R E PORT

Operationalizing Generative AI
for Better Business Outcomes

Sponsored by
SPONSOR PERSPECTIVE

Beyond boundaries: Innovate, adapt, excel

with generative AI

The most exciting technological advancements merit each headline, every

promise and all the buzz. Generative artificial intelligence (AI) has the
power to transform organizations’ processes, inform decision making and
give an edge over competitors. By introducing generative AI into workflows
and processes, organizations can realize tremendous benefits for their
employees, customers, and bottom line.

Realizing all of AI’s potential and maximizing its value requires careful
preparation. It also takes integration with existing technologies,
widespread employee adoption and a strong commitment and vision
to going beyond basic features like AI assistants and task automation
to enterprise-transforming initiatives. And organizations must properly
manage high-quality data and navigate the complex landscape of AI
regulations to optimize AI’s use.

As we guide organizations through exploring their ideas and aspirations

for ever higher returns on generative AI use case investments, Unisys
is committed to driving innovation and maximizing the impact of AI
across industries.

Also, as we explore the potential of generative AI, it becomes increasingly

clear that its impact extends far beyond initial implementations. Just as
people in the 1970’s, if magically presented with a smartphone, would
likely struggle to conjure all its innovative uses and features (maps, music,
a camera), so, too, are organizations grappling with the possible impacts
of generative AI when it comes to solving business problems. Those that
seize this opportunity by investing wisely can spark innovation, satisfy
regulatory standards and act as pioneers in this thrilling new age of AI.
With the right resources, mindset, strategies, and culture, achieving
business objectives by harnessing generative AI is more than possible.
It’s inevitable.
Operationalizing Generative AI
for Better Business Outcomes

As one of today’s most buzzed-about technologies, HIGHLIGHTS

generative AI (gen AI) promises to reshape entire
industries, revolutionizing the way businesses design
76% of respondents say their
products, connect with customers, and drive employee organization is using, piloting,
or exploring generative
productivity. Health care companies, for example, artifical intelligence (gen AI)
are using gen AI to facilitate early disease detection. for business purposes.

Similarly, pharmaceutical companies are discovering

gen AI’s power to predict results before proceeding 73% expect gen AI to be
with costly clinical trials, while product developers are significantly or extremely
valuable to their organization.
tapping into gen AI capabilities to create innovative
design prototypes.
41% agree their organization is
By leveraging vast amounts of data to create new content such as text, images, realizing a positive ROI from
music, videos, and source code based on inputs or prompts, gen AI has using gen AI.
prompted “a worldwide euphoria,” says Bill Wong, an artificial intelligence
(AI) research fellow and AI and data analytics lead at Info-Tech Research
Group in Toronto. “We’ve entered a new frontier with applications that we’ve
never really seen before.”
To learn more about gen AI’s potential for business transformation, Harvard
Business Review Analytic Services conducted a November 2023 survey of 500
members of the Harvard Business Review audience who are familiar with their
organization’s current state regarding the use of gen AI and decisions about
it. The findings reveal that 73% expect gen AI to be significantly or extremely
valuable to their organization. Overall, 76% say their organization is using,
piloting, or exploring gen AI for business purposes. FIGURE 1 Due to rounding, some figures in this report may not
add up to 100%.
But as with any new technology, organizations need to see value from their
gen AI investments to move beyond early experimentation. In fact, 80% of
respondents believe it’s important to prove an ROI on their gen AI initiatives,

Report | Operationalizing Generative AI for Better Business Outcomes

1
 Harvard Business Review Analytic Services

“We’ve entered a new frontier with applications that we’ve never really
seen before,” says Bill Wong, an artificial intelligence (AI) research
fellow and AI and data analytics lead at Info-Tech Research Group.

built and trained, the huge amount of computing power

required to process billions of user queries, along with the talent
needed to manage them, can turn a promising venture into a
money-losing proposition. Indeed, 39% of those respondents
who are exploring or actively using gen AI report unexpected
costs that were not initially budgeted for their gen AI efforts.
Organizations are now taking a hard look at the cost-benefit
calculus involving gen AI. “With AI, we had two to three years
to demonstrate a return,” says the chief information officer
(CIO) at a European animal food products producer. “With gen
AI, there is a much higher urgency coming from the board to
achieve a return on investment.”
This paper examines the importance of building a pathway
to a return on gen AI investments. It gauges the preparedness
of organizations to tackle AI projects with a clear strategy, a
strong business case, and ROI objectives. It also explores the
cost barriers to achieving financial gains with generative AI
and best practices for achieving greater value from gen AI.

