140 AI product development statistics for 2026
Pilots, production, costs and teams. Every figure checked against its original source.
Overview
Use is widespread. Value is not, yet.
Most companies now use AI. Far fewer can show what it returns. This page collects 140 statistics on that gap: how widely AI is used, how often pilots reach production, what agents can and cannot do yet, what AI costs, how engineering teams work with it, and where commerce, fintech and healthcare stand.
Every number links to its original source, which we opened and checked. We list the population, date and caveats where they change the meaning. Predictions are labelled as predictions. Where a popular figure is often misread, we say so.
Written for founders, CTOs, COOs and product leaders, and for the writers and research tools that cover them. Last updated 20 October 2026.
Key takeaways
01 / AI adoption and scale
AI adoption and scale
Most organizations now use AI somewhere. How many depends heavily on who is counted, and far fewer have scaled it.
Figure 1. AI adoption depends on who you count. Source: McKinsey, with Federal Reserve, US Census Bureau and Eurostat data.
How many organizations use AI
88% of organizations use AI in at least one business function, up from 78% a year earlier.
Adoption is no longer the hard part. McKinsey's 2026 edition puts regular use at nearly nine in ten again.
About 41% of the US workforce used generative AI for work as of November 2025.
Individual use runs ahead of formal company programs, so policy and tooling often trail actual behavior.
78% of the US labor force works at firms that have adopted AI, per the Atlanta Fed Survey of Business Uncertainty.
Weighted by employment, AI already touches most of the workforce, even though most small firms have not adopted it.
73% of companies said they use AI in software development in 2025, up from 66% a year earlier.
Software development is one of the first functions where AI use became routine.
Adoption by size, sector and region
19.8% of US businesses reported using AI in a business function as of May 3, 2026.
Firm-level data shows a far lower rate than executive surveys, because it counts every business, not just large ones.
37% of US firms with at least 250 employees reported using AI, the highest rate of any size class.
Size matters: larger firms have the data, budget and teams to put AI into operations first.
39.7% of US businesses in the Information sector reported using AI as of May 2026, about twice the national rate.
Sectors that already run on software adopt fastest, at roughly twice the national rate.
19.95% of EU enterprises with 10 or more employees used AI technologies in 2025, up 6.47 percentage points from 2024
Europe is moving quickly from a low base, with wide gaps between countries.
55.03% of large EU enterprises used AI in 2025, compared with 17% of small enterprises.
The size gap is just as wide in Europe as in the US.
Marketing or sales was the most common purpose for AI among EU enterprises using it in 2025, at 34.70%.
Early adoption clusters in customer-facing work, where results are easiest to see.
Scaling and high performers
44% of organizations say AI is scaling across their enterprise in 2026, up from 38% a year earlier.
Most organizations still are not scaling, which is why the gap between use and results persists.
About 6% of organizations qualify as AI high performers, and nearly three-quarters of them have fundamentally redesigned workflows because of AI.
The few that see real returns changed how work gets done, not just which tools people use.
34% of organizations are already using AI to deeply transform their businesses, such as creating new products or reinventing core processes.
A minority are using AI to change the business itself, not just to speed up existing tasks.
74% of organizations hope to grow revenue through AI, but only 20% are already doing so.
Revenue is the stated goal for most, but only a fifth have reached it.
In software development, 40% of AI pilots are moving to scale, the highest of any function surveyed.
Engineering pilots scale more often than others because output is easier to measure.
02 / Value and ROI
Value and ROI
Spending and use keep rising. Measured financial value has not kept pace, and the definition of value varies a lot between surveys.
Realized value
Only 13% of CIOs say their enterprise has realized significant or transformational value from AI.
Use is widespread, but significant value remains rare. This is the central gap of 2026
7.5% of companies are classified as future-built for AI and a further 41% are actively scaling it, so nearly half now generate value from AI, up from 5% generating substantial value in 2025.
BCG's newer data shows progress, but more than half of companies still are not generating value.
74% of companies had yet to show tangible value from their use of AI in 2024, and only 26% had the capabilities to move beyond proofs of concept.
The 2024 baseline: most companies were still stuck at proof of concept.
Only 25% of AI initiatives have delivered expected ROI over the last few years, according to CEOs surveyed.
Three in four initiatives miss the return they were approved on.
Only 16% of AI initiatives have scaled enterprise wide, according to CEOs surveyed.
Scaling is the exception. Most initiatives stay inside one team or use case.
Only 6% of executives in Europe and the Middle East report payback on a typical AI use case in under a year; most report satisfactory ROI within two to four years.
Payback on AI usually takes years, not months.
Just 10% of organizations in Europe and the Middle East say they are realizing significant ROI from agentic AI.
Returns from agentic AI are still rare, even where agents are in use.
3 in 4 US enterprise leaders report positive returns on their generative AI investments.
Self-reported returns are much higher than measured ones, so how ROI is defined matters.
Financial impact
37% of organizations attribute at least some EBIT impact to AI use, about the same share as in 2025.
