AI adoption is high. Operating model adaptation is not.
87.7%
say their organization uses AI coding assistants in product work.
36.1%
say AI is strengthening their operating model.
At a glance.
AI tools are now widely present in product and engineering organizations.
The product operating model has not adapted at the same pace.
The biggest gaps appear around strategy clarity, workflow integration, measurement, and governance.
The strongest signal is that AI adoption is ahead of organizational adaptation.
This is not a population-weighted benchmark of every product organization. It is a snapshot from Product Circle, Product Institute, and partner network followers. That likely means the sample over-indexes toward AI-aware leaders.
The tools changed. The work did not.
For the past two years, a lot of the public conversation about AI in product organizations has been a conversation about tools. Which assistant for which task. Which model for which workflow. Whether the latest plugin finally closes the loop. That conversation is not wrong. It is just not the most important one.
In this sample, the simple adoption question is no longer the most interesting one. 87.7% of 309 respondents report their organization uses AI coding assistants in product work. 85.4% report AI tools for product work. 69.9% have shipped AI-powered features in their product. The harder question is whether any of it is changing how work actually moves through the organization.
The five findings below are different facets of one structural fact: the tooling layer ran ahead of the operating model, and the operating model has not caught up.
The tools changed. The work did not.
Finding 1: Tool adoption is high, but transformation confidence is not.
87.7% of 309 respondents report AI coding assistants in product work. 85.4% report AI tools for product work. 71.5% have built internal AI tools or workflows. 69.9% have shipped AI-powered features into their product. Four of the top five adoption activities clear two-thirds of the sample.
Q6 · AI activities undertaken
Tool adoption is everywhere; training and restructuring lag behind.
AI coding assistants (Copilot, Cursor, etc.)87.7%
AI tools for product work (research, writing, analysis)85.4%
Internal AI tools or workflows71.5%
Shipped AI-powered features69.9%
Formal AI training39.2%
Hired AI specialists35.6%
Restructured roles or teams31.1%
Replaced a vendor with an internal AI build19.4%
Other1.3%
None of the above0.6%
Base: 309 respondents. Multi-select. Toggle between segments to see how adoption differs by org size.
The tool adoption headline is fairly uniform across the sample. The training and restructuring numbers underneath it are not. The split by org size shows that 500+ organizations are more likely to have run formal training and restructured roles than 1-50 organizations, but neither size cohort has closed the gap between tooling and the work around it.
Tool adoption is no longer the scarce variable in this sample. The harder question is whether the adoption is changing how work moves. When we asked which activity has produced the most measurable impact so far, 36.2% of 309 respondents named coding assistants and 24.3% said it is too early to tell. The measurable wins are concentrated where engineers benefit, not where customers do.
Q7 · Most measurable impact
Where AI has produced the most measurable impact, so far.
AI coding assistants36.2%
Too early to tell24.3%
AI product tools22.3%
Shipped AI features7.8%
Internal AI tools5.5%
Training1.9%
Hired AI specialists1.3%
Replaced a vendor0.3%
Restructured roles or teams0.3%
Base: 309 respondents. Single select.
The operating-model picture is the counterweight. Only 36.1% of 269 respondents say AI is strengthening their operating model. 22.7% say AI is exposing weaknesses that were already there. 6.3% say it is making things worse. Together, almost a third of respondents are describing an operating model that is in a worse place than before AI arrived.
That is not a story about tools failing. It is a story about tools landing on top of structures that were not redesigned to absorb them. The building layer is where most attention has gone and where adoption is already more mature.
“I would like leadership to be more committed to training everyone. The first six months were vague entreaties to use it more, and then it got a little aggressive. The training we finally got was a series of videos from a leader on one tool.”
Senior Product Manager
That quote captures one shape of the pattern: tools arrived, training did not, and what counted as "adoption" was mostly entreaty without redesign.
Finding 2: The bottleneck has moved upstream.
Q8 · Lifecycle impact ladder
AI impact concentrates in build and design; strategy and feedback loops are the laggards.