The Challenges of Calculating Costs

It’s easy to understand gen AI’s growing appeal. By generating
text, images, or other data using generative models,
respondents currently exploring or evaluating gen AI say
their organization is realizing benefits, including improved
employee productivity/efficiency (42%), the development of
new capabilities/innovations (34%), and a better utilization
of data (29%). FIGURE 2
The European animal food products maker has seen that
improved productivity and efficiency firsthand because
it’s currently using gen AI for a variety of tasks, including
analyzing procurement documents for critical details, such as
yet only slightly more than half (57%) agree their organization expiry dates on raw materials, to ensure the company meets
is seeing success from their use of gen AI, and just 41% agree regulatory and legal requirements. Other use cases for gen
their organization is realizing a positive ROI from using gen AI. AI include allowing the company’s junior sales associates to
Moreover, when it comes to calculating the hard-dollar instantaneously access information, such as trial research,
value of investing in this nascent technology, gen AI appears that would otherwise require months of training.
to be falling short of expectations. Only 20% of respondents But moving beyond reaping benefits to calculating gen
report successfully reducing operational costs, and a mere AI’s financial value is a challenging endeavor. Unlike cloud
11% say they’ve increased sales/revenue by embracing gen AI. technology, which enables companies to replace high-priced
Perhaps lost in all the excitement about gen AI is the reality on-premises servers with more cost-effective pay-as-you-go
that the cost of developing and maintaining the technology’s pricing structures, many businesses are still figuring out
software can be prohibitively high, offsetting any financial how to leverage gen AI and where it can have the greatest
gains. Even when large language models (LLMs) are properly financial impact.

Report | Operationalizing Generative AI for Better Business Outcomes

2
 Harvard Business Review Analytic Services

“There is a lack of understanding of

the cost implications of generative
artificial intelligence (gen AI) versus
its potential revenue,” says Rui
Lopes, director of new technology
assessment for Elekta AB.

associated with that? This is such a new space that it’s not
well understood.”
Better understanding the costs associated with gen AI
is of utmost importance to Lopes as Elekta increasingly
leverages the technology to improve operational efficiencies
and potentially enhance health services. For instance, Elekta
is exploring how gen AI can automate the compliance of
its products with the rules and guidelines of government
agencies like Health Canada by training models to answer
questions about regulations and flag potential breaches.
“There’s an awful lot of administration that’s required to
substantiate any claims that we make to the FDA [Food
and Drug Administration] or Health Canada,” says Lopes.
“Generative AI can hopefully lessen the burden over time.”
Organizations are also struggling to determine how best
to measure gen AI’s impact on a business. For instance, it’s
relatively easy for a retailer to calculate increased online
sales as a result of using predictive analytics for product
recommendations. Gen AI, on the other hand, is often used
as a tool for crafting marketing messages and creating email
responses—tasks that are tough to put a dollar value on.
“There isn’t always great data on tasks like how many
minutes in a day a customer service agent spends on writing
email responses and, similarly, if they spend 40% less time
writing that email, how much value that creates for the
firm,” says David Greenfield, chief product officer for SOVA,
a Milwaukee-based advisory, services, and product design
company that specializes in AI.
Years of media hype and false narratives regarding how
quickly gen AI can deliver results are also making it hard
“There is a lack of understanding of the cost implications of for organizations to establish accurate expectations of its
gen AI versus its potential revenue,” says Rui Lopes, director impact. Certainly, if executed strategically, gen AI can deliver
of new technology assessment for Elekta AB, a manufacturer impressive ROI. The problem is organizations often overlook
of radiation therapy and radiosurgery-related equipment how educating employees, preparing vast volumes of data,
for the treatment of cancer and brain disorders that is based integrating IT infrastructure, and training gen AI models can
in Stockholm. “The lowest-hanging fruit may be to build take time and seem to delay reaping rewards and realizing a
a chatbot. But what are the cost implications and the risks higher ROI. “It takes time, effort, and money to customize

Report | Operationalizing Generative AI for Better Business Outcomes

3
 Harvard Business Review Analytic Services

these applications to drive value for a given organization,”

says Info-Tech’s Wong.
New and innovative ways for weighing gen AI’s risks
and rewards are emerging. As part of this movement,
organizations are trying to learn the importance of aligning
gen AI goals with business objectives. They are exploring
how best to scale gen AI across an organization to prevent
surprises down the line. They want to figure out how to
reuse core capabilities to help stretch financial resources.
Organizations are also investing in employee training as a
critical part of driving adoption of gen AI tools. Together,
these considerations can empower organizations to properly
gauge gen AI’s financial impact and cut through the hype
for a more realistic assessment of this breakthrough
technology’s value.