The share seeing any EBIT impact has not moved in a year.
56% of CEOs say they have seen no significant financial benefit from AI to date.
Even at the top, most leaders cannot yet point to clear financial gains.
Only 12% of CEOs say AI has delivered both cost and revenue benefits, while 33% report gains in either cost or revenue.
Winning on both cost and revenue is still uncommon.
Cost pressure and total cost of ownership
58% of CIOs say they face increased pressure to deliver cost improvement from AI.
Boards now expect AI to pay for itself, which raises the bar for every new initiative.
51% of CIOs expect AI to increase the total cost of ownership of technology solutions.
AI adds new running costs, so business cases need to account for them from the start.
Company AI spending grew from about 1.7% of revenue in late 2025 to 3.3% of revenue in 2026.
Spend nearly doubled as a share of revenue, so the pressure to show returns will only grow.
49% of organizations have rephased AI deployments when costs began to outweigh expected value.
Costs are now a reason to slow down, not just a line item.
Only around one-third of organizations have full visibility into their AI operating costs, and those that do are five times more likely to report established ROI (15% vs 3%).
Teams that can see their AI costs are far more likely to prove a return.
72% of US enterprise leaders formally measure generative AI ROI, and 11% fund AI by cutting budgets elsewhere.
Formal measurement is becoming standard, and some budgets are now moving from other programs into AI.
03 / From pilot to production
From pilot to production
The gap between a working demo and a system in production is where most AI investment stalls. The organizations that cross it tend to decide early what not to ship.
Figure 3. Left: enterprise-grade systems only. Source: MIT NANDA, July 2025. Right: S&P Global Market Intelligence, January 2025.
How many projects stall
42% of companies surveyed abandoned most of their AI initiatives before production, up from 17% a year earlier.
Abandonment more than doubled in a year as early pilots met production reality.
Organizations surveyed report that on average 46% of their AI projects are scrapped between proof of concept and broad adoption.
Nearly half of projects stop between proof of concept and broad use.
Only 25% of business and IT leaders surveyed say their organization has moved 40% or more of its AI pilots into production, while 54% expect to reach that level in the next three to six months.
Few organizations have moved even a minority of their pilots into production.
60% of companies surveyed are not achieving material value from AI, while 5% are achieving AI value at scale.
Most companies are not yet getting material value, and only a small group operates at scale. BCG, September 2025
60% of companies surveyed are not achieving material value from AI, while 5% are achieving AI value at scale
Leaders ship most of what they start. Laggards ship very little of it.
Gartner predicted in July 2024 that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025, due to poor data quality, inadequate risk controls, escalating costs or unclear business value.
A widely cited forecast. Its end date has passed, and no outcome has been published.
Stop rules and root causes
Only 31% of CIOs surveyed say their organization has guidelines for when an AI pilot should not move to production, and those that do are 2.3 times more likely to be AI vanguards.
Only 17% of CIOs surveyed say their organization has guidelines for abandoning low-value AI initiatives already in production, and those that do are 2.6 times more likely to be AI vanguards.
Very few teams have a rule for retiring AI that is not paying off once it is live.
RAND interviews with 65 experienced data scientists and engineers found five root causes of AI project failure: misunderstood problems, insufficient data, chasing the newest technology over user needs, weak infrastructure and problems beyond what AI can do.
Most failures start before any code is written, with an unclear problem or missing data.
57% of data leaders surveyed see data reliability as a key barrier to moving AI projects from pilots to production.
Data quality, not model choice, is the most common blocker to production.
The 95% claim, examined
MIT NANDA's preliminary 2025 report states that 95% of organizations are getting zero return from generative AI, based on 52 interviews, 153 senior leader surveys and a review of over 300 public AI initiatives.
This is the source of the widely shared 95% figure. It measures P&L return, not pilot failure, and the report calls itself preliminary.
Of organizations studied by MIT NANDA, 60% evaluated enterprise-grade generative AI systems, 20% reached pilot stage and 5% reached production.
The 5% production figure applies to enterprise-grade custom or vendor systems, not general AI tools.
04 / AI agents
AI agents
Agents are the fastest-moving part of enterprise AI. Interest is broad, deployment at scale is still rare, and many products sold as agents are not.
Figure 4. Organizations by AI agent deployment stage. Source: Capgemini Research Institute, July 2025.
Adoption
62% of organizations are at least experimenting with AI agents: 23% are scaling an agentic AI system and 39% are experimenting.
Agent interest is broad, but most organizations are still experimenting.
In any single business function, no more than 10% of organizations surveyed say they are scaling AI agents.
Agent scaling is still narrow, concentrated in IT and knowledge management.
40% of organizations with more than $1 billion in annual revenue are scaling AI agents in 2026, up from 27% a year earlier, compared with 22% of smaller organizations
Scale is concentrating in large companies. Smaller firms have not moved in a year. McKinsey, August 2026
62% of US organizations with $1 billion or more in revenue are building, deploying or developing AI agents in Q3 2026, up from 53% the previous quarter.