Engineering & development50.2%
Design & prototyping45.3%
Documentation & knowledge management39.8%
Ideation & concept development36.6%
Data analysis & experimentation30.1%
Discovery & research29.4%
Strategic planning & roadmapping17.5%
QA & testing15.9%
Customer support & feedback loops12.6%
Collaboration & communication across teams9.3%
Base: 309 for most stages; 269 for collaboration and communication across teams (late-added row).
Half of respondents report High AI impact in engineering and development. Forty-five percent report High impact in design and prototyping. Documentation, ideation, data analysis, and discovery sit in the middle of the range. Strategic planning, QA, customer support, and cross-team collaboration sit in single or low double digits. Build and design are where AI impact is most visible. Strategy and feedback loops are the clear laggards.
That is not subtle. The shape of the ladder is the shape of a familiar product problem: the work closest to making things is the work that is easiest to instrument, easiest to measure, and easiest to accelerate with off-the-shelf tools. The work that lives in conversation, judgment, prioritization, and customer signal is the work that most people have not figured out how to compress with AI yet.
The data also shows that Product Managers say "We don't track AI-specific metrics" more often than C-level respondents (42.9% vs 15.8%). The people closest to the actual product work are also the ones most likely to say their organization is not tracking AI's impact at all.
“Delivery of designs and code got very fast. Delivery of good decisions became the new bottleneck.”
Product Manager
The lifecycle data underwrites the same observation in the quote: delivery accelerated. Decisions did not. The interesting question is no longer how fast the build layer can go. It is whether the judgment layer can keep up.
Finding 3: AI compounds operating-model maturity.
Q10 · AI effect on operating model
How AI is affecting the operating model, by org size and pre-AI maturity.
Strengthening the operating model36.1%
Exposing existing weaknesses22.7%
No noticeable change yet17.8%
Too early to tell17.1%
Making things worse6.3%
Base: 269 respondents. Toggle by org size or operating-model maturity to see the cuts that drive the finding.
The size signal in the toggle view is striking. Smaller organizations are more likely to say AI is strengthening their operating model than 500+ organizations. One plausible read is that smaller organizations have less legacy structure to fight, so AI rewires the operating model rather than running alongside it. Another is that larger organizations apply a higher bar before calling an operating-model change real. Both readings live in the data.
Most organizations did not enter the AI era with a clean operating model. 46.5% of 269 respondents describe their pre-AI operating model as functional but uneven. 28.6% describe it as immature or ad hoc. Only 19.0% describe it as mature and healthy. AI is landing on top of that baseline.
Q5 · Pre-AI operating-model maturity
The baseline AI is landing on: mostly functional-but-uneven.
Functional but uneven46.5%
Immature or ad hoc28.6%
Mature and healthy19.0%
No operating model5.2%
Not sure0.7%
Base: 269 respondents. Late-added question.
The data also shows mature and healthy organizations pick "Time to market" more often than immature or ad hoc respondents (54.8% vs 21.4%). Organizations with a mature pre-AI operating model are dramatically more likely to be tracking time-to-market as a metric AI has improved.
This may be less a pure impact story than a measurement-readiness story: organizations with mature operating models are more likely to know whether AI improved time-to-market at all. Either reading points back to the same question: without the operating muscles to measure, the conversation about AI value cannot get past anecdote.
Q21 · Metrics tracked, by op-model maturity
Mature operating models track more AI metrics; immature ones often track none.
Productivity49.8%
Cost37.6%
Not tracking AI-specific metrics33.6%
Time to market32.8%
Adoption rates30.6%
Quality25.8%
Revenue impact23.2%
Tracking but no clear metrics22.5%
Decision quality9.2%
Base: 271 respondents to the metrics question, split by Q5 operating-model maturity.
The organizations that were better run before AI appear better able to turn AI into outcomes. In this sample AI looks less like an equalizer and more like a multiplier.
“We have budget and little resistance, but no centralized team or strategy to integrate AI into the operating model. We have a lot of team-level initiatives and nothing that moves the needle at the org level.”
