Reimagining

Product
Development
with AI

HBR
Analytic Sponsored by
Services
PULSE
SURVEY
Reimagining Product
Development with AI

Many product development teams are turning to artificial HI GHL I GHTS

intelligence (AI) to speed up individual work, automate

repetitive tasks, and analyze customer information for
insights. But because AI is largely being used to achieve
87%
of survey respondents agree that
efficiency gains on an individual level, teams have barely in their organization’s industry,
begun to tap its strategic potential to accelerate and improve having highly effective product
development is critical to business
product development and boost innovation. success.

83%
agree that using artificial
intelligence can significantly
AI H A S T H Epotential to unlock “It’s really, really hard to ensure improve their organization’s
improvements across every stage of that what you learn in your discovery product development.
the product development process— and research phase actually trans-
and doing so will only become more lates into a desirable product that
important because organizations is viable and able to keep pace with
know their current product develop-
ment life cycles need improvement.
product evolution and growth,” says
Sareena Dalla Brookshire, an inde-
34%
In a September 2025 Harvard Business pendent product adviser and the say their organization’s current
Review Analytic Services survey of former chief product officer at Jersey approach to product development
332 members of the Harvard Busi- City, N.J.-based financial solutions is highly effective at producing the
ness Review audience, all of whom are provider BNY Pershing. “Especially if desired results.
knowledgeable about and involved in you’re building something that is truly
making decisions about their orga- innovative.”
nization’s product development, When it comes to product devel-
34% say their organization’s current opment, AI can enhance product
approach to product development groups’ cross-functional collaboration
is highly effective at producing the and iterative creative work, as well as
desired results. Another 16% describe help teams focus on customer needs
the current approach as not very or and make better decisions. In fact, Due to rounding, some figures in this report
not at all effective. the survey shows that organizations may not add up to 100%.

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

1
 Reimagining Product Development with AI

are optimistic about AI, with 83% agreeing that using it can
significantly improve their organization’s product develop-
ment. Yet its full value will ultimately come when companies
leverage it to rethink product development from end to end,
broadly reimagining how teams innovate and collaborate.
This report will illuminate how product development is
evolving in an AI-powered world. It will examine top chal-
lenges with current approaches to product development, the
benefits organizations seek by incorporating AI, and how
they are currently using and considering using AI. It will also
explore key obstacles to overcome in AI adoption and change
management considerations for bringing the technology to
product development.

Common Obstacles Slowing

Product Development
The rate of technological change and evolving consumer
expectations mean that product development teams have
to continually hone their approaches so that their offerings
remain relevant. In addition, having to respond to these
various demands means teams have multiple competing
priorities to balance. “Each phase of the process has its own
unique set of pain points,” explains Brookshire.
Survey respondents report multiple challenges their
organizations have faced when turning a product vision
into a reality. The most-cited ones are limited customer
feedback/insights and lack of strategic alignment across
teams (each cited by 44% of respondents), followed by lack
of cross-team collaboration (i.e., siloed work) and slow pro-
cesses (each cited by 42%). FIGURE 1
One of the biggest factors underlying these challenges is
that it is fundamentally difficult to know what products or
features to build and then to get alignment among stake-
holders. According to Teresa Torres, product discovery
coach at Bend, Ore.-based consultancy Product Talk, this
friction is why cultivating a “product mindset” is critical
for successful teams. A great product both is based on the

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

2
 Reimagining Product Development with AI

The product development cycle should be adjusted

to AI, not just have it layered on top.
Marcel Semmler, chief product officer, Bauer Media Group