The Pursuit of Value

Having tools and methods to measure financial impact are
one thing, but it’s incumbent on organizations to identify
the conditions where they can be put to the best use. The
first step in building such a pathway to a return on gen AI
investments is determining what an organization wants
to achieve and whether these goals properly align with
business objectives.
“One mistake we often see is organizations looking first at
AI capabilities and then thinking about how they could use
them internally,” says SOVA’s Greenfield. “It’s most effective
to think about your business first and what are the areas
that could drive success and then think about how AI could
improve the processes that drive these business needs.”
As it is, organizations’ reasons for investing in gen AI
vary significantly, and those investments have resulted in
varying degrees of success: Among respondents who say their
organization is currently using, piloting, or exploring gen AI,
67% cite improving productivity/efficiency as a key driver
compared to 42% who say it’s a benefit their organization
is currently realizing from gen AI. Forty-five percent say
their organization is motivated to develop new capabilities/
innovations, with 34% achieving this goal. And 32% say can demonstrate “tangible results. We try to have the ROI
better utilizing data is a reason to invest in gen AI; similarly, back within a year to ensure that management is eager to
29% say it’s a reaped benefit. FIGURE 3 A smaller proportion start with the next initiative.” Indeed, among those who are
of respondents say their organization is looking to achieve using or exploring gen AI, 55% agree that their organization
financial gains such as reducing operational costs (30%) and is realizing tangible business value from its use.
increasing sales/revenue (16%). The next step to achieving an ROI on gen AI is developing a
In fact, for all the hype surrounding gen AI’s ability to deep understanding of the costs involved and how these costs
cut costs by replacing human beings with sophisticated might offset any financial gains, such as upticks in employee
algorithms, the European CIO says, “If I go to the board and productivity or improved customer service. For example,
say, ‘Here is our investment in gen AI and this is the savings,’ investments in cloud infrastructure and specialized hardware
it’s a difficult sell for us because we’re a low-margin business such as graphics processors are critical for supporting and
[anyway].” Rather, he says, his organization works hand in training gen AI models. Because LLMs are only as effective
hand with customers to identify high-value use cases that as the data used to train them, organizations must clean

Report | Operationalizing Generative AI for Better Business Outcomes

4
“One mistake we often see is
organizations looking first at AI
capabilities and then thinking
about how they could use them
internally. It’s most effective
to think about your business
first and what are the areas
that could drive success and
then think about how AI could
improve the processes that
drive these business needs.”
David Greenfield, chief product officer, SOVA
 Harvard Business Review Analytic Services

“It’s really about selecting use cases that are scalable. If we can
invent a gen AI application once and then scale it all over the world,
it can bring a lot of benefits for a global company,” says the chief
information officer for a European animal food products producer.