Product Manager
That quote names the operating-model story in one respondent's own words: budget present, resistance low, and still nothing that moves the needle at the organizational level, because the work to integrate AI into the operating model has not been done.
Finding 4: Executives and PMs are not seeing the same strategy.
Q11 · Top adoption challenges
Top adoption challenges across the sample.
Unclear ROI50.5%
Integration with existing systems46.0%
Data privacy and security41.1%
No clear AI strategy38.2%
Limited technical understanding21.4%
Budget constraints18.4%
Internal resistance to change12.3%
Build vs buy decisions11.7%
Other8.1%
Base: 309 respondents. Multi-select.
The single largest cross-tab gap in the dataset is not about tools, industry, or geography. It is about role.
Who cites "no clear AI strategy" as a top challenge?
Product managers
61.9%
Gap
42.9pp
Gap between the people closest to strategy-setting and the people expected to translate it into product work.
C-level respondents
19.0%
Base: 309 respondents. Q11 top adoption challenges split by Q2 role.
Product Managers identify "no clear AI strategy" as a top challenge 42.9 percentage points more often than C-level respondents (61.9% vs 19.0%). The careful read matters. The gap may not mean executives have no AI strategy. It may mean that whatever strategy exists at the investment level has not been translated into the operating guidance PMs need at the workflow level: priorities, decision rights, review rituals, metric definitions, what to use AI for, what to deliberately not use AI for.
That gap describes a translation problem, not a comprehension problem. The people closest to strategy-setting and the people expected to translate it into daily product work are reporting different versions of the same organization.
“AI is being implemented bottom-up, but value is realized top-down. Start with a company-level view, translate it into a small number of economic bets, align every team to them, and measure outcomes, not activity.”
Other
That quote names the diagnosis from inside an organization wrestling with the same gap: AI gets implemented bottom-up, value gets realized top-down, and closing the gap means a company-level point of view that travels all the way into team-level workflows.
Finding 5: Leaders fear moving too slow even after adopting the tools.
Q12 · Biggest risk in current AI posture
Moving too slow is the dominant fear, but it is not lonely.
Moving too slowly and falling behind35.3%
Over-investing in tools without changing how the org works27.8%
Moving too fast without guardrails18.8%
Other5.8%
Misaligned leadership4.5%
Knowledge loss as people lean on AI3.9%
Becoming too dependent on a single vendor3.9%
Base: 309 respondents. Single select.
When we asked respondents to pick the single biggest risk in their organization's current AI posture, 35.3% of 309 picked moving too slowly and falling behind. The headline number is the dominant fear, but the nuance is just as important. 27.8% picked over-investing in tools without changing how the organization works. 18.8% picked moving too fast without adequate guardrails. The fear of slowness is dominant; the second-order fears are not lonely.
The framing that AI is a runaway train that needs to be slowed down does not match how product leaders feel about their own organizations. The dominant fear is the opposite: that competitors are extracting more value, and that the cost of inaction is greater than the cost of overreach. Almost twice as many respondents fear being too slow as fear being too fast. The over-investing share is a quiet warning that the second-order risk, of buying without redesigning, is already visible in the data.
Q9 · Response to the EU AI Act
EU AI Act awareness: uneven, and concentrated in Europe.
Not aware of the regulation36.9%
No action taken yet19.4%
Assessing applicability15.2%
Does not apply to us14.9%
Audit complete13.6%
Base: 309 respondents. Toggle by geography to see how awareness differs across the US, Europe, and rest of world.
On the EU AI Act, the headline is sharper than the five raw percentages. 36.9% of 309 respondents are not aware of the regulation. 15.2% are assessing. 13.6% have completed an audit. Not-aware is the largest category, bigger than assessing and completed-audit combined.
A caveat sits underneath those numbers. The EU AI Act can apply outside Europe; "does not apply to us" may be correct for some respondents and a misread for others. This data alone cannot distinguish.
When asked what they would want if they had a "magic wand", this is what a C-level wished for:
“Published security and privacy guardrails for compliance, instead of relying on an AI policy committee to review every proposal ad hoc.”
