and integrating solutions (see Fig. 10).
icant resources and specialized skills to build and can take months to
train. “Training a large language model from scratch is out of reach “Being able to integrate new tools into existing systems is crucial,”
for most organizations,” Ransbotham says. “The amount of computing Ammanath says. “Teams need to have the skills to ensure seamless
power it takes is overwhelming for all but the largest companies.” application and maximize the tools’ ability to drive business value.”

Still, some companies may want to build highly customized solu- Because generative AI tools are still evolving, companies must also
tions that don’t yet exist in the market, or that require the use of have the flexibility to adapt as new iterations are introduced. Many
highly sensitive data — for example, a patient diagnostic tool. are currently exploring several types of tools for different purposes.

Companies that want to create new solutions can either build Through research, trial, and error, organizations can learn which
them in-house or partner with a vendor. If they decide to use a tools are most effective for their use cases and employees.

MIT SMR CONNECTIONS 11

SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

“With generative AI, it’s more

Early Adopter Characteristic No. 6: Obtaining important than ever to exercise
the Right Skills for Generative AI Deployment human judgment. AI can only do
Both early adopters and planners recognized the need for key
what it’s told, based on teachings
technical skills for creating or customizing generative AI solutions,
and both cited data modeling as the most important (see Fig. 11).
and computer instructions.”
RANDY BEAN
These findings suggest that organizations currently deploying AUTHOR AND ADVISER TO FORTUNE 1000
generative AI are more acutely aware of its technical require- COMPANIES ON DATA AND AI LEADERSHIP
ments. That puts them in a better position to seek out the talent
they need, particularly for high-level data skills, which are often
expensive and hard to come by.
Davenport isn’t surprised by the data-related concerns. “Organiza-
Before they can start using sophisticated data tools, however, or- tions have been struggling with data silos for years,” he notes. Data
ganizations must build a strong data foundation by organizing and inconsistency is another big problem — one that directly affects
consolidating their data, making sure it’s free from errors and in- generative AI accuracy, he adds.
consistencies and available to the people who need it.
“Say you want generative AI to tell you about recent trends in sales
For many companies, that’s a tall order. When asked about their of a detergent product in the southeast U.S.,” Davenport continues.
greatest overall data challenges, early adopters and planners alike “Six different databases may define that region six different ways.
cited — among other issues — costs, lack of needed skills, and data Your data and metadata need to be standardized and accurate to
accuracy, cleanliness, and accessibility. provide meaningful results.”

Data Snapshot:
Different Leadership Roles, Different Outlooks

Not surprisingly, data leaders appear to be further along with using generative AI for analytics than their business counterparts, accord-
ing to the research. Seventy-eight percent of data executives reported that they’re already using the technology, compared to 58% of
business leaders.

In addition, data leaders are clearly more optimistic about the technology’s potential to positively impact their organizations’ financial
performance in the next year. Seventy-eight percent of that group project “major” or “moderate” revenue increases from the use of
generative AI for analytics, compared to 60% of business leaders.

The findings underscore another important theme from the survey: the importance of ensuring that data and business leaders are
aligned in their approach to using generative AI, including how they measure success.

Consultant Eric Siegel, author of The AI Playbook (MIT Press, 2024), says many AI projects fail because leaders overemphasize technical
metrics. “Instead, the focus should be on business metrics — such as revenue, profit, savings, and number of customers acquired,” he
wrote in a recent article for MIT Sloan Management Review.

“These straightforward, salient metrics gauge the fundamental notions of success,” and, he adds: “They’re core to building a
much-needed bridge between business and data science teams.”

12 MIT SMR CONNECTIONS

 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

Keeping a Human in the Loop

Because generative AI is a new technology that can produce un- “It didn’t realize the person was really saying, ‘This offer stinks, and
expected results, it’s also essential that people monitor its output what can be done about it?’” Bean explains.
and make changes as necessary, experts say.
Having humans review AI-generated content provides opportu-
“With generative AI, it’s more important than ever to exercise hu- nities to correct mistakes and avoid problematic use cases. Re-
man judgment,” Bean says. “AI can only do what it’s told, based on searchers are also developing ways to have generative AI source
teachings and computer instructions.” its information and explain its decisions (explainable AI, or XAI),
which will promote trust and enable humans to correct errors or
Even when instructions are correct, programmed responses can misinterpretations closer to the source.
sometimes miss nuances a human would pick up. As an example,
Bean describes a company that tried using generative AI to present In the meantime, as training models incorporate more data, they are
an offer to a job candidate. learning to take on a host of increasingly complex capabilities.

“So that’s the offer?” the displeased candidate asked. “In the future, people will have many AIs performing tasks for
them, giving them time for higher-level reflection,” Bean says. “Ev-
“Yes, that’s the offer,” the AI replied, and went on to fill in the details. erybody will be like a CEO.”