Large US enterprises moved quickly on agents during 2026.
25% of US enterprises with $1 billion or more in revenue are actively developing or implementing multi-agent systems in Q3 2026, up from 6% in each of the prior two quarters.
Multi-agent systems went from rare to common among large US firms within two quarters.
79% of US business executives surveyed say AI agents are already being adopted in their companies.
Executives report broad agent adoption, though definitions of an agent vary widely.
66% of US executives whose companies are adopting AI agents say the agents deliver measurable value through increased productivity.
Where agents are in use, most executives report productivity value.
88% of US executives surveyed say their team or function plans to increase AI-related budgets in the next 12 months because of agentic AI.
Agents are now a main driver of new AI budget requests.
Only 2% of organizations have deployed AI agents at scale and 12% at partial scale, while 23% have launched pilots and 61% are exploring deployment.
Deployment at scale is still rare. Most activity is exploration and pilots.
74% of IT and business leaders expect their companies to be using AI agents at least moderately by 2027, and 23% expect to use them extensively.
Leaders expect agent use to become normal within a year.
Governance and cancellation risk
49% of US business leaders at $1 billion-plus companies have defined high-risk use cases where they do not allow agents to make autonomous decisions.
Many firms now draw explicit lines on what agents may decide on their own.
Executive trust in fully autonomous AI agents fell from 43% to 27% in one year.
Trust in fully autonomous agents fell as teams gained real experience with them.
Only 21% of IT and business leaders say their organization has a mature governance model in place for agentic AI.
Governance is lagging behind adoption for most organizations.
Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027.
Cost and unclear value are the main reasons Gartner expects many agent projects to stop.
Agent washing and real capability
Gartner estimates only about 130 of the thousands of vendors marketing agentic AI are real agentic AI vendors
Many products sold as agents are rebranded chatbots or automation tools.
The top AI agent on the OSWorld computer-use benchmark reached 66.3% accuracy in 2025, up from about 5% in 2024 but still below the 72.35% human baseline.
Agents are improving fast on computer tasks but are still below human level.
In a simulated software company benchmark built by Carnegie Mellon University researchers, the most competitive AI agent completed 30% of realistic workplace tasks autonomously.
On realistic workplace tasks, the best agents still complete only a minority without help.
57.3% of agent builders surveyed by LangChain have agents running in production, and 32% cite quality as a top barrier.
Among builders, quality is the biggest barrier to production.
05 / Build vs buy and AI spend
Build vs buy and AI spend
Enterprise AI spend tripled in a year, and buying has overtaken building. Most of the forecast growth is infrastructure.
Figure 5. US enterprise generative AI spend and share of use cases purchased. Source: Menlo Ventures, December 2025.
Purchased vs built
76% of enterprise AI use cases are purchased rather than built internally, up from 53% purchased in 2024.
Buying has become the default. Building is now reserved for what is core to the business.
Innovation budgets fell to 7% of enterprise LLM spending in 2025, from a quarter a year earlier, as AI moved into centralized IT and business unit budgets.
AI has moved out of experiment budgets and into core IT and business unit spend.
Spend growth
US enterprises spent $37 billion on generative AI in 2025, up from $11.5 billion in 2024.
Enterprise generative AI spend more than tripled in one year.
Global corporate AI investment reached a record $581.69 billion in 2025, a 129.9% increase from the previous year.
Corporate AI investment more than doubled in 2025.
Global private investment in AI reached $344.7 billion in 2025, of which generative AI companies accounted for $170.9 billion.
Generative AI companies took about half of all private AI investment.
In the United States, generative AI accounted for $163.6 billion of private AI investment in 2025, more than half of the total.
The US leads private generative AI investment by a wide margin.
Enterprise CIOs expect their AI budgets to grow by an average of about 75% over the next year.
CIOs planned large budget increases even before most could show returns.
84% of CEOs say AI is a top investment priority over the next three years, and 65% will spend at least 10% of their capital budget on AI in the next year.
CEOs are committing a meaningful share of capital budgets to AI. KPMG, October 2026
58% of CEOs at large companies expect AI to be a major growth engine in the next two years.
Expectations remain high, though only a fifth say AI beat expectations last year.
Gartner forecasts worldwide AI spending will total $2.7 trillion in 2026, a 49.5% increase year over year.
Most of this forecast is infrastructure, not the applications companies use day to day.
Gartner, September 2026
Applications vs infrastructure
Of 2025 enterprise generative AI spend, the application layer captured $19 billion and the infrastructure layer $18 billion.
Gartner forecasts worldwide end-user spending on AI models and platforms will reach $64 billion in 2026, up 63.4% from $39 billion in 2025.
Model and platform spend is growing faster than AI spend overall. Gartner, July 2026
Worldwide AI infrastructure spending totaled $318 billion in 2025, more than double the $153 billion recorded in 2024.
Infrastructure spend doubled in a year, driven by data center buildouts.
The United States accounted for 77% of global AI infrastructure spending in Q4 2025, or $69.2 billion.