their data by removing random information, inaccuracies, The Tally on Talent

missing values, and bias; identify and fully integrate data Beyond technology expenditures, developing and managing
silos and unstructured data; and ensure data consistency large AI models requires qualified talent—a high-priced
across the enterprise—labor-intensive and complex tasks commodity in today’s competitive labor market. “In today’s
that can inflate the costs of preparing for gen AI. Running market, starting salaries can range from $200,000 to
LLMs, designed to perform billions of calculations at a time, $350,000 for a good prompt engineer with just one year of
also requires huge amounts of costly computing power, not experience,” says Wong.
to mention the cumulative costs of licensing agreements In-house talent must also develop an understanding of the
and monthly subscription fees for gen AI applications intricacies of gen AI systems. “The promise of generative AI
and features. was that people without engineering or technical competence
Complicating matters, says Vaibhav Kumar, senior director would be able to participate in this revolution and that domain
at the Association of Data Scientists in Karnataka, India, experts could control their destiny through the tools created
is the fact that “companies aren’t always sure how much for generative AI,” says Lopes. “But it’s been more difficult
infrastructure they’ll need, such as cloud support and servers, than the [product] brochure suggests.”
until the implementation begins.” To bridge this knowledge gap and get workers up to speed,
Even organizations that take the time to carefully calculate many organizations are investing in employee training. For
the costs of gen AI can encounter unanticipated expenses if example, the European CIO says his company now offers its
they fail to factor in scalability. In these early days of gen AI, employees digital literacy training through the use of fun and
many organizations are experimenting with pilot projects that interactive learning modules so that they “can set direction,
may not consume the same resources when fully deployed. vision, and potentially identify opportunities” for gen AI
One way to control costs is by building on the infrastructure throughout the organization.
and capabilities created during the experimentation By establishing clear objectives, calculating costs, and
phases of a project. For example, the European CIO says, bolstering talent, organizations can begin the tough task of
by establishing a cloud-based foundation for gen AI, his creating a strategy for deploying gen AI in a way that is most
organization can enable multiple use cases with a high ROI likely to deliver ROI.
while managing costs. “Some organizations may find gen AI generates benefits,
“It’s really about selecting use cases that are scalable,” he but unless they develop a plan up front, they will deploy
says. “If we can invent a gen AI application once and then numerous silo-based systems, all unintegrated, because
scale it all over the world, it can bring a lot of benefits for a they lack guidance on where they should invest and how to
global company.” prioritize their investments,” says Wong.
Consider, for example, a retail company that creates a In the case of Elekta, a slow and steady approach to gen
digital assistant to generate automatic responses to customer AI is proving to be the most effective. “First we deploy
inquiries and then reuses its user interface, search capabilities, applications in a controlled environment internally so that
and LLMs to design a tool for investigating shipment delays. we can assess, analyze, and understand how these solutions
In addition to reusing core capabilities for scalability, live and breathe in a simulated environment,” says Lopes.
standardization can also help organizations better calculate Recent proofs of concept at Elekta include exploring how
the costs of gen AI. For instance, Elekta's Lopes says the gen AI can improve efficiencies in preparing tender offers,
company has developed “a standardized way of deploying accelerate and improve writing code for software, and improve
small explorations so that they’re easier to measure” over time. document management.
“It can seem counterintuitive for the innovation process One reason for this careful approach, Lopes asserts, is
but it’s actually very helpful for organizations to have a the fact that gen AI deployments often require “refocusing
real handle on evaluating the value and costs of each of its capacity. We have to stop what we’re doing to do something
applications rather than just a gut feel,” he says. else.” That’s because many of today’s organizations are

Report | Operationalizing Generative AI for Better Business Outcomes

6
 Harvard Business Review Analytic Services

Deploying a gen AI-powered chatbot,

for instance, may require shifting
strategic assets, such as cloud
infrastructure or engineering teams,
from one project to another. Gradually
rolling out gen AI initiatives grants
organizations the time needed to
carefully reallocate and rebalance
these assets.

operating with limited human and technological resources.

As a result, deploying a gen AI-powered chatbot, for
instance, may require shifting strategic assets, such as
cloud infrastructure or engineering teams, from one project
to another. Gradually rolling out gen AI initiatives grants
organizations the time needed to carefully reallocate and
rebalance these assets.
Deciding which projects are most deserving of additional
resources, however, isn’t easy. Fortunately, robust metrics
can aid in the prioritization process, enabling leaders to build
a strong and defensible business case for new investment.
In some instances, these metrics are straightforward. For
example, the European CIO says his organization has long
relied on human agents to translate its employee training
modules “into 15 main languages and sublanguages” for its
nearly 10,000 employees worldwide. With gen AI, however,
the European animal food products producer has managed
to reduce the time it takes to translate the company’s training
manuals “from weeks to minutes”—a boon for employee
productivity as measured by reduced hours spent on mundane
translation tasks.
In fact, 62% of respondents say their organization uses costs by replacing salaried workers with sophisticated
productivity/efficiency gains to assess the performance of gen algorithms. “But what will prevent that from happening in
AI initiatives, followed by reduced operational costs (48%), a hurry is that we don’t really entirely understand where
customer experience/satisfaction improvement (46%), and these technologies can replace people,” says Manav Raj,
increased innovation (41%). FIGURE 4 assistant professor of management at the Wharton School,
In the case of Elekta, Lopes asserts the company’s engineers University of Pennsylvania, in Philadelphia. “It’s a learning
rely on gen AI to create software code for its health care and iterative process.”
products. “Factors such as increased coding efficiency and code In fact, sometimes the decision to hold off on deploying
completion rates have been really useful and have provided the gen AI can be just as costly as any up-front investment.
most tangible metrics of efficiency for gen AI today,” he says. For example, when it comes to using digital assistants, the
But gen AI’s benefits don’t always fit the formula for an European CIO says he’s “not convinced yet that we’ve seen a
obvious ROI. For instance, gen AI can slash operational return on investment. But the question is, ‘If we don’t use it,

