Almost every company uses AI. Almost none have made it run.
This is the data on the gap that defines 2026: the small group of teams turning AI into reliable, running systems, and the majority still doing the work by hand. More than 50 statistics, each one sourced, pulled from over 20 primary reports.
There is a number that quietly explains most of what is happening with AI at work right now. 88% of organizations use it. 7% have actually scaled it.
Everything else on this page is a footnote to that gap. The story of 2026 is not adoption. Adoption is a solved problem, near-universal, climbing every quarter. The story is the distance between a company that has AI switched on somewhere and a company that has turned it into a system that runs on its own, reliably, without someone babysitting every output.
We call that distance the AI divide. On one side are the teams that got AI to do real work: agents that run, results they trust, hours they got back. On the other side, the larger group, are the teams still copying from one tab and pasting into the next, now with a chatbot open in a third. They adopted AI. They just never got it to do the job.
What follows is the evidence for where that line sits and why. Six sections, 51 statistics, every one traced to a named source. Read it as a map of who is winning the AI divide, who is stuck, and what separates the two. The short version: the winners run tested systems with a human keeping watch. The stuck ones handed the keys to an agent and hoped.
Key takeaways
- 88% of organizations now use AI in at least one function. Only 7% have fully scaled it. The adoption-to-impact gap is the whole story of 2026. (Source: McKinsey, State of AI 2025)
- Trust in fully autonomous AI agents fell from 43% to 27% in a single year. Reality arrived the moment agents met production. (Source: Capgemini Research Institute, 2025)
- Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027, blaming unclear value, runaway cost, and weak risk controls. (Source: Gartner, 2025)
- Workers lose an average of 17.3 hours a week to mindless, easily automatable tasks. That is close to half the work week, gone. (Source: Zapier, 2025)
- Only about 5% of companies - BCG's "future-built" leaders - systematically generate value from AI. Nearly 60% report little or no impact at all. (Source: BCG, 2025)
01AI and automation adoption
Use is near-universal. Maturity is not.
Read the adoption headlines and you would think the transformation already happened. 88% of organizations use AI somewhere. 72% touch generative AI. Nearly two-thirds are experimenting with agents. Those numbers are real, and they climb every quarter.
Then you look at the second column. The share of companies that have actually scaled AI across the organization sits at 7%. In any single business function, no more than one in ten say they are scaling agents. That is the shape of the divide in one chart: a tall bar of curiosity, a short bar of results. Adopting AI turns out to be easy. Getting it to run the work is the hard part, and it is where almost everyone is still stuck.
The adoption-maturity gap
Share of organizations at each stage, from trying AI to running it at scale.
of organizations report regularly using AI in at least one business function, up from 78% a year earlier.
(Source: McKinsey, State of AI 2025)of organizations report using generative AI, up from just 33% in 2024. The curve is steep.
(Source: McKinsey, State of AI 2025)are scaling an agentic AI system somewhere in the enterprise; another 39% have started experimenting with agents.
(Source: McKinsey, State of AI 2025)In any given business function, no more than one in ten organizations say they are scaling AI agents.
(Source: McKinsey, State of AI 2025)- Only 7% of respondents report that AI is fully scaled across their organization. The other 93% are somewhere on the runway. (Source: McKinsey, State of AI 2025)
- 87% of sales organizations already use some form of AI, and 54% of sellers say they have already worked alongside an AI agent. (Source: Salesforce, State of Sales)
- 66% of marketers globally use AI in their roles, and 91% of marketing leaders say their teams use AI to assist them. (Source: HubSpot, State of AI)
- 43% of organizations now use AI in HR, up from 26% in 2024, led by recruiting at 51%. (Source: SHRM, 2025 Talent Trends)
- 73% of companies use AI in software development, up from 66% a year earlier. (Source: Bain and Company, 2025)
- 74% of companies now rank AI as a top-three strategic priority, up from 60% in 2024. (Source: Bain and Company, 2025)
- 46% of leaders say their companies are already using agents to fully automate workflows or processes. (Source: Microsoft, 2025 Work Trend Index)
The curve of people trying AI and the curve of people running it have split apart. Which side of that split you land on is the only question that matters now.
On the 81-point gap between 88% adoption and 7% scale.
02The access gap: who gets to automate
Demand is coming from the business teams. The tools were built for developers.
Here is the part most adoption surveys miss. Automation is not just spreading, it is moving toward the people closest to the work, the ones in sales, marketing, HR, and operations who feel the manual grind every day. But almost every tool built to automate that work still assumes you can think like an engineer: nodes, flowcharts, integrations, logic you have to wire by hand.