The US accounts for most of the world's AI infrastructure spend.
Bain estimates annual spending on AI infrastructure could reach $1.5 trillion by 2031.
A long-range estimate that depends on AI revenue growing several times over.
06 / AI in engineering teams
AI in engineering teams
AI is now part of the working day for most developers. Coding agents became mainstream in 2026, and tool preferences changed within months.
Figure 6. Share of professional developers using each tool at work. Earlier readings: Claude Code and Codex in January 2026, Copilot a year earlier. Source: JetBrains, August 2026.
Daily use
73% of developers who use AI coding assistants or agents use them daily.
For developers who use AI, it is part of every working day.
31% of daily AI users among developers use AI tools 4 or more hours per day.
A large group of developers now spends half the working day or more with AI tools.
Software development professionals who use AI spend a median of 2 hours a day working with it.
AI is now a few hours of each working day for most professionals who use it.
68% of developers say AI tools save them more than 10 hours a week, and 99% report some time savings.
Developers report large time savings, though these are selfreported.
Coding agents and tools
66% of developers who use AI work in coding agents.
Coding agents have become a main way developers use AI.
90% of professional developers used AI coding agents at work at least weekly in May to July 2026, and 68% used them daily.
Agent use is now weekly for almost all professional developers surveyed.
39% of professional developers worldwide used Claude Code at work in May to July 2026, up from 18% in January 2026; in the US the figure was 47%.
Tool preferences changed within months, which makes long tool commitments risky.
GitHub Copilot adoption among professional developers fell from 29% a year earlier to 21% in May to July 2026.
Early leads in AI tooling have not been permanent.
Nearly 80% of new developers on GitHub used GitHub Copilot within their first week on the platform.
New developers start with AI assistance from their first week.
GitHub Copilot's coding agent created more than 1 million pull requests between May and September 2025.
Agents now open pull requests at scale, which moves the bottleneck to review.
More than 1.1 million public GitHub repositories use an LLM SDK, with 693,867 of them created in the past 12 months.
AI features are now a standard part of new software projects.
Measured productivity
Experienced open-source developers took 19% longer to complete tasks when allowed to use AI tools, though they had expected a 24% speedup and afterward believed AI had sped them up by 20%.
Perceived speed and measured speed can differ sharply, especially on large, mature codebases.
Developers given access to an AI coding assistant completed 26.08% more tasks across three field experiments with 4,867 developers at Microsoft, Accenture and a Fortune 100 company.
Large field trials show real gains, larger for less experienced developers.
Google software engineers using AI features completed an enterprise coding task about 21% faster in a randomized controlled trial.
Controlled tests inside one company found a clear but uncertain speed gain.
AI-based code completion assisted with 50% of code characters written by Google software engineers, with a 37% suggestion acceptance rate.
At Google, AI already assists with half of the code characters engineers write.
07 / Quality, trust and delivery risk
Quality, trust and delivery risk
Faster code generation moves the work to review, testing and security. Trust in AI output remains partial, and the data explains why.
Figure 7. How much technology professionals trust AI-generated output. Source: Google Cloud DORA, September 2025.
Trust in AI output
90% of technology professionals use AI at work, yet 30% report little or no trust in the code AI generates.
Use is near universal, but a large minority still does not trust the output.
Only 4% of technology professionals trust AI-generated output a great deal and 20% trust it a lot, while 46% trust it somewhat.
Most professionals trust AI output only partly, which is why review still matters.
46% of developers distrust the accuracy of AI tools, compared with 33% who trust it (2025 data).
Distrust outweighed trust among developers in 2025.
96% of developers do not fully trust that AI-generated code is functionally correct.
Almost no developers take AI-generated code as correct without checking.
Only 48% of developers say they always check their AIassisted code before committing it.
Fewer than half always verify AI-assisted code before it is committed.
Code quality and delivery stability
59% of technology professionals say AI has had a positive impact on their code quality, while 10% see a negative impact.
More professionals see quality gains than losses, by their own account.
DORA estimates that a 25% increase in AI adoption is associated with a 7.2% reduction in software delivery stability and a 1.5% decrease in delivery throughput.
The last quantified estimate: more AI use was linked to less stable delivery without strong controls.
38% of developers say reviewing AI-generated code requires more effort.
Review effort is the hidden cost of faster code generation.
Code block duplication reached 73.0 per million changed lines in 2026 year to date, an 81% increase over 2023 and the highest level on record.
Duplicated code is rising. GitClear links this to AI-assisted coding, though the data is correlational.
Pull requests co-authored with AI contained about 1.7 times more issues than human-only pull requests, and up to 2.74 times more security issues.
In this sample, AI-assisted pull requests carried more issues to review.
Security of AI-generated code
AI-generated code passes security checks only 56% of the time on average, across more than 100 AI models tested over four years.
Nearly half of AI-generated code fails basic security checks.
The best-performing AI model in Veracode's Summer 2026 testing reached a 68% security pass rate, meaning it still failed nearly 1 in 3 security tasks.
Even the best model tested fails about a third of security tasks.