C-level
That quote points at the unglamorous version of the question: governance that reduces uncertainty tends to speed adoption rather than slow it. Whether that pattern holds in this sample is one of the open questions the report does not try to answer.
In their own words.
Across three open-text questions (the magic-wand question, structural changes leaders want to make, and metric follow-ups), the patterns are remarkably consistent. We surface some representative quotes below.
Leaders want vision and strategy more than tools.
“Our weak spot has always been strategy and focus. Sometimes it feels like leadership thinks AI will let us do everything fast, so we never have to make choices.”
Senior Product Manager
Speed is up, but the bottleneck moved to deciding what to build.
“Engineering is accelerating with no clear thought on how it impacts the rest of the org. The CEO is just throwing more stuff at the wall faster, without a clear strategy.”
Director / Head of Product
Fluency is uneven, and the gap is widening.
“Give product the same tools as engineering. Product is allowed the basic assistant while engineering uses the better one. That creates a gap in quality and speed.”
Senior Product Manager
Roles and team shapes are being rewritten.
“Moving from defined product, engineering, and design roles towards unified builder or maker roles, and rebuilding performance management for that shape.”
Director / Head of Product
The human element will still be there.
“Make senior leaders humble and care about people. If teamwork breaks down in an egoistic race for power inside your own company, and people get scared, the transformation will be very hard.”
Senior Product Manager
Questions for product leaders.
This report stops short of prescription. It does point to a few questions product leaders may need to answer inside their own organizations.
01
Where is AI already changing output, but not yet changing the workflow around that output?
02
Which decisions are now faster, and which decisions have simply become more crowded with information?
03
What would prove that AI is strengthening the operating model, not just increasing tool usage?
04
Where does AI strategy stop being communicated and start being executed?
05
Which AI bets are creating compounding advantage, and which are creating compounding overhead?
Three archetypes we saw in the data.
Three patterns recur across the dataset. They are not formal scores; they are recognizable shapes of how organizations are absorbing AI. Most organizations span more than one archetype. The point is not to box anyone in. It is to give readers a way to locate the work they are doing.
Archetype
Tool adopters
AI footprint is real, but the system around it has not changed. Among extended-section respondents, 33.6% of 271 currently do not track AI-specific metrics; this group is over-represented there.
Archetype
Program builders
Formal training, AI-specific hiring, and restructured roles appear together. AI adoption becomes a program with budget, governance, and review rituals. Typically 500+ org size.
Archetype
Operating-model redesigners
Restructured roles plus shifted team composition. Smaller organizations are over-indexed. AI is treated as a direct change in how work moves, not a workstream alongside the org chart.
These archetypes describe orientations toward AI, not outcomes.
What this is not saying.
This report is not arguing that AI adoption is failing. It is not arguing that product teams need fewer tools. It is observing that, in this sample, the next layer of variance appears to sit in the system around the tools rather than in the tools themselves.
The data here is directional, not predictive. The sample is senior-product-heavy and likely over-indexes toward AI-aware organizations. The correlations have not been adjusted for overlapping variables, so segment effects (industry, role, geography, company size) are not adjusted for one another. The lifecycle and operating-model findings are descriptive of this sample, not causal claims about all product organizations. Where the report uses soft language ("may," "appears," "in this sample") we mean it.
Methodology.
The survey ran from late April 2026 through May 2026 and closed with 309 complete responses. Some of those are core-complete (first 12 questions) and some are extended-complete. Distribution was through Product Circle, Product Institute, and partner channels, weighted toward senior product roles. Three late-added questions (Q1 country and two operating-model questions) are reported on a base of 269 to 272 respondents; other questions use the full base of 309. Country was available for 272 respondents; within that base, the sample covered 40 countries. The full reconstruction lives in Appendix A1.
Appendix.
A1. Methodology in detail.