Report | Operationalizing Generative AI for Better Business Outcomes

7
 Harvard Business Review Analytic Services

“Leadership should establish responsible AI—the human-based

principles that organizations need to guide their developers in
the development and deployment of this technology to minimize
unintended consequences that may occur,” says Info-Tech Research
Group’s Wong.

are we reducing our efficiency compared to competitors who Cautiously Optimistic

are using the technology?’” Certainly, there are no guarantees that the rewards of using
Organizations could have difficulty rallying team members algorithms to create business value will outweigh the risks.
around gen AI when the payoff isn’t always clear. Strong However, there are signs that significant battles are being won
leadership, according to Greenfield, carries the day in such in the boardroom—for example, presently among respondents
situations. “One of the key success factors we see from using or exploring gen AI, 65% say that the value they get from
leadership is in encouraging gen AI adoption, learning, and investing in gen AI is worth the cost.
experimentation,” he says. “Employees need support in the But reaching beyond experimentation to scaling and
up-front phases where they are testing and learning and not fully operationalizing gen AI systems in ways that deliver
yet reaping the rewards.” continuous and measurable financial returns will require
Good leadership can also activate responsible AI, a growing clear objectives so that gen AI goals are aligned with corporate
movement among organizations using artificial intelligence. policies. Accurate cost calculations that factor in scalability
“Leadership should establish responsible AI—the human- can prevent financial misfires. The right talent can turn gen
based principles that organizations need to guide their AI capabilities into bottom-line benefits. Organizations must
developers in the development and deployment of this also begin to embrace a new perspective on measuring success.
technology to minimize unintended consequences that may Because for all of today’s preliminary wins, the European
occur,” says Wong. CIO says, “When it comes to generative AI, it’s a little bit too
For example, management can foster a corporate culture early to claim victory.”
that prioritizes principles such as fairness, transparency, and
accountability to ensure gen AI models not only deliver value
but do so in a way that empowers employees and positively
impacts society and the environment. Influential managers
can also encourage CFOs to develop a higher tolerance for
indirect future business outcomes from gen AI. Given the costs
associated with launching gen AI initiatives, an immediate
ROI isn’t always possible or even likely. However, CFOs who
are taught it’s acceptable to extend their timeline on returns
could compound their results.
Developing strategies for the deployment and use of gen
AI is becoming increasingly important as expenditures on
everything from IT infrastructure to employee training
increase. While 61% of respondents say their organization
is currently making financial investments toward the use of
gen AI, 70% expect overall financial investment in gen AI to
increase in the next 12 months, making it more important
than ever to find ways to derive value from it.

Report | Operationalizing Generative AI for Better Business Outcomes

8
M E T H O D O LO GY A N D PA R T I C I PA N T P R O F I L E

Harvard Business Review Analytic Services surveyed 500 members of the Harvard
Business Review audience via an online survey fielded in November 2023. Respondents
qualified to complete the survey if they were familiar with their organization’s current
state regarding the use of gen AI and decisions about it.

Size of Organization Seniority Industry Sectors Job Functions Regions

22% 26% 13% 17% 51%

10,000 or more Executive Technology General/executive North America
employees management/ management
board members
12% 23%
32% Education 12% Europe
1,000-9,999 43% HR/training
employees Senior
management 10% 13%
Business/ 9% Asia Pacific
12% professional IT
500-999 20% services
employees Middle 6%
management 9% Latin America
10% Marketing/
21% Financial services communications
100-499 11% 5%
employees Other grades All other functions, Middle East/Africa
9% less than 8% each
Government/
12% not-for-profit 1%
50-99 Other
employees
9%
Health care

All other sectors,

less than 8% each

Figures may not add up to 100% due to rounding.

ABOUT US

Harvard Business Review Analytic Services is an independent commercial research

unit within Harvard Business Review Group, conducting research and comparative
analysis on important management challenges and emerging business opportunities.
Seeking to provide business intelligence and peer-group insight, each report is
published based on the findings of original quantitative and/or qualitative research
and analysis. Quantitative surveys are conducted with the HBR Advisory Council,
HBR’s global research panel, and qualitative research is conducted with senior
business executives and subject-matter experts from within and beyond the Harvard
Business Review author community. Email us at [email protected].

hbr.org/hbr-analytic-services

Copyright © 2024 Harvard Business School Publishing. MCCRE48930424