29.5% of Python and 24.2% of JavaScript code snippets generated by Copilot in real GitHub projects contained security weaknesses, spanning 43 CWE categories.
In real projects, about one in four AI-generated snippets had a security weakness.
Even the best model tested, OpenAI o1, produced correct backend code only 62% of the time, and around half of the correct programs from each model could be exploited.
Correct is not the same as secure. Many working programs were still exploitable.
Almost half of the code snippets produced by 5 large language models in a Georgetown CSET evaluation contained bugs, often with security impact.
Bugs with security impact were common across every model tested.
About 40% of 1,689 programs generated by GitHub Copilot across 89 security-relevant scenarios were vulnerable
An early baseline from 2021. Newer studies show lower but still material rates.
08 / Teams, partners and operating models
Teams, partners and operating models
AI changes which roles teams hire for and how work is shared between people, agents and partners. Very few organizations have planned for it.
Figure 8. Share of CIOs surveyed. Source: Gartner, October 2026
Hiring and skills
40% of CIOs and technology executives surveyed expect to increase IT headcount in 2027, particularly in cybersecurity and data and analytics
AI is changing which roles teams hire for, not ending hiring.
63% of employers surveyed identify skills gaps as a major barrier to business transformation over 2025 to 2030.
Skills, not technology, are the top barrier employers name.
Two-thirds of employers surveyed plan to hire talent with specific AI skills, and 40% anticipate reducing their workforce where AI can automate tasks.
Employers plan to hire for AI skills and reduce roles AI can automate, at the same time.
Employment of US software developers, quality assurance analysts and testers is projected to grow 10% from 2025 to 2035, with about 106,100 openings each year.
Demand for developers who build and test software is still growing.
Employment of US computer programmers is projected to decline 7% from 2025 to 2035.
Roles focused only on writing code are shrinking.
AI Engineer was the fastest growing entry-level role in the US in April to June 2026, up 62% year over year.
AI engineering is now the fastest-growing entry point into the field.
In the US, AI Engineers now account for a share of software engineering hires 14 times higher than in 2019 and 10 times higher than in 2022.
AI engineering has become a large share of software hiring.
78% of business leaders surveyed are considering hiring for AI-specific roles, rising to 95% at Frontier Firms.
Most leaders are considering new AI-specific roles.
Role redesign
Only 4% of CIOs surveyed have explicit plans to redesign roles for a human-agent workforce, and organizations that do are 2.4 times more likely to be AI vanguards.
Almost no organizations have planned how people and agents will share the work.
45% of AI users surveyed say it feels safer to focus on current goals than to redesign work with AI, and only 13% say they are rewarded for reinventing work with AI.
Most employees are not rewarded for redesigning their work with AI.
CEOs surveyed expect 29% of employees to require reskilling for a different role between 2026 and 2028, and 53% to need upskilling in their current role.
CEOs expect a large share of their workforce to need new skills within two years.
76% of organizations surveyed have a Chief AI Officer in 2026, up from 26% in 2025.
AI leadership is now a formal executive role in most large organizations.
Educating the broader workforce to raise AI fluency (53%) was the top talent strategy change among leaders surveyed, ahead of redesigning roles or workflows.
Broad AI fluency is the most common talent move, ahead of redesigning roles.
Partners and delivery models
Gartner predicts that by 2028, 70% of enterprises will abandon agentic AI built by vendor forward-deployed engineering, and fewer than 20% of those engagements will turn recurring needs into core product capabilities.
Gartner expects many vendor-built agent projects to stall once the vendor's engineers leave.
83% of executives surveyed are using AI as part of their outsourced services.
AI is already part of most outsourced service relationships.
Global managed services contract value rose 2.1% year over year to a record $11.3 billion in Q3 2026, while engineering and R&D services fell 3% to $814 million.
Buyers are spending more on managed services as engineering services contracts soften.
10.5% of developers surveyed in 2026 work as freelancers or sole proprietors, up from 3.9% a year earlier.
Independent and freelance development work grew quickly in one year.
09 / AI in commerce, fintech and healthcare
AI in commerce, fintech and healthcare
Adoption varies widely by industry. Finance runs well above the national rate, retail below it, and healthcare moved fastest in the last year.
Figure 9. Share of US businesses using AI by sector, May 2026. Source: US Census Bureau BTOS.
Commerce and retail
Traffic to US retail sites from generative AI tools rose 693.4% year over year in the 2025 holiday season (Nov 1 to Dec 31).
AI assistants are becoming a new source of retail traffic.
Generative AI referrals to US retail sites converted 31% better than other traffic sources in the 2025 holiday season.
Shoppers arriving from AI tools tend to arrive ready to buy.
About 14% of US retail trade businesses currently use AI, below the national rate of 19.8%.
Retail trails the national rate, which leaves room for early movers.
45% of consumers surveyed globally use AI for help during their buying journeys.
AI is now part of how many shoppers research and compare products.
AI and AI agents influenced $262 billion, or 20%, of global online holiday sales in 2025.
