Articles · bdautomated
AI Agent Statistics 2026: Every Number Checked at Its Source
Published:
Every number,
checked.
The AI agent statistics everyone quotes, traced to the original reports.
The AI agent numbers you keep seeing are mostly real. The trouble is what people take them to mean. In McKinsey's global survey, 62 percent of organizations are trying AI agents, 23 percent have scaled one somewhere, and in any single department it is no more than 10 percent. Across all U.S. businesses of every size, the Census Bureau counts one in five using AI at all. The famous "95 percent get zero return" is a six-month profit test on 153 survey answers. Every number on this page was checked against the report it came from.
You have probably seen two headlines in the same week: almost nobody has AI agents running, and almost everybody does. Both were quoting real surveys. This page exists to show you what each survey counted, so you can tell which number applies to a company like yours.
Five things the numbers actually say
- Trying agents is common, running them is rare. 62 percent of organizations are at least experimenting with AI agents, 23 percent are scaling one somewhere, and in any single business function no more than 10 percent are (McKinsey, 2025).
- The “adoption” headlines count intent. Surveys of executives report 79 percent (PwC, April 2025) and 61 percent (IBM, spring 2025) adopting agents. A survey that checked what respondents meant by “agent” found 14 percent had implemented one (Capgemini, July 2025).
- Money on the books is still rare. 39 percent of organizations report any earnings impact from AI, and about 6 percent are “high performers” with 5 percent or more of earnings from AI (McKinsey, 2025). Only 7 percent of senior leaders worldwide report an established return (KPMG, June 2026).
- Trust is falling as use rises. Trust in fully autonomous agents dropped from 43 percent to 27 percent of large organizations in a year (Capgemini), and only 21 percent of companies planning agents have a mature way to govern them (Deloitte, January 2026).
- Expected job cuts run far ahead of recorded ones. 32 percent of organizations expect AI to cut their workforce next year (McKinsey); AI-related employment decreases occurred in 2 percent of U.S. firms (U.S. Census Bureau, April 2026).
How many companies use AI agents?
Nine surveys give nine answers, from 10 percent to 79 percent. None of them is wrong. They asked different people different questions.
Three things move the number.
Who was asked. PwC asked 308 U.S. executives and 79 percent said agents were already being adopted in their companies. IBM asked 2,000 CEOs and 61 percent said they were actively adopting agents and preparing to implement them at scale. LangChain asked over 1,300 people who build agents for a living and 57 percent had one in production. Executives are describing what they intend to do. Builders are describing their own work. Neither group is a random sample of businesses.
What counts as “using”. McKinsey separates experimenting (39 percent) from scaling (23 percent), where scaling means expanding the deployment inside at least one business function. Most of those scaling do so in one or two functions only, which is why the per-function figure is no more than 10 percent. Capgemini went further and re-checked 900 of its 1,500 respondents to make sure they meant agents rather than AI assistants: 14 percent had implemented agents, 2 percent at full scale, 23 percent had pilots, and 61 percent were preparing or exploring.
What the thing being counted is. Menlo Ventures asked 495 U.S. enterprise AI decision-makers to describe what they had built. Only 16 percent of enterprise deployments qualified as true agents, meaning a system that plans, acts, observes the result and adapts. The rest were fixed sequences of steps wrapped around a model call. Gartner puts the same point from the vendor side: it estimates only about 130 of the thousands of companies selling “agentic AI” are real, and calls the rest agent washing (Gartner, June 2025).
Two surveys sit in the middle and are worth knowing. Google Cloud found 52 percent of executives actively using agents, and 39 percent with more than ten launched, but it only surveyed enterprises that already had generative AI in production. KPMG found 53 percent of large U.S. companies deploying agents in spring 2026, level with the quarter before, while the share running several agents together across a workflow doubled from 9 percent to 18 percent.
Where agents are, department by department
Stanford’s 2026 AI Index redrew McKinsey’s function-level data. In IT, the department with the most agent activity, 69 percent of organizations reported no agent use at all. In knowledge management it was 66 percent. Scaled use sat in the single digits for nearly every function. The one place scaled use is common is the technology sector, where it reaches 24 percent in software engineering, 22 percent in IT and 21 percent in service operations.