The survey ran from late April 2026 through May 2026 and closed with 309 complete responses. Distribution was through Product Circle, Product Institute, and partner channels, weighted toward senior product roles. The sample is not population-weighted; it should be read as a directional snapshot of an AI-aware, senior-product audience rather than a benchmark of every product organization.
Three questions were added shortly after the survey launched: Q1 (country), Q5 (pre-AI operating-model maturity), and Q10 (AI effect on operating model). The respondents missing them form a nested coverage pattern, which is the signature of structural form edits rather than respondents skipping questions. Q1 reports on a base of 272 respondents and Q5 and Q10 on 269; every other core question uses the full base of 309. The extended section (Q14 through Q23) has a base of around 271, declining slightly through the section as respondents drop off; each extended question is reported on its own n in Appendix A2.
Two terms in this report are editorial labels rather than survey fields. "Regulated industries" is self-reported by respondents on Q4 and covers fintech, healthcare, legal, insurance, and similar sectors; the label is the respondent's own assessment, not an external classification. "Product leaders" is shorthand for the audience that completed the survey, inferred from the role mix on Q2, not a separate variable.
A2. Full per-question results.
The per-question results are collapsed by default. Expand any question to view its distribution.
Q1. Where are you based?
Base: 272.
United States30.5%
United Kingdom9.2%
Germany8.5%
India5.9%
Portugal5.5%
Canada4.4%
Spain3.7%
Romania3.3%
Brazil2.9%
Australia2.6%
Netherlands2.2%
Ireland2.2%
Switzerland1.8%
Sweden1.5%
Denmark1.5%
Finland1.1%
Norway1.1%
France1.1%
Indonesia0.7%
Belgium0.7%
Poland0.7%
Argentina0.7%
Colombia0.7%
Italy0.7%
New Zealand0.7%
Bangladesh0.7%
Guatemala0.4%
Serbia0.4%
Iran0.4%
Iceland0.4%
Nigeria0.4%
Kenya0.4%
United Arab Emirates0.4%
Austria0.4%
Malaysia0.4%
Chile0.4%
Thailand0.4%
Singapore0.4%
South Africa0.4%
Taiwan0.4%
Q2. What best describes your current role?
Base: 309.
Director / Head of Product25.2%
Senior Product Manager24.9%
C-level (CEO, CTO, CPO, etc.)13.6%
Product Manager13.6%
VP of Product12.9%
Other4.5%
Engineering Leader2.3%
Founder1.6%
Design Leader1.3%
Q3. What is the size of your product and engineering organization?
Base: 309.
11–5026.5%
51–15024.6%
500+18.4%
151–50017.2%
1–1013.3%
Q4. What industry does your company primarily operate in?
Q5. Before your AI adoption journey, how mature was your product operating model?
Base: 269.
Functional but uneven46.5%
Immature or ad hoc28.6%
Mature and healthy19.0%
No operating model5.2%
Not sure0.7%
Q6. Which of the following has your product organization undertaken in the past 12 months?
Base: 309. Multi-select.
AI coding assistants (Copilot, Cursor, etc.)87.7%
AI tools for product work (research, writing, analysis)85.4%
Internal AI tools or workflows71.5%
Shipped AI-powered features69.9%
Formal AI training39.2%
Hired AI specialists35.6%
Restructured roles or teams31.1%
Replaced a vendor with an internal AI build19.4%
Other1.3%
None of the above0.6%
Q7. Which of these has had the most measurable impact?
Base: 309.
AI coding assistants36.2%
Too early to tell24.3%
AI product tools22.3%
Shipped AI features7.8%
Internal AI tools5.5%
Training1.9%
Hired AI specialists1.3%
Replaced a vendor0.3%
Restructured roles or teams0.3%
Q8. Where in the product development lifecycle is AI having the most positive impact?
Base: 309 for most stages, 269 for collaboration and communication across teams (a late-added row).
Engineering & development50.2%
Design & prototyping45.3%
Documentation & knowledge management39.8%
Ideation & concept development36.6%
Data analysis & experimentation30.1%
Discovery & research29.4%
Strategic planning & roadmapping17.5%
QA & testing15.9%
Customer support & feedback loops12.6%
Collaboration & communication across teams9.3%
Q9. How is your organization responding to the EU AI Act?