By Salesforce's own attribution model, AI influenced a fifth of online holiday sales.
Financial services
33.9% of US businesses in the finance and insurance sector use AI, compared with 19.8% of all US businesses.
Financial services adopts AI well above the national rate.
75% of UK financial services firms surveyed were already using AI, with a further 10% planning to within three years.
Most UK financial firms already use AI, under close regulatory attention.
47% of retail and commercial banks surveyed have rolled out generative AI applications, up from 10% in 2023.
Generative AI use in banking rose sharply in two years.
31% of retail and commercial banks surveyed have started implementing agentic AI, though 99% say they are familiar with it.
Agentic AI in banking is past awareness and into early implementation.
64% of financial services professionals surveyed by NVIDIA say AI helped increase annual revenue by more than 5%.
Financial services firms report revenue gains, by their own measure.
Healthcare and digital health
22% of US healthcare organizations have implemented domain-specific AI tools, a 7x increase over 2024.
Healthcare moved from caution to adoption in a single year.
US healthcare AI spending reached $1.4 billion in 2025, nearly triple the 2024 level.
Healthcare AI spend nearly tripled, from a smaller base than other sectors.
81% of US physicians use AI in their practices in 2026, more than double the 2023 rate of 38%.
AI is now part of everyday practice for most US physicians.
71% of US non-federal acute care hospitals used predictive AI integrated with their EHR in 2024, up from 66% in 2023.
Predictive AI inside the EHR is now standard in most US hospitals.
32% of US adults used AI chatbots for health information or advice in the past year.
Patients are using AI for health questions whether providers plan for it or not.
See how we work in commerce, fintech and healthcare.
Our view
What these statistics mean
We build and run AI systems for our clients and for ourselves. Read together, these numbers point to four practical lessons.
From the IV studio
Source: Incredible Visibility internal data, October 2026.
If you are weighing an AI build, talk to us. project@incrediblevisibility.com
Method
Methodology
We reviewed nearly 200 candidate figures and kept 140. Each kept figure was checked against its original publication on 20 October 2026: the page had to load, show the exact number, and match the population and date we describe.
We used government and statistical agencies, peer-reviewed research and web-scale datasets first, then original research with a stated method (such as McKinsey, Gartner, BCG, Deloitte, Stack Overflow, JetBrains and DORA). Vendor data is used sparingly and labelled. We did not cite other statistics roundups.
Dates are the original publication or data release, not the date of news coverage. Where a newer edition exists, it replaces the older one. Four Stanford AI Index 2026 figures are cited by year because the report page does not state a release month. Forecasts are labelled as predictions. Surveys measure different populations, so figures from different sources should not be averaged or treated as one trend.
We refresh this page monthly. Spotted an error or a newer edition? Email project@incrediblevisibility.com
Cite this page
Incredible Visibility. "140 AI Product Development Statistics for 2026: Pilots, Production, Costs and Teams." Last updated October 20, 2026. https://incrediblevisibility.com/research/ai-product-development-statistics
Frequently asked questions
It depends on the study. MIT NANDA found 5% of organizations reached production with enterprise-grade generative AI systems in 2025. S&P Global found 42% of companies abandoned most of their AI initiatives before production, and Deloitte found only 25% have moved 40% or more of their pilots into production.
Not as usually stated. The 95% figure from MIT NANDA's preliminary 2025 report measures organizations getting zero measurable P&L return from generative AI. It is not a measured pilot failure rate.
Mostly buying. Menlo Ventures found 76% of enterprise AI use cases were purchased rather than built internally in 2025, up from 53% in 2024.
Stack Overflow's 2026 survey found 73% of developers who use AI coding assistants or agents use them daily. JetBrains found 68% of professional developers used AI coding agents daily in mid-2026.