Size matters as much as sector. Among people who build agents, 67 percent of those at organizations with more than 10,000 employees had agents in production, against 50 percent at organizations under 100 people (LangChain).
The number for businesses of every size
Every survey above leans toward large companies. The U.S. Census Bureau asks a nationally representative sample of businesses of every size whether they used AI in any business function in the past two weeks. As of 3 May 2026 the answer was 19.8 percent. Between December 2025 and May 2026 it hovered between 17 and 20 percent. Firms with at least 250 employees were at 37 percent; firms with four or fewer employees, under 20 percent. That is AI of any kind. The Census does not yet publish an agent-specific figure, so for a typical business, agents running in a given department is a single-digit story.
What are the agents doing?
In large organizations, agent use is most common in IT and knowledge management, for service-desk work and deep research (McKinsey). Among people building agents, customer service is the most common primary use case at 26.5 percent, then research and data analysis at 24.4 percent and internal workflow automation at 18 percent (LangChain). Among enterprises already running agents, the most common uses are customer service and experience (49 percent), marketing (46 percent), security operations (46 percent) and tech support (45 percent) (Google Cloud).
For ordinary businesses the picture is simpler. Among U.S. firms that use AI, the most common functions are sales and marketing (52 percent), strategy and business development (45 percent) and IT (41 percent), and 57 percent use AI in three or fewer functions (U.S. Census Bureau). Two-thirds of those firms, 66 percent, use AI only to help people with tasks.
How people hand work to AI is measurable from usage data rather than surveys. On Claude.ai, 52 percent of conversations are collaborative, where the person learns, iterates or asks for feedback, and 45 percent are handed over entirely (Anthropic, January 2026). Business use through the API is different: 77 percent of it follows automation patterns, mostly full delegation of a task (Anthropic, September 2025). The share of consumer conversations where a whole task was handed over rose from 27 percent to 39 percent during 2025, then eased back. Across occupations, about 49 percent have seen at least a quarter of their tasks performed with Claude (Anthropic, March 2026); that is a count of tasks people have used Claude for, and it says nothing about jobs.
Microsoft found that 49 percent of Microsoft 365 Copilot conversations support cognitive work such as analyzing, solving problems and evaluating, and that 86 percent of AI users treat the output as a starting point rather than an answer. Agents in the Microsoft 365 ecosystem grew 15 times year over year, a multiple from a small base that Microsoft does not publish.
Is it paying off?
This is where the numbers disagree most.
Start with the strict measures. McKinsey asks about EBIT, which is profit before interest and tax. 39 percent of organizations attribute any earnings impact to AI, and most of those say it is under 5 percent of earnings. About 6 percent are what McKinsey calls AI high performers: at least 5 percent of earnings from AI, and value they describe as significant. KPMG asked 2,145 senior leaders in 20 countries in spring 2026 and found 7 percent with an established return, while 24 percent were under pressure to prove value to investors. CEOs told IBM that only 25 percent of their AI initiatives had delivered the expected return and only 16 percent had scaled company-wide.
Now the loose ones. Google Cloud reports that 74 percent of executives achieved a return within the first year. Two things sit inside that sentence: only companies already running generative AI were surveyed, and “return” means a return on at least one use case. PwC reports that among companies adopting agents, 66 percent see higher productivity, 57 percent cost savings, 55 percent faster decisions and 54 percent better customer experience. OpenAI surveyed 9,000 workers at almost 100 of its own enterprise customers: 75 percent said AI improved the speed or quality of their output, and they reported saving 40 to 60 minutes a day. Gallup found 65 percent of employees at organizations that implemented AI say it improved their productivity, and only 14 percent strongly agree it has transformed how work gets done.
The best-measured productivity numbers come from controlled studies, each on one narrow task: gains of 14 to 15 percent in customer support, 26 percent in software development and 50 percent in marketing output, with smaller gains in work that needs deeper reasoning (Stanford AI Index 2026).
The 95 percent, decoded
MIT Project NANDA reported in July 2025 that 95 percent of organizations are getting zero return from generative AI. Here is what that measured. The research ran from January to June 2025. It combined 52 structured interviews, 153 survey responses collected at four industry conferences, and a review of 300 publicly disclosed AI initiatives. “Zero return” means no measurable profit-and-loss impact, observed within roughly six months of a pilot. The authors label the findings preliminary. The same report says over 80 percent of organizations had piloted general tools like ChatGPT and nearly 40 percent had deployed them, while custom enterprise tools went from 60 percent evaluated to 20 percent piloted to 5 percent in production.