company’s competitive positioning and meets the needs of product development. Eighty-one percent also agree that
its target customers. “Not everyone has that mindset, so it implementing AI in product development will help their
can feel like product teams are making arbitrary decisions organization stay competitive with other organizations.
sometimes,” she explains. The survey shows that the most highly sought-after
These dynamics are exacerbated when product devel- benefits of AI in product development are related to speed,
opment decision makers aren’t close to emerging customer cost, and quality. The top two benefits are faster time to
needs. The best kind of leadership, says Torres, is when value (mentioned by 44%) and faster iteration loops (40%),
executives set the strategic direction and empower prod- followed by lower product development costs and improved
uct teams to carry it out. “You’re not just letting product quality of work (both 31%), according to respondents whose
teams do anything they want,” she says. “You’re asking organizations are using or considering using any of the prod-
them to go out and discover the right products given your uct development AI use cases listed.
strategic context.” Unfortunately, Torres adds, that strategic At the same time, unlocking AI’s full capabilities will
approach is still missing at too many companies. require a long-term commitment. Determining the full
Data issues are another common challenge facing extent of what these technologies can do for product devel-
product teams, from siloed information to a lack of tools opment will take curiosity, patience, and a willingness to
or workflows for sharing customer insights effectively. experiment. “Everyone acknowledges AI will change the
Without a single source of truth, product teams and their way things are done, but there’s a wide spectrum of opinion
collaborative partners struggle to make well-informed about to what extent it will erase things or replace things or
decisions. “Data is a huge problem in product development substitute things,” says Brookshire, the independent prod-
organizations because without data, everything you do is uct adviser.
just gut feeling,” says Marcel Semmler, chief product officer According to Bauer’s Semmler, the goal of integrating AI
at Hamburg, Germany-based multimedia conglomerate shouldn’t just be to automate existing processes, which will
Bauer Media Group. “And gut feeling is always worse than produce limited, if useful, gains. Instead, product executives
looking at data.” should consider how to use the technology to rethink entire
Getting product development wrong can have serious workflows. “Look at, from the ground up, how your product
business consequences. Survey respondents report their development cycle works. Where are the flaws? Where are
organizations have experienced myriad negative outcomes the problems?” he says. “The product development cycle
as a result of challenges in their product development pro- should be adjusted to AI, not just have it layered on top.”
cess, including resources (time, money) being used ineffi-
ciently (mentioned by 60%), products being slow to market
(53%), and products not achieving revenue targets (52%).
Finding Early Wins with AI
While redesigning entire workflows with AI should be a
long-term goal, organizations can achieve more immediate
Bringing Artificial Intelligence impact by applying the technology to existing tasks and pro-
into the Mix cesses today.
Many organizations are optimistic about how AI will Steven Eppinger, professor of management science and
enhance their product teams’ work. In the survey, for exam- innovation at the Massachusetts Institute of Technology
ple, 81% agree that AI will accelerate their organization’s (MIT) Sloan School of Management, agrees with taking

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

3
 Reimagining Product Development with AI

a short-term view at first, suggesting executives use AI to

improve the established steps of product development before
trying to reimagine them. As teams learn to harness the
technology, they can expand its use into additional areas. “I
would identify ways to have AI assist at every stage,” he says.
“Over time it’s going to help with more and more parts of the
whole workflow.”
In the survey, respondents were asked about both current
uses of AI in product development and use cases their organi-
zations are considering. The most common current use cases
are creating materials such as product briefs (mentioned by
39%), generating code (35%), and brainstorming product
ideas (33%). The top use cases being considered from the
list are automating processes (47%) and analyzing customer
information/feedback (42%). FIGURE 2
In highlighting AI’s ability to automate work and create
materials, Eppinger says it’s worth investigating which repet-
itive or routine tasks it can be trained to do. Even if handing
a task to AI “merely” saves someone an hour or two, they’re
now freed up for more creative and strategic work. For exam-
ple, putting together a so-called release package—a bundle
of all the changes, assets, and documentation for a specific
product or update—might take an employee hours. Training
an AI system to compile the necessary materials could lead
to considerable time savings for each release. “That kind of
usage might be below the radar of more-strategic thinkers,
but it just saves engineers and developers tons and tons of
time,” Eppinger says.
Beyond increasing productivity, another area where AI
can make an impact right now is in discovery. Discovery has
always been vital, but as Semmler points out, it’s time-con-
suming. “Ideation, A/B testing, running experiments—every
company wants to do all of that, but only the small minority
are really doing it because experimentation can take weeks
to get to a result,” he explains. AI tools, however, can help
discovery efforts happen faster, especially those that involve
a number of participants, such as A/B testing and focus
groups. It does so by shortening the time needed to gather
and interpret user signals, as well as by simulating user

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

4
 Reimagining Product Development with AI

AI tools, however, can help

discovery efforts happen faster,
especially those that involve a
number of participants, such as
A/B testing and focus groups.

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

5
 Reimagining Product Development with AI

“With AI prototyping tools, we can actually to build, let’s just release something,’” she says. But that
approach can create complicated products that are hard to
make the design look like a real product. use and don’t clearly address customer needs. “I think a lot
Customers can interact with it. We can import of companies are going to go through this learning curve in a
their data into it,” says Teresa Torres, product very unpleasant way,” she warns.

discovery coach, Product Talk.