That mismatch is the access gap. IT is more than ten times likelier to lead an AI push than any of the teams actually drowning in the busywork. The demand lives with business teams. The supply was designed for the department that already had technical skill. Bridge that gap and you scale AI. Leave it, and you get exactly what the data shows: a handful of companies pulling value while the majority watch their tools sit unused.
of business leaders report AI-critical skill shortages in their workforce today.
(Source: World Economic Forum, 2025)By 2026, citizen developers at large enterprises will outnumber professional developers four to one.
(Source: Gartner, via Kissflow)By 2026, low-code tools will account for three-quarters of all new application development.
(Source: Gartner, via InfoWorld)- 94% of business leaders report AI-critical skill shortages in their workforce today, and the skills gap is the single biggest barrier to transformation, cited by 63% of employers. (Source: World Economic Forum, 2025)
- 78% of enterprises struggle to integrate AI with existing systems, and 35% point to AI skill gaps among employees as a barrier. (Source: Zapier AI Resistance Survey, 2025)
- By 2026, more than 90% of organizations will feel the IT skills shortage, adding up to roughly $5.5 trillion in losses from delays, quality issues, and lost revenue. (Source: IDC FutureScape, via CFO Dive)
- By 2026, developers outside formal IT will make up at least 80% of the low-code user base, up from 60% in 2021. (Source: Gartner, via InfoWorld)
- By 2029, low-code platforms will power 80% of the world's mission-critical applications. (Source: Gartner, via Kissflow)
- Only about 5% of companies (BCG's "future-built" leaders) systematically generate value from AI, while nearly 60% report little or no impact. (Source: BCG, The Widening AI Value Gap, 2025)
- "Future-built" firms expect twice the revenue increase and 40% greater cost reductions than the laggards. (Source: BCG, 2025)
- Just 39% of organizations report any enterprise-level profit impact from AI, and only about 5.5% attribute more than 5% of operating profit to it. (Source: McKinsey, State of AI 2025)
03The cost of manual work
The demand for automation is not abstract. It is measured in hours.
Ask why business teams want automation and the honest answer has nothing to do with AI. It is the hours. The report someone rebuilds every two weeks. The list an SDR exports every Monday. The CVs an HR manager reads and summarizes by hand, one after another. Work a machine should have been doing years ago.
The numbers here are the fuel behind the whole adoption wave. Workers give up an average of 17.3 hours a week to tasks they describe as mindless. Sales reps spend barely a third of their day actually selling. This is the pain the winning side of the AI divide got rid of, and the pain the stuck side is still paying in full, every single week.
Hours lost per week to manual, repetitive work
Three different measures, three different studies, one consistent verdict.
- Employees spend an average of 17.3 hours a week on mindless, repetitive tasks that are easily automated, almost half the work week. (Source: Zapier, 2025)
- Sales reps spend only 30% of their time actually selling. The rest goes to admin, data entry, and internal busywork. (Source: Salesforce, State of Sales)
- 62% of the workday goes to repetitive, mundane tasks rather than skilled, strategic work. (Source: Asana, Anatomy of Work)
- Over 40% of workers spend at least a quarter of their week on manual, repetitive tasks. (Source: Smartsheet)
- Nearly 60% of workers say they could save six or more hours a week, almost a full workday, if the repetitive parts of their job were automated. (Source: Smartsheet)
- Office workers spend more than three hours a day on manual, repetitive computer tasks, about 60 hours a month. (Source: Automation Anywhere / OnePoll)
- Marketing teams spend an average of 14.5 hours a week managing and collecting customer data. (Source: Treasure Data, via Coupler.io)
- Employees are interrupted every 2 minutes during core work hours, roughly 275 times a day. (Source: Microsoft, Work Trend Index 2025)
- 97% of workers believe automation would benefit their organization; 66% specifically cite cutting out human error. (Source: Smartsheet)
Sales reps sell 30% of the time. The other 70% is a human doing what a machine should. That is not a productivity problem. That is the market for automation, stated in one number.
On why the demand for automation is not a forecast, it is arithmetic.
04The ROI, when it actually works
Operationalized AI pays. The catch lives in that word: operationalized.
None of the reliability trouble in the next section means the returns are fake. They are not. When a company gets AI past the pilot and into daily operation, the payback is large and fast: an average of $3.70 back for every dollar in, and more than three times that for the teams at the top.
Read this section as the prize for clearing the divide. Every ROI figure below comes from an organization that did the unglamorous part: put the system into production, kept it running, measured it. The returns are what winning the AI divide looks like on a balance sheet. The reason so few companies post numbers like these is the reason the next section exists.
average return for every $1 invested in generative AI. The top 5% of leaders report $10.30 per dollar.
(Source: IDC, Microsoft-sponsored, 2025)of executives report achieving ROI on generative AI within the first year; 88% of early adopters report ROI on at least one use case.