So the honest sentence is: in a small 2025 sample, 95 percent of organizations could not show a profit-and-loss effect within six months. That is consistent with McKinsey’s 39 percent reporting any earnings impact over a longer window. It does not say that 95 percent of AI projects fail.
Two more figures on the road from pilot to production. Only 25 percent of organizations have moved 40 percent or more of their AI pilots into production, and 34 percent say they are using AI to deeply transform their business while 37 percent use it at a surface level (Deloitte). And 76 percent of enterprise AI use cases are now bought rather than built in-house, up from 53 percent a year earlier; AI purchases reach production 47 percent of the time, against 25 percent for ordinary software (Menlo Ventures).
What goes wrong
The prediction everyone quotes as a fact. Gartner predicts that over 40 percent of agentic AI projects will be canceled by the end of 2027, because of rising costs, unclear business value or inadequate risk controls. It is a forecast from June 2025. Nobody has counted the cancellations yet. Gartner has a second 40 percent, and that one is about software products: it predicts 40 percent of enterprise applications will include task-specific agents by the end of 2026, up from less than 5 percent (Gartner, August 2025). The two get mixed up constantly.
Trust fell in a year. Trust in fully autonomous agents fell from 43 percent to 27 percent of large organizations in one year; Capgemini reads that as business reality replacing early enthusiasm. Only 15 percent of business processes are expected to run semi- or fully autonomously within 12 months. Fewer than one in five organizations report high data readiness, and over 80 percent say they lack mature AI infrastructure (Capgemini).
Guardrails lag plans. Close to three-quarters of companies plan to deploy agentic AI within two years, and only 21 percent of them have a mature model for governing agents (Deloitte). U.S. executives trust agents most with data analysis (38 percent) and least with financial transactions (20 percent) and autonomous employee interactions (22 percent); they rank cybersecurity and cost as the top challenges at 34 percent each (PwC). Among people building agents, quality is the top barrier to production at 32 percent; 89 percent have monitoring in place but only 52 percent run evaluations (LangChain).
Cost is the new constraint. One-third of senior leaders cite limited understanding of usage costs as a key challenge in deploying agents, and leaders with strong cost visibility are five times as likely to report a return, 15 percent against 3 percent (KPMG, global). The most chosen way to manage agent risk over the next year, picked by 52 percent of large U.S. companies for the second quarter running, is a human-in-the-loop model: a person checks the outputs without watching every action (KPMG, U.S.).
Decisions are being handed over anyway. CEOs say 25 percent of operational decisions where the rules can be written down are already made by AI without human intervention, and they expect 48 percent by 2030. 64 percent are comfortable making major strategic decisions based on AI-generated input, and 76 percent of their organizations now have a Chief AI Officer, up from 26 percent a year earlier (IBM, May 2026).
Small businesses: why 58 percent and 20 percent are both right
The U.S. Chamber of Commerce surveyed 3,870 small businesses in June 2025 and found 58 percent saying they use generative AI, up from 40 percent in 2024 and 23 percent in 2023. The Census Bureau found one in five businesses using AI in a business function. The gap is the definition. The Chamber counts a business that uses a chatbot to draft emails. The Census counts AI in the way the business produces or sells. For a small business both are useful: most owners have tried the tools, and most have not yet put them into a business function.
The size gap runs through everything. Firms with 250 or more employees are at 37 percent in the Census data; firms with four or fewer are under 20 percent, and use among firms with fewer than 20 employees did not change significantly between December 2025 and May 2026. Among small businesses using AI, 82 percent increased their workforce over the past year (U.S. Chamber), which says AI users were growing, and does not say AI caused the growth.
AI and jobs: expected versus recorded
Expectations first. 32 percent of organizations expect AI to reduce their total workforce by 3 percent or more in the coming year, 43 percent expect little or no change and 13 percent expect growth. Across business functions, a median of 30 percent expect a decrease next year while a median of 17 percent reported one over the past year (McKinsey). Then the recorded numbers. AI-related employment decreases occurred in only 2 percent of U.S. firms between November 2025 and January 2026 (U.S. Census Bureau). The one group where a clear effect shows so far is young software developers: employment for developers aged 22 to 25 has fallen nearly 20 percent from 2024 (Stanford AI Index 2026).