Figure 11: Early adopters

Technical Skills: Top
Priorities*
Data modeling 49%
Build machine learning/predictive models 41%
Skills respondents consider most Natural language processing (generative AI, LLMs) 41%
critical for effective use of generative
AI in analytics: Communication (business goals) 39%
Programming languages 38%
Data governance experience 38%

Planners

Data modeling 49%

Communication (business goals) 37%
Build machine learning & predictive models 36%
Data governance experience 34%
Data storytelling 30%
Natural language processing (generative AI, LLMs) 28%

*Multiple selections allowed

MIT SMR CONNECTIONS 13

SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

Checklist: Getting Started With Data Snapshot:

Generative AI for Analytics Different Places, Different Priorities
Here are some tips that experts suggest to avoid headaches
when using generative AI for analytics and to gain
Survey respondents’ reasons for adopting generative AI for
maximum value from your deployments:
analytics vary somewhat from region to region. Fully 51% of
early adopters based in the Asia-Pacific Japan (APJ) region
3 Align your generative AI strategy with your business
— specifically, those with organizations headquartered in
strategy. Develop and regularly update a timeline for
Australia, New Zealand, India, or Japan — identify improv-
deployments. Focus on business intelligence to drive
ing customer service and satisfaction as a top priority. So
decisions that can generate ROI and increase revenue.
do 43% of those from Europe and the U.K.
3 Establish a center of excellence or dedicated team to
evaluate proposed generative AI projects. This group But 48% of early adopters from the United States and
will examine how such projects might improve Canada cite increasing data-analyst efficiency as the most
productivity, support new capabilities, save money, important reason for adopting the technology. For those
generate revenue, or provide other benefits — and respondents, boosting customer service and satisfaction
track the results. Strive to staff the center or team
ranks a distant second, cited by 35%.
with a representative mix of business, technical, and
data specialists.
The stronger emphasis on data-analyst efficiency may simply
3 Develop a robust measurement framework for gauging reflect the fact that the United States has invested the most
the return on your generative AI investments. in AI over the past five years — more than $548.5 billion in the
past five years, according to Homeland Security Today —
3 Build and maintain strong partnerships between your and is also home to many of the top generative AI companies.
business and data teams. Collaborate regularly on
projects, making sure that everyone on both teams
speaks the same language.

3 Document the technical skills required to run effective

generative AI models and develop training programs
in-house or seek help from partners.
Conclusion: The Path to Generative AI Success
3 Determine what tools are required to best support In the next few years, generative AI will have a profound effect on
generative AI development and deployment: third- organizations in nearly every industry — whether or not they ac-
party, in-house, or a combination of the two. tually use it. Companies that incorporate the technology into their
business decisions stand to gain insights that can help them ex-
3 Eliminate data silos to provide everyone in the
pand to new markets, boost sales, improve products and services,
organization access to those tools.
or create entirely new revenue-generating opportunities that lift
3 Get your data house in order. After all, AI is only as them far above competitors. Some are already realizing these
good as the information you feed it. benefits, making it increasingly difficult for those standing on the
sidelines to catch up.
3 Protect your data with strong privacy, security,
compliance, and access controls.
But rushing into haphazard deployments won’t end well. Instead,

3 Always keep humans in the loop to ensure quality leaders who create strategies to maximize the technology’s value
and accuracy. for their organization, foster collaboration between business and
data teams, employ the right tools and partnerships, and close-
ly track generative AI spending and impact are the most likely to
achieve far-reaching and long-lasting results. l

14 MIT SMR CONNECTIONS

 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

S P O N S O R ’S V I E W P O I N T

Using Generative AI
to Tackle Business
Cindi Howson is an
analytics and business
thought leader and
expert who specializes in

Challenges and
meeting business needs
with technology. As chief
data strategy officer at

Create Real Value

ThoughtSpot, she advises
major clients on data
and AI strategy and best
practices for becoming a
data-driven organization,
influences the company’s
product strategy, and
interviews the industry’s
top data and analytics
leaders on The Data We are at a seminal moment in history when generative AI is poised to transform business
Chief podcast. models and humanity itself. The data and analytics industry has spent a disproportionate
amount of time and money — more than $100 billion annually — to collect data that has, until
now, been locked away in the hands of experts. Generative AI is poised to change that — for
some. The differences between analytics leaders and laggards have never been more stark.

Many of the early generative AI use cases focused on using generative AI for communicating —
for instance, putting bots on top of troves of PDFs and textual data — or for creating content
such as images, slides, and code. Using generative AI on top of structured corporate data
holds promise, but given the constraints of large language models in terms of math, halluci-
nations, privacy protection, and defining specific time periods, there are limitations to how
far this technology can go alone. You may tolerate hallucinations in AI-generated recipes,
but it’s obviously not acceptable when it comes to analyzing sales numbers or employee
performance reviews.