New Mindsets, Team Structures,
and Ways of Collaborating
Evolving product development for the AI age requires a shift
in mindset. Some companies still view products as finite,
behavior, generating test variations, and processing large linear projects with defined starts and ends. But a more for-
volumes of data—helping teams get a faster first read on what ward-looking perspective, explains Torres, treats products
to test with real customers. AI can therefore speed up the time as ongoing efforts and emphasizes gathering feedback, iter-
to insight and improve decision making—helping teams feel ating, and updating products post-launch. In this approach,
more confident that they’re building validated products that which is especially useful for digital offerings, though rel-
will deliver the right business outcomes. evant to physical products too, teams continuously gather
Transforming prototyping is also among the fundamental data about which features are working well or not and use
shifts that AI will enable, says Product Talk’s Torres. Cur- that data for the next iteration.
rently, product teams spend a lot of time mocking up designs Executives also need to consider how roles and team
and walking customers through those designs to demonstrate structures are changing with AI, thanks to enhanced flexi-
how a product might work. “With AI prototyping tools, we can bility. Traditionally, product groups have been based around
actually make the design look like a real product,” she says. what Brookshire calls “siloed roles and responsibilities”:
“Customers can interact with it. We can import their data into Engineers did the engineering, designers did the design-
it.” This method of prototyping can help gather the high-qual- ing, and so on. New technologies are democratizing craft
ity feedback that is essential for designing great products. The and helping people expand their skills. “This is going to be
survey finds that although relatively few organizations are an adjustment as people start to step outside their function
currently using AI for designing prototypes (22%) and testing and think more broadly about the product,” Brookshire
prototype designs (14%), greater numbers are considering says. The blending of traditional roles is likely to continue,
using AI in those areas (36% and 37%, respectively). she adds, changing how executives think about organizing
Still, Torres cautions against getting carried away in the teams and talent.
pursuit of AI-driven efficiency. If teams aren’t careful, trying But as the lines between traditional roles blur, organiza-
to increase speed and lower costs can lead to cutting corners. tions should work to improve cross-functional collaboration,
AI tools are lowering the cost and complexity of building cautions Torres. Often, companies create products with each
products, and Torres has seen that weaken teams’ disci- function handling its own tasks, such as the product devel-
pline for making decisions focused on business goals. “A lot opment team gathering the project requirements, the design
of teams are already starting to frame this as, ‘Why do any team doing the design elements, and the engineering team
discovery? Why get feedback from customers? If it’s cheap building the product. That approach tends to create extra

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

6
 Reimagining Product Development with AI

But as the lines between

traditional roles blur, organizations
should work to improve cross-
functional collaboration.

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

7
 Reimagining Product Development with AI

“It doesn’t really matter whether something Using AI for collaboration can also include integrating it
came from you or the AI or came from the AI as a member of the team. MIT’s Eppinger says this approach
is relatively rare so far given that most organizations still use
and then you embellished it. You’re just using AI as an individual, rather than a team, productivity booster.
the AI in conjunction with your team-based He points to brainstorming as one collaborative task where
the technology could make a real impact. Though a person or
process,” says Steven Eppinger, professor, group can come up with plenty of ideas, AI can come up with
MIT Sloan School of Management. them much faster and at scale. Teams can look at AI’s sugges-
tions alongside employees’ ideas and then feed the best ones
back into the AI tool to iterate further. “It doesn’t really matter
whether something came from you or the AI or came from the
AI and then you embellished it,” Eppinger says. “You’re just
using the AI in conjunction with your team-based process.”
work, she says, because it leads to less effective decision Leveraging AI for teams could also mean deploying it in
making. “You end up asking engineers to build things that spaces where teams work together, such as visual collab-
are less than optimal because they weren’t involved in the oration platforms. These platforms have a digital canvas/
decisions about what to build,” Torres explains. More effec- whiteboard or similar interface to support collaboration and
tive is for key decisions to be made with input from a range information sharing for product teams, sometimes with AI
of functions, which helps ensure that teams are aligned features that automate and organize teams’ work. In the sur-
around common goals. vey, 50% of respondents say their organization uses a visual
collaboration platform for product development, with 81% of
that group describing it as either critical or important to their
product development workflows and processes.
Driving Collaboration and Innovation Another aspect of AI-driven product development is to
Along with accruing quick wins from AI, executives should reimagine current workflows. “The analogy I’d use is classic
also think about how the technology can be used to rede- digital transformation,” says Brookshire. “The best companies
sign collaboration across the entire product develop- didn’t just digitize existing processes—they rethought entire
ment process. processes from start to finish with a digital-first approach. It’s
One critical aspect of doing this is to consider how AI can the same with AI.” There aren’t many established best prac-
be used for team collaboration rather than just for individual tices here yet, and what the reimagining looks like in product
productivity. Without this team-level emphasis, executives development may vary depending on a company’s goals,
are unlikely to realize the technology’s full potential and industry, and other factors. Brookshire suggests product exec-
impact across the organization. Notably, the survey finds utives look for opportunities to redesign and automate work-
that 62% of respondents say AI’s uses in product develop- flows, find deeper customer insights, create more confidence
ment at their organization so far focus primarily on indi- around the products being built, and make delivery roadmaps
vidual work rather than collective work. Only 10% report more accurate. Once products are in the market, she also sug-
that the primary focus is on helping teams collaborate to gests organizations improve their ability to gather data about
get work done, while 20% say AI usage is equally focused on product performance, analyze it, and use it to enhance their
both approaches. processes further.