The people side is where the surveys agree. CEOs say only 25 percent of their workforce uses AI regularly, although 86 percent believe employees have the skills, and 83 percent say AI success depends more on people’s adoption than on technology (IBM). They expect 29 percent of employees to need reskilling for a different role and 53 percent to need upskilling for their current one by 2028. Microsoft finds that organizational factors such as culture and manager support explain 67 percent of reported AI impact and individual factors 32 percent, and that only 26 percent of AI users say their leadership is clearly aligned on AI. Gallup finds employees whose manager actively supports AI are 1.7 times as likely to use it weekly, and only 36 percent strongly agree their manager does. 61 percent of large organizations report rising employee anxiety about agents, and over half believe agents will displace more jobs than they create (Capgemini).
What this means if you run a company
Use the department number. The figure that describes your situation is the one for a single function: no more than 10 percent of organizations have agents scaled in any one department, and in most departments two-thirds have none. Nobody is behind. The companies that are ahead have agents running in one or two functions.
Measure at the company level and expect the loose numbers to flatter. A survey of people who already use AI will say it helps, and they are probably right. The questions that decide whether it was worth the money are the strict ones: earnings, established return, pilots that reached production. On those, the share saying yes is between 6 and 39 percent in the surveys above.
Copy what the 6 percent do. McKinsey’s high performers are nearly three times as likely to have fundamentally redesigned a workflow, and more likely to have defined when a model’s output needs a person to check it (McKinsey). Deloitte’s successful companies start with lower-risk use cases and build governance as they go. KPMG’s leaders with cost visibility get returns five times as often. None of that is about which model you bought.
Keep a person at the point where a mistake would cost you. 52 percent of large U.S. companies chose a human-in-the-loop model as their main safeguard, and trust in agents rises with experience: 47 percent of organizations implementing agents report above-average trust, against 37 percent still exploring (Capgemini). The pattern is to let the agent draft, research and check, and let a person approve anything that leaves the company, then widen what the agent may do as it proves reliable. That is how we install agents at bdautomated: each one arrives with a job, a trigger and a level of trust the customer sets, from draft-and-wait upward.
Buy before you build, and start in the functions that already use AI. 76 percent of enterprise AI use cases are now bought, and ready-made tools reach production faster (Menlo Ventures). In ordinary firms the functions already using AI are sales and marketing, strategy and IT (U.S. Census Bureau), which is where an agent has data to work with.
The 2026 McKinsey edition: not yet verified by us
McKinsey published a 2026 edition of its survey on 25 August 2026. Its page returned an error to our reader and to our fetching tools on 15 September 2026, so none of its figures appear in the table below; every McKinsey row on this page is from the November 2025 edition, which we opened. For the record, The Register, a secondary source, reports that the 2026 survey had 1,719 respondents, that 37 percent attribute at least some EBIT impact to AI, that 6 percent are high performers, that 40 percent of respondents at organizations above one billion in revenue are scaling agents (27 percent a year earlier), that 20 percent say operating costs constrained their AI use, and that 39 percent expect AI-related job cuts against 32 percent the year before. We will add those rows once the original opens. If you can open it, tell us at contact@bdautomated.com.
How we checked every number
Every figure on this page passed four checks before it went in.
- The number is in the original document. Another roundup or a press summary does not count. If we could not open the original, the number was left out.
- The exact place is recorded: the section, exhibit, figure or page, with a verbatim quote. Both are in the dataset.
- What it measures is written down in plain words: who was asked, how many, when, and what the question counted.
- It is set against the other sources. Where they disagree, the disagreement is shown, never averaged away.
What we left out, and why. Market-size forecasts: the reports behind most of them are paid, so rung one fails, and the published totals disagree with each other by more than the size of the market. The often-quoted S&P Global finding that 42 percent of companies abandoned most of their AI initiatives in 2025: S&P’s own page would not open for us, so it is not here. Any figure we could only find inside another statistics roundup. McKinsey’s 2026 edition, for the reason above.