Base: 309.
Not aware of the regulation36.9%
No action taken yet19.4%
Assessing applicability15.2%
Does not apply to us14.9%
Audit complete13.6%
Q10. How is AI affecting your product operating model?
Base: 269.
Strengthening the operating model36.1%
Exposing existing weaknesses22.7%
No noticeable change yet17.8%
Too early to tell17.1%
Making things worse6.3%
Q11. What are the top challenges your organization faces in adopting AI?
Base: 309. Multi-select.
Unclear ROI50.5%
Integration with existing systems46.0%
Data privacy and security41.1%
No clear AI strategy38.2%
Limited technical understanding21.4%
Budget constraints18.4%
Internal resistance to change12.3%
Build vs buy decisions11.7%
Other8.1%
Q12. What is the single biggest risk in your organization's current approach to AI?
Base: 309.
Moving too slowly and falling behind35.3%
Over-investing in tools without changing how the org works27.8%
Moving too fast without guardrails18.8%
Other5.8%
Misaligned leadership4.5%
Knowledge loss as people lean on AI3.9%
Becoming too dependent on a single vendor3.9%
Q14. How has your product team size changed in the past 12 months?
Base: 272.
About the same26.5%
Smaller26.5%
Shifted composition24.6%
Grown17.3%
Fully restructured5.1%
Q15. How much weight do you give to AI fluency when hiring for product roles today vs. 12 months ago?
Base: 272.
Strong preference43.4%
Core requirement28.3%
Mentioned but not weighted15.1%
Hiring paused7.7%
No change in criteria5.5%
Q16. Has your team built internal tools that replaced something you previously bought or outsourced?
Base: 271.
Experimented but not shipped36.2%
Yes, multiple18.1%
Evaluating17.3%
Yes, one14.4%
No14.0%
Q17. How dependent is your product org on a single AI provider?
Base: 271.
Diversified across providers41.0%
Primary provider, experimenting with others34.3%
Heavily dependent on one provider13.7%
Minimal provider use8.5%
Provider-flexible / abstracted2.6%
Q18. How does leadership currently make decisions about AI initiatives?
Base: 271.
Case-by-case approvals40.2%
Supportive but informal24.7%
Clear strategy and prioritization18.5%
Ad hoc13.7%
Slows decisions down3.0%
Q19. Who owns AI transformation in your organization?
Base: 271.
CPO / CTO owns it29.5%
Distributed across leadership28.8%
Informal18.1%
Dedicated AI / transformation role11.8%
No clear owner10.7%
External advisor1.1%
Q20. How are AI initiatives currently resourced?
Base: 271.
Early agentic work41.3%
Basic automation34.7%
Structured workflows14.8%
Not beyond basic automation6.3%
Advanced agentic work3.0%
Q21. What metrics does your organization currently track to evaluate the success of AI adoption?
Base: 271. Multi-select.
Productivity49.8%
Cost37.6%
Not tracking AI-specific metrics33.6%
Time to market32.8%
Adoption rates30.6%
Quality25.8%
Revenue impact23.2%
Tracking but no clear metrics22.5%
Decision quality9.2%
Q23. Self-rated AI readiness (1 to 10).
Base: 271. Mean 5.40, median 5.0.
7.019.9%
5.018.5%
4.014.4%
6.012.9%
8.012.2%
3.011.8%
2.05.5%
9.02.2%
1.01.8%
10.00.7%
A3. Notable correlations.
The cross-tab miner ranks pairwise findings by effect size (percentage-point gap between the highest and lowest segment, with both segments meeting an n threshold). The findings below are above-threshold results that were not already surfaced in the narrative tier. Correlations have not been adjusted for overlapping variables, so segment effects (industry, role, geography, company size) are not adjusted for one another. Read them as descriptive of this sample, not as independent effects.