“With generative AI, you have the opportunity

to deliver a data strategy that helps people
solve real-world challenges while delivering
unprecedented value.”

MIT SMR CONNECTIONS 15

SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

Data, digital, and business leaders must unite to seize the gener-
ative AI opportunity. They need to work collaboratively to align
ABOUT THOUGHTSPOT
their data and AI strategies to their organizations’ business strat-
The world’s most innovative companies use AI-powered analytics
egies and prioritize use cases for the technology. ThoughtSpot
fromThoughtSpot to empower everyone in their organizations
sponsored an MIT SMR Connections survey and this resulting
with the ability to ask and answer data questions, create and
report to provide you with industry insights and best practices
interact with data-driven insights, and use those insights to make
to increase your chances of success. The research findings can
informed decisions.ThoughtSpot delivers a complete platform
also help you benchmark your organization’s efforts. for the modern data stack, spanning the spectrum of business
intelligence needs from ad hoc novel analysis with a code-first
With generative AI, you have the opportunity to deliver a data approach to code-free self-service exploration and AI-driven
strategy that helps people solve real-world challenges while monitoring powered by natural language. Companies like
delivering unprecedented value. Fully 48% of early adopters of T-Mobile, CVS Health, Daimler, Capital One, and Huel have
generative AI for analytics expect greater than 100% ROI over turned toThoughtSpot to transform their data-driven cultures.
three years, according to the survey. To learn more, visit www.thoughtspot.com.

But success isn’t inherent. In this quickly evolving era of data

and analytics, it’s easy for fear of failure, hallucinations, and messy
data to sabotage the best intentions. Complacency may feel safer,
but it only furthers the widening gap, jeopardizing both your
business and your career. I encourage you to assess the differ-
ences between the early adopters and those still in the planning
stage and to use that information to guide your own AI-powered
analytics implementation.

16 MIT SMR CONNECTIONS

 MANAGER’S GUIDE —
SURVEY RESEARCH
TRANSFORM YOURREPORT | GENERATIVE
WORKFORCE WITH SKILLS
AI FOR
FORDATA
MACHINE
AND ANALYTICS
LEARNING

Methodology
MIT SMR Connections conducted a global online survey, sponsored by ThoughtSpot, that drew responses from
1,000 data and business leaders from companies of various sizes in a broad range of industries and locations.
Kadence International fielded the survey in the spring of 2024. We examined the data based on respondents’ roles,
geographical locations, company size, and other factors. To provide a rich context for discussion of the quantitative
research results, we interviewed several authors, academics, consultants, and industry practitioners. These individuals
provided insight into current trends and future priorities about the use of generative AI for data and analytics.

Demographic Information
Roles* Top Industries*
All respondents were C-suite, SVP, VP, IT/tech/software/telecom: 34%
or director-level executives. Financial services: 18%
Manufacturing: 8%
Data Leaders: 45% Retail: 7%
Non-Data Leaders:* 55% Construction and retail: 6%
*Non-data leaders include executives in finance, go-to-market, IT,
Health care: 5%
marketing, operations, product engineering, and sales roles. Professional services: 5%
*Remainder divided among other industries
Organization Size*
100-499: 19% Headquarters Location
500-999: 26% U.K., Europe*: 37%
1,000-4,999: 30% U.S., Canada: 31%
5,000-9,999: 16% Australia, New Zealand, India, Japan: 29%
10,000+: 10% Other: 3%
*Results reflect rounding.
*France, Germany, Netherlands, Sweden, Switzerland, U.K.

MIT SMR CONNECTIONS 17

SURVEY RESEARCH REPORT | GENERATIVEMANAGER’S
AI FOR DATAGUIDE
AND ANALYTICS
— TRANSFORM YOUR WORKFORCE WITH SKILLS FOR MACHINE LEARNING

Acknowledgments
MIT SMR Connections is grateful to the following subject matter
experts who shared their insights in interviews for this report:

Beena Ammanath, global head of the Deloitte AI Institute

Randy Bean, author, speaker, and adviser to Fortune 1000 companies on data and AI leadership

Thomas H. Davenport, author, speaker, President’s Distinguished Professor of Information Technology and

Management, Babson College, and fellow of the MIT Initiative on the Digital Economy

Anil Kumar, vice president of data and AI engineering, Verizon

Marc Labelle, global director of the enterprise data office, Ecolab

Sam Ransbotham, professor of business analytics, Carroll School of Management, Boston College

Michael Schrage, author, speaker, and research fellow with the MIT Sloan School of Management’s Initiative on

the Digital Economy

We’d also like to thank the following for their contributions:

Todd Fitz, Fuel Media, graphic designer

Teresa Meek, writer

Christopher Storey, independent data analyst

Kadence International, research partner

18 MIT SMR CONNECTIONS