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

8
 Reimagining Product Development with AI

“The best companies didn’t just

digitize existing processes—they
rethought entire processes from start
to finish with a digital-first approach.
It’s the same with AI.”

Sareena Dalla Brookshire,

independent product adviser and
former chief product officer,
BNY Pershing

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

9
 Reimagining Product Development with AI

I think where AI lacks is not being able to do intuitive

decision making or think outside the box, like looking for
pitfalls or potential roadblocks you might not be able to
preemptively think about.
Akhil Seth, head of product (London office), New York City-based financial
services provider

Taking these steps is no small task, yet it’s likely to be how

AI brings the greatest long-term value to product develop- Current Limits of AI
ment—and will only become more important as executives AI is evolving quickly, yet executives should be mindful
see their competitors exploring these areas. “At the highest of areas where using the technology requires extra care.
level, I do think it’s a restructuring and redesigning of the Chief among them is decision making. At its core, AI is a
entire process,” Brookshire says. pattern-matching technology, and so while it can offer an
The good news for executives, adds Torres, is that as they answer to any question put to it, Eppinger says whether that
integrate AI into their workflows, some areas for improve- answer has any validity depends strongly on the question.
ment become obvious. Among these areas is the creation of He cites the example of ranking 100 product ideas, which
product requirement documents (PRDs), which outline all an expert could do by drawing on their skills and knowledge.
the details and specifications of what will be built. PRDs are Even if someone couldn’t articulate the difference between
standard practice in product development, but the advent of ideas 86 and 83, the person would certainly be right that
AI can make them seem woefully outdated. “That document both are much worse than idea two. “I would trust your judg-
originated in the factory world, where we were trying to ment, because you’ve got all kinds of considerations in your
define six months of work before we started,” Torres points head,” he says. However, what gives Eppinger pause with AI
out. “In a truly agile world, we’re defining two weeks of is that it’s often unclear why AI systems make decisions in
work.” AI can easily produce a 30-page PRD, and teams are the manner they do, even when a person prompts them with
using it to do so, but then other teams use AI to summarize specific criteria. “I’m not saying AI will never be able to make
the document, Torres says, so what’s the point of making it? decisions well, but so far, that’s not the type of thing I want
“What’s weird about product development is we have all to trust it for,” he notes.
these vestiges of this old world that we don’t live in any- In addition, executives should be cautious about relying
more that are still hanging on,” Torres says. “We can ship on AI to perform beyond its programming. The technology
updates to our customers daily. We can iterate as often as we is good at coming up with a result based on data it’s been
need to. It’s much easier to learn from our customers and trained on, but it’s less reliable at making an educated guess.
make improvements, if we are willing. But too many teams An expert could make a logical leap based on their experi-
skip this step. They finish a project and they move on to the ence; AI doesn’t have that capability. “I think where AI lacks
next. They forget to ask, ‘Does this actually work for our is not being able to do intuitive decision making or think
customers?’” outside the box, like looking for pitfalls or potential road-
The survey finds there’s ample opportunity to reimagine blocks you might not be able to preemptively think about,”
the steps of product development, since most steps aren’t says Akhil Seth, head of product in the London office of a
working as well as they could be. The earlier steps and the New York City-based financial services provider.
final one are particularly in need of improvement: Around Other limits of AI are less about its technological power
one-third of respondents point to generating and evaluat- and more about practical considerations—those tasks or
ing product ideas (36%), conducting market research (36%), contexts that people are simply better suited for. When it
product roadmapping and planning (33%), and mapping the comes to customer research, for instance, AI may be useful
customer journey (32%) as underperforming. Just behind for analyzing data, Semmler explains, but even the most
these stages, about a quarter (24%) say the final part of the advanced system can’t replace the person-to-person con-
process, supporting the product post-launch, is one of the versations that can unearth valuable insights. “Going into
steps most in need of improvement. FIGURE 3 your customer’s office for a day to understand how they’re