Show 20 notable correlations
Q11 (top adoption challenges), by role: Product Manager respondents pick “No clear AI strategy” 42.9pp more often than C-level respondents (61.9% vs 19.0%).
Q1 (country), by role: VP of Product respondents pick “United States” 42.1pp more often than Product Manager respondents (51.4% vs 9.4%).
Q15 (AI fluency in hiring), by industry: Professional services & consulting respondents pick “Strong preference” 38.0pp more often than E-commerce, retail & consumer respondents (58.8% vs 20.8%).
Q18 (leadership decision-making on AI), by industry: E-commerce, retail & consumer respondents pick “Case-by-case approvals” 36.5pp more often than Professional services & consulting respondents (54.2% vs 17.6%).
Q12 (biggest risk), by industry: Professional services & consulting respondents pick “Over-investing in tools without changing how the org works” 36.0pp more often than E-commerce, retail & consumer respondents (54.5% vs 18.5%).
Q21 (AI success metrics), by operating-model maturity: Mature and healthy respondents pick “Time to market” 33.3pp more often than Immature / ad-hoc respondents (54.8% vs 21.4%).
Q18 (leadership decision-making on AI), by industry: Professional services & consulting respondents pick “Clear strategy and prioritization” 32.8pp more often than E-commerce, retail & consumer respondents (41.2% vs 8.3%).
Q6 (AI activities undertaken), by org size: 500+ respondents pick “Hired AI specialists” 31.6pp more often than 1–50 respondents (54.4% vs 22.8%).
Q9 (EU AI Act response), by role: C-level respondents pick “No action taken yet” 30.3pp more often than Senior Product Manager respondents (38.1% vs 7.8%).
Q15 (AI fluency in hiring), by role: VP of Product respondents pick “Core requirement” 30.1pp more often than Product Manager respondents (47.2% vs 17.1%).
Q3 (org size), by role: C-level respondents pick “51–150” 29.7pp more often than Director / Head of Product respondents (47.6% vs 17.9%).
Q7 (most measurable AI impact), by industry: Professional services & consulting respondents pick “AI product tools” 29.7pp more often than Regulated industries respondents (50.0% vs 20.3%).
Q5 (pre-AI operating-model maturity), by industry: SaaS respondents pick “Immature or ad hoc” 29.4pp more often than Professional services & consulting respondents (35.3% vs 5.9%).
Q3 (org size), by industry: E-commerce, retail & consumer respondents pick “151–500” 28.8pp more often than Professional services & consulting respondents (33.3% vs 4.5%).
Q10 (AI effect on operating model), by org size: 1–50 respondents pick “Strengthening the operating model” 28.5pp more often than 500+ respondents (48.1% vs 19.6%).
Q9 (EU AI Act response), by geography: USA respondents pick “Not aware of the regulation” 28.2pp more often than Europe respondents (53.0% vs 24.8%).
Q21 (AI success metrics), by role: Product Manager respondents pick “Not tracking AI-specific metrics” 27.1pp more often than C-level respondents (42.9% vs 15.8%).
Q7 (most measurable AI impact), by geography: USA respondents pick “AI coding assistants” 26.2pp more often than Rest of world respondents (43.4% vs 17.2%).
Q11 (top adoption challenges), by role: Director / Head of Product respondents pick “Integration with existing systems” 25.3pp more often than Product Manager respondents (53.8% vs 28.6%).
Q7 (most measurable AI impact), by industry: Regulated industries respondents pick “Too early to tell” 25.1pp more often than Professional services & consulting respondents (34.2% vs 9.1%).
A4. Bucket definitions.
Org size refers to the "product and engineering organization" size, and is collapsed into three buckets for cross-tab stability. The raw form options 1-10 and 11-50 are combined into 1-50. The options 51-150 and 151-500 are combined into 51-500. The 500+ option is kept as is. The full distributions on the original five buckets are reported in Appendix A2.
Geography is collapsed into three buckets: USA (United States only), Europe (United Kingdom, Germany, Portugal, Spain, Romania, France, Netherlands, Ireland, Switzerland, Sweden, Denmark, Finland, Norway, Belgium, Poland, Italy, Austria, Iceland, Serbia), and Rest of world (all other countries in the dataset). The full country-level distribution is in Appendix A2 under Q1.