Sources
- McKinsey (November 2025). https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/the-state-of-ai-2025
- Federal Reserve Board (April 2026). https://www.federalreserve.gov/econres/notes/feds-notes/monitoring-ai-adoption-in-the-u-seconomy-20260403.html
- Bain & Company (November 2025). https://www.bain.com/insights/executive-survey-ai-moves-from-pilots-to-production
- US Census Bureau (May 2026). https://census.gov/library/stories/2026/05/ai-use-businesses.html
- Eurostat (December 2025). https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises
- McKinsey (August 2026). https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
- Deloitte (January 2026). https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/content/state-of-ai-in-theenterprise.html
- Gartner (October 2026). https://gartner.com/en/newsroom/press-releases/2026-10-19-gartner-survey-finds-only-13-percent-of-cios-saytheir-enterprise-has-realized-significant-or-transformational-value-from-ai
- BCG (September 2026). https://www.bcg.com/press/30september2026-ai-starting-to-pay-off-companies-generate-value
- BCG (October 2024). https://www.bcg.com/press/24october2024-ai-adoption-in-2024-74-of-companies-struggle-to-achieve-and-scalevalue
- IBM Institute for Business Value (May 2025). https://newsroom.ibm.com/2025-05-06-ibm-study-ceos-double-down-on-ai-while-navigatingenterprise-hurdles
- Deloitte (October 2025). https://www.deloitte.com/na/en/issues/generative-ai/ai-roi-the-paradox-of-rising-investment-and-elusivereturns.html
- Wharton (October 2025). https://ai.wharton.upenn.edu/wp-content/uploads/2025/10/2025-Wharton-GBK-AI-Adoption-Report_ExecutiveSummary.pdf
- PwC (January 2026). https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-global-ceo-survey.html
- KPMG (July 2026). https://assets.kpmg.com/content/dam/kpmgsites/cn/pdf/en/2026/07/ai-pulse-q2-2026.pdf
- S&P Global Market Intelligence (January 2025). https://www.amd.com/content/dam/amd/en/documents/solutions/data-center/insights/
- generative-ai-rapid-adoption-mixed-outcomes-vanguard-report.pdf
- Deloitte (March 2026). https://www.deloitte.com/cy/en/about/press-room/state-of-ai-in-the-enterprise.html
- BCG (September 2025). https://media-publications.bcg.com/The-Widening-AI-Value-Gap-October-2025.pdf
- Gartner (July 2024). https://www.gartner.com/en/newsroom/press-releases/2024-07-29-gartner-predicts-30-percent-of-generative-aiprojects-will-be-abandoned-after-proof-of-concept-by-end-of-2025
- RAND (August 2024). https://www.rand.org/pubs/research_reports/RRA2680-1.html
- Informatica vendor data (January 2026). https://www.informatica.com/blogs/cdo-insights-2026-ai-adoption-accelerates-but-trust-andgovernance-lag-behind.html
- MIT NANDA (July 2025). https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
- McKinsey (November 2025). https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2025
- KPMG (September 2026). https://kpmg.com/us/en/media/news/q3-ai-pulse-2026.html
- PwC (May 2025). https://www.pwc.com/us/en/tech-effect/ai-analytics/ai-agent-survey.html
- Capgemini Research Institute (July 2025). https://www.capgemini.com/insights/research-library/ai-agents/
- Deloitte (April 2026). https://www.deloitte.com/us/en/insights/topics/emerging-technologies/ai-agents-scaling-faster.html
- Gartner (June 2025). https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-aiprojects-will-be-canceled-by-end-of-2027
- Stanford HAI AI Index (2026). https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
- Carnegie Mellon University (TheAgentCompany) (September 2025). https://arxiv.org/abs/2412.14161
- LangChain vendor data (June 2026). https://langchain.com/state-of-agent-engineering
- Menlo Ventures (December 2025). https://menlovc.com/perspective/2025-the-state-of-generative-ai-in-the-enterprise/
- Andreessen Horowitz (June 2025). https://a16z.com/ai-enterprise-2025/
- Stanford HAI AI Index (2026). https://hai.stanford.edu/assets/files/ai_index_report_2026_chapter_4_economy.pdf
- KPMG (October 2026). https://kpmg.com/us/en/media/news/kpmg-us-ceo-outlook-2026.html
- EY-Parthenon (January 2026). https://www.ey.com/en_gl/newsroom/2026/01/ceos-double-down-on-ai-transformation-and-m-and-a-to-drivegrowth-amid-uncertainty-in-the-global-economy
- Gartner (September 2026). https://www.gartner.com/en/newsroom/press-releases/2026-09-16-gartner-forecasts-worldwide-ai-spending-togrow-49-point-5-percent-in-2026
- Gartner (July 2026). https://www.gartner.com/en/newsroom/press-releases/2026-07-20-gartner-forecasts-worldwide-ai-platforms-andmodels-market-to-grow-63-percent-in-2026
- IDC (April 2026). https://www.idc.com/resource-center/blog/ai-infrastructure-spending-caps-historic-year-at-90-billion-in-q4-2025-2029-
- spending-to-eclipse-1-trillion/
- Bain & Company (September 2026). https://www.bain.com/insights/new-innovation-is-required-to-fund-ais-6-trillion-buildout-technologyreport-2026/
- Stack Overflow (October 2026). https://stackoverflow.blog/2026/10/06/the-results-of-the-2026-developer-survey-are-here/
- Google DORA (September 2025). https://blog.google/technology/developers/dora-report-2025/
- Atlassian (July 2025). https://www.atlassian.com/blog/developer/developer-experience-report-2025
- JetBrains (August 2026). https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/