How to read the table. “Who was asked” tells you whether a number is about large companies, all businesses, executives, workers or agent builders. Most disagreements between statistics disappear once you read that column.
Corrections. If a number here does not match its source, write to contact@bdautomated.com with the row number. Corrections are made in place and listed in the changelog at the end.
Updates. This page is refreshed in place when a source publishes a new edition. The address never changes.
The sources: all 75 numbers
The full dataset, with a verbatim quote for every row, is free to download as CSV or JSON under a CC BY 4.0 licence. Cite bdautomated and link to this page.
| # | The number | Source | Date | Who was asked | Where in the document | What it says |
|---|---|---|---|---|---|---|
| 1 | 88% of organizations regularly use AI in at least one business function. | McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation | 2025-11-05 | 1,993 respondents in 105 nations, fielded 25 June to 29 July 2025, weighted by each nation's share of global GDP | section 'AI use continues to broaden but remains primarily in pilot phases', Exhibit 1 | “88 percent report regular AI use in at least one business function, compared with 78 percent a year ago” |
| 2 | 79% of organizations regularly use generative AI in at least one business function. | Stanford Institute for Human-Centered AI, AI Index Report 2026, chapter 4 (Economy) | 2026-04 | reproduces McKinsey's 2025 survey of 1,993 respondents; self-reported, directional | Figure 4.3.1, 'Share of respondents who say their organization uses AI in at least one function, 2017-25' | “79% of respondents reporting that their organizations regularly use generative AI in at least one business function, compared to 71% in 2024” |
| 3 | 19.8% of all U.S. businesses used AI in any business function in the two weeks before 3 May 2026. | U.S. Census Bureau, Large Firms With at Least 20 Employees Biggest AI Users (BTOS) | 2026-05-26 | Business Trends and Outlook Survey, biweekly, nationally representative of U.S. businesses; data 14 Dec 2025 to 3 May 2026 | section 'AI Trends Across Sectors' | “the AI use rates in the Information (39.7%) and Finance and Insurance (33.9%) sectors were both higher than the national rate (19.8%)” |
| 4 | 37% of U.S. firms with at least 250 employees reported using AI; under 20% of firms with four or fewer employees did. | U.S. Census Bureau, Large Firms With at Least 20 Employees Biggest AI Users (BTOS) | 2026-05-26 | Business Trends and Outlook Survey, biweekly, nationally representative of U.S. businesses; data 14 Dec 2025 to 3 May 2026 | section 'U.S. Business AI Use', Figure 2 | “37% of firms with at least 250 employees reported using AI in their business operations” |
| 5 | 18% of U.S. firms used AI in a business function between November 2025 and January 2026, or 32% when weighted by how many people they employ. | U.S. Census Bureau, The Microstructure of AI Diffusion (working paper CES-WP-26-25) | 2026-04-15 | BTOS AI supplement, nationally representative, reference period Nov 2025 to Jan 2026 (working paper CES-WP-26-25) | abstract | “18% of firms used AI in a business function, rising to 32% on an employment-weighted basis; adoption is expected to reach 22% within six months” |
| 6 | 58% of U.S. small businesses say they use generative AI, up from 40% in 2024 and 23% in 2023. | U.S. Chamber of Commerce (with Teneo Research), Empowering Small Business: The Impact of Technology on U.S. Small Business, fourth edition | 2025-08-18 | 3,870 U.S. small businesses with fewer than 250 employees, surveyed online 6 to 26 June 2025 | Introduction, key findings list | “58% of small businesses self-identified that they use generative AI—up from 40% in 2024 and more than double compared to 2023 (23%)” |
| 7 | 52% of U.S. employees use AI at work at least a few times a year; 30% a few times a week or more; 15% daily. | Gallup, Global Indicator: Artificial Intelligence (workplace AI adoption tracker) | 2026-05 | Gallup Panel, probability-based random sample of U.S. adults employed full or part time; figures as of May 2026 | 'What Percentage of Employees Use AI at Work?' | “As of May 2026, 15% of U.S. employees use AI daily in their role, 30% use it a few times a week or more, and 52% use it a few times a year or more” |