(Source: Google Cloud, The ROI of AI 2025)average productivity gain for workers using generative AI, across three real-world studies.
(Source: Nielsen Norman Group, 2023)ROI over three years for a composite Microsoft 365 Copilot customer, saving users 9 hours a month.
(Source: Forrester, Total Economic Impact, 2025)- Generative AI could add the equivalent of $2.6 to $4.4 trillion a year across 63 use cases. (Source: McKinsey, 2023)
- 52% of executives say their organizations have deployed AI agents in production, and 39% have launched more than ten. (Source: Google Cloud, 2025)
- 78% of organizations expect to increase overall AI spending in the next fiscal year. (Source: Deloitte, State of Generative AI, 2025)
05Reliability, trust, and the governance gap
This is where 2026 turns. Adoption is outrunning reliability, and the market has started to notice.
If the first four sections explained the AI divide, this one explains what put companies on the wrong side of it. Almost every failure has the same cause. A team stood up an agent, handed it real autonomy, and skipped the boring engineering discipline: testing, guardrails, a human in the loop. The agent worked in the demo. Then it met production.
The evidence is blunt. Trust in autonomous agents is falling, not rising. More than half of organizations have watched an agent exceed the permissions it was given. Gartner expects four in ten agent projects to be dead by 2027. This is not a case against AI. It is a case against unsupervised, untested autonomy, and it is the single clearest line separating the companies scaling AI from the ones quietly canceling it.
| Signal | Finding | Source |
|---|---|---|
| Projects at risk | 40%+ of agentic AI projects will be canceled by the end of 2027 | Gartner, 2025 |
| Trust reversing | Confidence in fully autonomous agents fell from 43% to 27% in one year | Capgemini, 2025 |
| Agents overstepping | 53% of organizations have had AI agents exceed intended permissions | Cloud Security Alliance / Zenity, 2026 |
| Governance lagging | Two-thirds of tech executives say AI adoption is outpacing their governance | IBM, 2026 |
| Oversight valued | 90% of executives view human involvement in AI workflows as positive or cost-neutral | Capgemini, 2025 |
Trust in autonomous agents is going the wrong way
Confidence in fully autonomous AI agents, year over year.
- Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027, blaming escalating costs, unclear value, and inadequate risk controls. (Source: Gartner, 2025)
- Confidence in fully autonomous AI agents dropped from 43% to 27% in a single year; only 40% of organizations trust agents to manage tasks on their own. (Source: Capgemini Research Institute, 2025)
- Only 46% of people globally are willing to trust AI systems, even though 66% already use AI regularly. (Source: KPMG and University of Melbourne, 2025)
- 74% of respondents name inaccuracy and 72% name cybersecurity as highly relevant AI risks. (Source: McKinsey, State of AI Trust 2026)
- 53% of organizations have had AI agents exceed intended permissions, and 47% had a security incident involving an agent in the past year. (Source: Cloud Security Alliance / Zenity, 2026)
- Only 21% of organizations have a mature governance model for agentic AI, meaning roughly 80% do not. (Source: Deloitte, State of AI in the Enterprise 2026)
- Enterprises faced an average of 54 AI-agent incidents each in the past year, 37% of which led to data exposure or a security breach. (Source: IBM Institute for Business Value, 2026)
- Nearly two in five (about 40%) executives believe the risks of implementing AI agents currently outweigh the benefits. (Source: Capgemini, 2025)
Would you run production code that never passed QA? With most AI-agent builders, you already are. That is the whole reliability gap, in one uncomfortable question.
On why 40% of agent projects are heading for cancellation.
06Market growth
The spend behind all of this is compounding fast, and agents are the fastest-growing slice.
For all the trouble in section five, nobody is pulling money out. The opposite: every segment of the automation market is growing, and the money is flowing hardest toward the newest and least proven layer, AI agents, growing at a 46% annual clip off a smaller base.
That is the tension this whole page circles. Capital is racing into agents faster than reliability practice can keep up. The market is betting big on the exact category most likely to get canceled. Which means the winners will not be whoever spends the most. They will be whoever turns that spend into systems that actually keep running.
Automation market segments by 2025 value
In US dollars. Agents are the smallest slice today and the fastest-growing.