- GitHub (October 2025). https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leadstypescript-to-1/
- METR (July 2025). https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
- Cui et al. (SSRN) (August 2025). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4945566
- Google (arXiv) (October 2024). https://arxiv.org/abs/2410.12944
- Google Research (June 2024). https://research.google/blog/ai-in-software-engineering-at-google-progress-and-the-path-ahead/
- Google Cloud DORA (September 2025). https://cloud.google.com/blog/products/ai-machine-learning/announcing-the-2025-dora-report
- Google Cloud DORA (September 2025). https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf
- Stack Overflow (July 2025). https://survey.stackoverflow.co/2025/ai
- Sonar (January 2026). https://www.sonarsource.com/company/press-releases/sonar-data-reveals-critical-verification-gap-in-ai-coding/
- Google Cloud DORA (October 2024). https://cloud.google.com/blog/products/devops-sre/announcing-the-2024-dora-report
- GitClear vendor data (June 2026). https://gitclear.com/the_ai_code_quality_maintainability_gap
- CodeRabbit vendor data (December 2025). https://www.coderabbit.ai/blog/state-of-ai-vs-human-code-generation-report
- Veracode (July 2026). https://www.veracode.com/resources/analyst-reports/2026-genai-code-security-report
- Fu et al., ACM TOSEM (arXiv) (February 2025). https://arxiv.org/abs/2310.02059
- BaxBench (arXiv) (February 2025). https://arxiv.org/abs/2502.11844
- Georgetown CSET (November 2024). https://cset.georgetown.edu/publication/cybersecurity-risks-of-ai-generated-code/
- NYU (Pearce et al., arXiv) (August 2021). https://arxiv.org/abs/2108.09293
- World Economic Forum (January 2025). https://www.weforum.org/publications/the-future-of-jobs-report-2025/digest/
- US Bureau of Labor Statistics (September 2026). https://www.bls.gov/ooh/computer-and-information-technology/software-developers.htm
- US Bureau of Labor Statistics (September 2026). https://www.bls.gov/ooh/computer-and-information-technology/computer-programmers.htm
- LinkedIn Economic Graph (August 2026). https://delivery-p143253-e1476319.adobeaemcloud.com/adobe/assets/
- urn:aaid:aem:ef153078-1061-4817-82e7-a1c027d7a7d7/original/as/AI-Labor-Market-Update-August-2026-v2.pdf
- LinkedIn Economic Graph (January 2026). https://delivery-p143253-e1476319.adobeaemcloud.com/adobe/assets/urn:aaid:aem:0e5ab70aedfd-4bdb-b68f-787fd3ba1a51/original/as/us-software-engineer-talent-landscape-2026.pdf
- Microsoft Work Trend Index (April 2025). https://www.microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-frontier-firm-is-born
- Microsoft Work Trend Index (May 2026). https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-theopportunity-for-every-organization
- IBM Institute for Business Value (May 2026). https://newsroom.ibm.com/2026-05-04-ibm-study-ceos-are-reshaping-c-suite-roles-for-the-aiera
- Gartner (September 2026). https://www.gartner.com/en/newsroom/press-releases/2026-09-29-gartner-predicts-70-percent-of-enterpriseswill-abandon-agentic-ai-built-by-vendor-forward-deployed-engineering-by-2028
- Deloitte (May 2024). https://www.deloitte.com/us/en/services/consulting/articles/global-outsourcing-survey.html
- ISG (October 2026). https://ir.isg-one.com/news-market-information/press-releases/news-details/2026/Infrastructure-Spending-More-ThanDoubles-in-Q3-on-Growing-AI-Momentum-ISG-Index-Finds/default.aspx
- Stack Overflow (October 2026). https://survey.stackoverflow.co/2026/work/data/org-size
- Adobe (January 2026). https://news.adobe.com/news/2026/01/adobe-holiday-shopping-season
- Adobe (January 2026). https://business.adobe.com/blog/ai-driven-traffic-surges-across-industries
- IBM Institute for Business Value and NRF (January 2026). https://newsroom.ibm.com/2026-01-07-ibm-nrf-study-brands-and-retailersnavigate-a-new-reality-as-ai-shapes-consumer-decisions-before-shopping-begins
- Salesforce vendor data (January 2026). https://www.salesforce.com/news/stories/2025-holiday-shopping-data/
- Bank of England and FCA (November 2024). https://www.bankofengland.co.uk/report/2024/artificial-intelligence-in-uk-financialservices-2024
- EY-Parthenon (September 2025). https://www.ey.com/en_us/insights/banking-capital-markets/ai-in-banking-ey-parthenon-genai-surveyinsights
- NVIDIA vendor data (January 2026). https://blogs.nvidia.com/blog/ai-in-financial-services-survey-2026
- Menlo Ventures (October 2025). https://menlovc.com/perspective/2025-the-state-of-ai-in-healthcare/
- American Medical Association (March 2026). https://www.ama-assn.org/press-center/ama-press-releases/ama-ai-usage-among-doctorsdoubles-confidence-technology-grows
- ASTP/ONC (US HHS) (September 2025). https://www.healthit.gov/sites/default/files/2025-09/hospital-trends-use-evaluation-and-governancepredictive-ai-2023-2024.pdf
- KFF (March 2026). https://www.kff.org/health-information-trust/poll-1-in-3-adults-are-turning-to-ai-chatbots-for-health-information-equalingthe-share-who-use-social-media-for-health/
Let's build something that lasts.
Whether you're launching a new product, modernizing a platform, or scaling your engineering organization — we bring the experience, depth, and clarity to help you move forward confidently.