| 8 | 47% of U.S. employees say their organization has integrated AI tools; only 25% say it has communicated a clear plan for doing so. | Gallup, Global Indicator: Artificial Intelligence (workplace AI adoption tracker) | 2026-05 | Gallup Panel, probability-based random sample of U.S. adults employed full or part time; figures as of May 2026 | 'What Percentage of Companies Have Integrated AI?' and 'Do Most Companies Have a Clear AI Strategy?' | “As of May 2026, 47% of U.S. employees say their organization has integrated AI technology or tools to improve organizational practices” |
| 9 | 62% of organizations are at least experimenting with AI agents. | McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation | 2025-11-05 | 1,993 respondents in 105 nations, fielded 25 June to 29 July 2025, weighted by each nation's share of global GDP | Key findings, item 2, and Exhibit 2 | “Sixty-two percent of survey respondents say their organizations are at least experimenting with AI agents” |
| 10 | 23% of organizations are scaling an AI agent system somewhere in the company; another 39% are experimenting. | McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation | 2025-11-05 | 1,993 respondents in 105 nations, fielded 25 June to 29 July 2025, weighted by each nation's share of global GDP | section 'Many organizations are already experimenting with AI agents', Exhibit 2 | “Twenty-three percent of respondents report their organizations are scaling an agentic AI system somewhere in their enterprises (that is, expanding the deployment and adoption of the technology within a least one business function), and an additional 39 percent say they have begun experimenting with AI agents” |
| 11 | In any single business function, no more than 10% of organizations are scaling AI agents. | McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation | 2025-11-05 | 1,993 respondents in 105 nations, fielded 25 June to 29 July 2025, weighted by each nation's share of global GDP | same section, Exhibit 2 | “In any given business function, no more than 10 percent of respondents say their organizations are scaling AI agents” |
| 12 | In IT, the function with the most agent activity, 69% of organizations report no agent use at all; in knowledge management, 66%. | Stanford Institute for Human-Centered AI, AI Index Report 2026, chapter 4 (Economy) | 2026-04 | reproduces McKinsey's 2025 survey of 1,993 respondents; self-reported, directional | Figure 4.3.7, 'Stage of AI agent use by business function, 2025' | “Even in functions with the most activity, including IT and knowledge management, about two-thirds or more of respondents reported no use” |
| 13 | In the technology sector, scaled agent use reaches 24% in software engineering, 22% in IT and 21% in service operations. | Stanford Institute for Human-Centered AI, AI Index Report 2026, chapter 4 (Economy) | 2026-04 | reproduces McKinsey's 2025 survey of 1,993 respondents; self-reported, directional | Figure 4.3.8 | “the technology sector had comparatively higher rates of scaled agent use in software engineering (24%), IT (22%), and service operations (21%)” |
| 14 | 79% of U.S. executives say AI agents are already being adopted in their companies. | PwC, PwC's AI Agent Survey | 2025-05 | 308 U.S. executives (C-suite 33%, VP 13%, director 54%), surveyed 22 to 28 April 2025 | 'Key trends about AI agents in the enterprise' | “Seventy-nine percent say AI agents are already being adopted in their companies” |
| 15 | Of companies adopting agents, 35% say they do so broadly and 17% say agents are fully adopted in almost all workflows; 18% of companies are not using agents at all. | PwC, PwC's AI Agent Survey | 2025-05 | 308 U.S. executives (C-suite 33%, VP 13%, director 54%), surveyed 22 to 28 April 2025 | section '1. Budgets are surging' (the 35% and 17%) and section '4. Only a few businesses have their eyes on the prize' (the 18%) | “Of those companies adopting AI agents, 35% say they're doing so broadly, while another 17% say AI agents are being fully adopted in almost all workflows and functions” |
| 16 | 14% of large organizations have implemented AI agents (12% at partial scale, 2% at full scale); 23% have pilots; 61% are preparing or exploring. | Capgemini Research Institute, Rise of agentic AI: How trust is the key to human-AI collaboration | 2025-07-16 | 1,500 leaders at director level and above, 14 countries, large organizations; 900 re-confirmed their adoption answer | Executive summary | “Already, 14% of organizations have implemented AI agents at partial (12%) or full scale (2%) and nearly one-quarter (23%) have launched pilots, while another 61% are preparing for or exploring deployment” |