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

10
 Reimagining Product Development with AI

working and their problems—this kind of closeness will,

for a long time, remain part of product [development]’s
heart,” he asserts. Before handing over critical tasks to the
technology, product executives should be sure they’ve con-
sidered which ones still need and may always require the
human touch.
Another practical consideration is that AI makes mis-
takes, such as inventing fake citations to support wrong con-
clusions. AI tools often warn users to check the technology’s
work, but as long as hallucinations and similar issues persist,
product teams—and organizations overall—will continue to
need human experts to monitor and oversee AI-generated
work. Torres recommends that product development teams
first focus on using AI in domains where their employees
have expertise so they can catch what AI gets wrong and
build an awareness of how to detect its errors. “Keep a really
skeptical mindset when you’re working in a domain you
don’t really understand,” she says, “because then [you’ll] be
in a position to catch what [AI] gets wrong.”

Managing Data and Change

AI adoption must be accompanied by a company-wide effort
around increasing data quality as well as setting up good
governance for data systems and how sensitive information
is used. Improving quality means, at minimum, breaking
down data silos, cleaning data sets, and ensuring teams have
easy access to all the information they need to make good
decisions and collaborate. In addition, while decisions about
who can access data and how it can be shared are not solely
the province of product development, they should be made
by relevant executives, such as those in IT and cybersecurity,
before AI systems are put in place. “From a change man-
agement standpoint, the rules of the road around data and
security are really important,” says Brookshire. Special care
should be taken in industries with stricter data laws, regula-
tory requirements, or privacy concerns, such as finance and
health care, she adds.

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

11
 Reimagining Product Development with AI

The survey finds that the top obstacles to using AI for

product development are data issues and lack of know-
how. The four hurdles mentioned the most all fell into
those categories: data security concerns (cited by 42%
of respondents); lack of technical expertise/staff (41%);
data issues such as silos, low quality, and incompleteness
(39%); and lack of guidance on how to use AI in workflows
(38%). FIGURE 4
When investing in AI, executives would do well to give
everyone throughout the organization the same tools, says
Seth, the London head of product for the New York City-
based financial services provider. If the selection of tools is
left to each individual or team, organizations will quickly
have a disparate set of systems that may not integrate with
each other or give everyone the same capabilities, which
can lead to tool sprawl and technical debt. To build the
workforce’s AI skills equally, Seth says, everyone should
have the same tools, similar to how all employees use the
same application for email. “The more standardization you
do across tools, the more you put people at the same start-
ing point,” he explains.
Ultimately, AI alone won’t make a product team success-
ful, argues Eppinger, who says a company’s competitive
differentiator is not its technology but the teams that use
it—something executives shouldn’t forget. “We all have
access to more or less the same approaches and tools,” he
points out. “So having really good staff and thinking about
innovation right and having a good strategy—I think those
things will differentiate us when we’re all using AI.”

HARVARD B USI N ES S R E V I EW A N A LY TI C SE RV I CES

12
 ME T H OD O LO GY A N D PA RTIC IPA N T P ROF IL E

Harvard Business Review Analytic Services surveyed 332 members of the Harvard Business Review
audience via an online survey fielded in September 2025. Respondents qualified to complete the
survey if they work at an organization that does product development, they’re knowledgeable about
the product development process, and their role is involved in making decisions about it.

ORGANIZATION SIZE SENIORITY JOB FUNCTIONS

28% 34% 31%

10,000 or more employees Executive management/board members R&D/innovation/product development

28% 39% 25%

1,000–9,999 employees Senior management General/executive management

10% 16% 9%
500–999 employees Middle management IT

23% 11% All other functions less than 7% each.

100–499 employees Other grades

11%
50–99 employees

INDUSTRIES REGIONS

21% 42%
Manufacturing North America

19% 29%
Technology Europe

15% 18%
Financial services Asia Pacific

8% 6%
Health care Latin America

6%
All other sectors less than 8% each.
Middle East/Africa
 VISIT US ONLINE

[Link]/hbr-analytic-services

Harvard Business Review Analytic Services is an independent

commercial research unit within the 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
hbranalyticservices@[Link].

CRE73700326 Copyright © 2026 Harvard Business School Publishing.