Operating-model maturity (Q5) is collapsed from five options into four. Mature-healthy, functional-uneven, and immature-ad-hoc are kept as is. The two low-frequency options (none, not-sure) are combined into "other" to preserve cell counts in cross-tabs.
A5. What this report excludes.
This report does not identify individual respondents. All respondents answered an anonymous survey. Where the qualitative themes section quotes verbatim text, the quotes are screened for company names, product names, and other identifying details before publication. Quotes that could not be safely de-identified are excluded entirely.
This report does not include commercial calls to action. Product Circle and Product Institute are co-publishers; the report's purpose is to document the state of AI in product organizations as honestly as the data allows. Readers will find no gated downloads, no contact forms, and no embedded marketing offers.
This report does not present per-country detail beyond the top of the distribution. Geography is collapsed into three buckets for all cross-tabs to preserve cell counts. The full country-level distribution is in Appendix A2 under Q1 for transparency, but no per-country claims are made in the analysis.
This report does not adjust correlations for overlapping variables. The cross-tab miner reports pairwise gaps. It does not regress one segment variable against another. US respondents are more likely to be VPs in this sample, services respondents cluster by company size, and so on. Read the correlations as descriptive of this sample, not as independent effects.
This report does not auto-update. It is a fixed snapshot of the dataset as of 2026-05-25. Future editions will be published as new snapshots rather than as live revisions to this one.
A6. Definitions.
AI-native product tools. Per the survey wording: "AI tools for product work (research, writing, analysis)." This includes AI-powered research synthesizers, writing assistants, product-analytics tools with embedded AI, prompt-driven artifact generators, and similar product-management-facing tools.
Internal AI tools. Per the survey wording: "Internal AI tools or workflows built by your team." Things teams built themselves that did not exist before, typically replacing manual work, vendor capability, or off-the-shelf integration.
Coding assistants in product work. Per the survey wording: "AI coding assistants (Copilot, Cursor, etc.)." Used by engineering inside the product organization. Adoption is measured at the organization level, not the individual level: 87.7% means 87.7% of respondents report their organization uses these tools in product work, not that 87.7% of individuals use them personally.
Operating model. The set of choices that determine how product work moves: roles, decision rights, rituals, review processes, prioritization mechanics, metrics, and tooling. Distinct from strategy. The report's central observation is that AI has reached the tools but not yet the operating model.
Regulated industries. Self-reported by respondents on Q4. Includes fintech, healthcare, legal, insurance, and similar sectors where compliance materially shapes product work. The label is the respondent's own assessment, not an external classification.
Operating-model maturity (pre-AI). A self-reported pre-AI assessment from Q5. Buckets:
Mature and healthy. Clear strategy, roles, and ways of working.
Functional but uneven. Worked in parts of the org, weaker in others.
Immature or ad hoc. More reactive than intentional.
None / Not sure. Collapsed into "Other" for cross-tab stability.
Transformation ownership. Per the survey question Q19: who owns the AI transformation effort. Options range from CPO/CTO ownership, dedicated AI/transformation role, distributed across leadership, informal, no clear owner, or external advisor. The report treats "dedicated role" and "CPO/CTO" together as concentrated ownership; "distributed" and "no clear owner" as diffuse ownership.
Late-added questions. Three questions added after the survey launched: Q1 (country), Q5 (pre-AI operating-model maturity), and Q10 (AI effect on operating model). Their base is 269 to 272 respondents, not the full 309. All cross-tabs cite the smaller base where these questions are involved. See Appendix A1 for the reconstruction.
How to cite.
Product Circle & Product Institute. (2026). State of AI in Product 2026 [Research report]. https://productcircle.co/state-of-ai-2026
Acknowledgments.
To the 309 product leaders who answered an anonymous survey honestly, and everyone who helped the survey reach a broader audience. Without you this report does not exist.