- The global AI agents market was worth $7.84B in 2025 and is projected to reach $52.62B by 2030, a 46.3% annual growth rate. (Source: MarketsandMarkets, 2025)
- The global workflow automation market was worth $23.77B in 2025, growing to $40.77B by 2031. (Source: Mordor Intelligence, 2025)
- The robotic process automation market sits at $28.31B in 2025, projected to reach roughly $247B by 2035 (24.2% annual growth). (Source: Precedence Research, 2025)
- Gartner forecast low-code development technologies to reach $44.5B by 2026, up 19% over four years. (Source: Gartner, via InfoWorld)
- AI agents accounted for 17% of total AI value in 2025 and are projected to reach 29% by 2028. (Source: BCG, 2025)
What the numbers add up to
Put the six sections next to each other and a single argument falls out. Demand for automation is real and enormous: 17.3 hours a week, 30% selling time, a market compounding past $37 billion. Adoption is basically finished: 88% are in. And yet 7% have scaled, trust in autonomous agents is falling, and Gartner is calling the funeral for 40% of agent projects.
The companies on the winning side of the AI divide are not the ones with the most AI. They are the ones who treated an agent like software instead of magic: tested it, put guardrails around it, kept a human on the review, and only then let it run. That is the unglamorous discipline that turns 7% into a real number instead of an aspiration.
The uncomfortable part, if you are on the other side, is that the gap is compounding. Every quarter the winners get more hours back and more systems running while the stuck side adds another chatbot to another manual process. The tools were the barrier, not the people. The teams that get automation they can actually use, without a developer or a flowchart in sight, are the ones who close the gap. Everyone else keeps doing the work by hand and calling it AI adoption.
Frequently asked questions
How many companies actually use AI in 2026?
88% of organizations report using AI in at least one business function, up from 78% a year earlier. But only 7% have fully scaled it across the organization. Use is near-universal; mature, at-scale deployment is rare. (Source: McKinsey, State of AI 2025)
Why do AI agent projects fail?
Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. Confidence in fully autonomous agents also fell from 43% to 27% in a single year. The common thread is autonomy deployed without testing, guardrails, or human oversight. (Sources: Gartner, 2025; Capgemini, 2025)
How much time do employees lose to manual work?
Workers spend an average of 17.3 hours a week on repetitive, easily automatable tasks, and sales reps spend only 30% of their time actually selling. Nearly 60% of workers say they could save six or more hours a week if the repetitive parts of their job were automated. (Sources: Zapier, 2025; Salesforce; Smartsheet)
What is the ROI of automation and AI?
Organizations report an average return of $3.70 for every $1 invested in generative AI, with the top 5% of leaders reporting $10.30 per dollar. 74% of executives report achieving ROI within the first year. The returns are real once the system is operationalized and running in production. (Sources: IDC, Microsoft-sponsored, 2025; Google Cloud, 2025)
Who is actually driving AI adoption inside companies?
IT is over 10x more likely to lead AI acceleration than sales, marketing, HR, and customer service teams, even though those business teams carry most of the manual work. Meanwhile citizen developers are projected to outnumber professional developers four to one at large enterprises by 2026. The demand sits with business teams; the tools were built for the technical department. (Sources: Zapier AI Resistance Survey, 2025; Gartner via Kissflow)
Sources
Every statistic on this page traces to one of the reports below. Attributions are given in plain text throughout; consult each primary report for full methodology.
McKinsey, The State of AI 2025.
McKinsey, The economic potential of generative AI, 2023.
McKinsey, State of AI Trust in 2026.
Gartner, low-code forecast (via InfoWorld), 2022.
Gartner, agentic AI project cancellations, 2025.
Gartner, citizen development (via Kissflow).
Capgemini Research Institute, Rise of Agentic AI, 2025.
Microsoft, 2025 Work Trend Index.
Deloitte, State of Generative AI / State of AI in the Enterprise, 2025-2026.
IBM Institute for Business Value, AI Control Gap Study, 2026.
Cloud Security Alliance / Zenity, AI Agent Security Study, 2026.
KPMG and University of Melbourne, Trust in AI Global Study, 2025.
IDC (Microsoft-sponsored), Business Opportunity of AI, 2025.
IDC FutureScape (via CFO Dive), IT skills shortage forecast.
Google Cloud, The ROI of AI, 2025.
Forrester, Total Economic Impact of Microsoft 365 Copilot, 2025.
Nielsen Norman Group, AI productivity synthesis, 2023.
BCG, The Widening AI Value Gap, 2025.
Bain and Company, Executive Survey: AI Moves from Pilots to Production, 2025.
Salesforce, State of Sales.
HubSpot, State of AI for Marketers.
SHRM, 2025 Talent Trends: The Role of AI in HR.
World Economic Forum, Future of Jobs 2025.
Smartsheet automation survey.
Zapier, most-dreaded-tasks and AI Resistance surveys, 2025.
Asana, Anatomy of Work.
Automation Anywhere / OnePoll, most-hated office tasks.
Treasure Data (via Coupler.io), marketing data workload.
MarketsandMarkets; Mordor Intelligence; Precedence Research; Fortune Business Insights (market sizing).