| 17 | 52% of executives whose companies already run generative AI in production say their organization is actively using AI agents; 39% say it has launched more than ten. | Google Cloud (with National Research Group), The ROI of AI 2025 (press release) | 2025-09-04 | 3,466 senior leaders of enterprises across 24 countries, all with generative AI already deployed | 'The Rise of AI Agents', key findings | “More than half (52%) of executives report their organization is actively using AI agents, with 39% reporting their company has launched more than ten” |
| 18 | 53% of large U.S. companies are deploying AI agents (55% the quarter before); the share orchestrating several agents across a workflow doubled from 9% to 18%. | KPMG, AI Quarterly Pulse Survey, Q2 2026 (U.S.) | 2026-06 | 204 U.S. C-suite and business leaders at organizations with annual revenue of 1 billion dollars or more, 28 April to 25 May 2026 | 'AI agent deployment and trends' | “AI agent deployment holds steady at 53% Compared to last quarter 55%” |
| 19 | 61% of CEOs say they are actively adopting AI agents and preparing to implement them at scale. | IBM Institute for Business Value, 2025 CEO Study (press release) | 2025-05-06 | 2,000 CEOs from 33 countries and 24 industries, February to April 2025, with Oxford Economics | opening paragraphs | “61% confirm they are actively adopting AI agents today and preparing to implement them at scale” |
| 20 | 57% of professionals who build AI agents have agents in production; 30% are developing them with plans to deploy. | LangChain, State of Agent Engineering | 2026-06-12 | over 1,300 professionals building AI agents: engineers, product managers, business leaders and executives | 'Large enterprises are leading adoption' | “More than half of respondents surveyed (57.3%) now have agents running in production environments, with another 30.4% actively developing agents with concrete plans to deploy them” |
| 21 | Among organizations with more than 10,000 employees, 67% of agent builders have agents in production, against 50% at organizations under 100 employees. | LangChain, State of Agent Engineering | 2026-06-12 | over 1,300 professionals building AI agents: engineers, product managers, business leaders and executives | 'What changes at scale?' | “Of 10k+ size orgs, 67% had agents in production, with 24% actively developing with plans for production— versus for <100 size orgs, 50% had agents in production” |
| 22 | Only 16% of enterprise AI deployments (27% at startups) qualify as true agents that plan, act, observe and adapt; the rest are fixed-sequence workflows around a model call. | Menlo Ventures, 2025: The State of Generative AI in the Enterprise | 2025-12-09 | 495 U.S. enterprise AI decision-makers at companies actively using AI tools, 7 to 25 November 2025 | 'AI Infrastructure: A Modern AI Stack Still in Development' | “Only 16% of enterprise and 27% of startup deployments qualify as true agents—systems where an LLM plans and executes actions, observes feedback, and adapts its behavior—while most are still built around fixed-sequence or routing-based workflows wrapped around a single model call” |
| 23 | Close to three-quarters of companies plan to deploy agentic AI within two years, but only 21% have a mature model for governing agents. | Deloitte AI Institute, The State of AI in the Enterprise 2026: The Untapped Edge (press release and report page) | 2026-01-21 | 3,235 business and IT leaders across 24 countries and six industries, August to September 2025 | press release, 'For Agentic AI, governance and growth go hand in hand' | “Agentic AI is poised for growth with close to three-quarters of companies planning to deploy Agentic AI within two years. Yet only 21% of those companies report having a mature model for agent governance” |
| 24 | In a January 2025 Gartner poll, 19% of organizations had made significant investments in agentic AI, 42% conservative investments, 8% none, and 31% were waiting or unsure. | Gartner, Gartner press release, 25 June 2025 | 2025-06-25 | 3,412 webinar attendees polled in January 2025 | third paragraph | “According to a January 2025 Gartner poll of 3,412 webinar attendees, 19% said their organization had made significant investments in agentic AI, 42% had made conservative investments, 8% no investments, with the remaining 31% taking a wait and see approach or are unsure” |
| 25 | 16% of the AI users Microsoft surveyed are 'Frontier Professionals': they use agents for multi-step work, redesign workflows and share AI practices. | Microsoft WorkLab, 2026 Work Trend Index Annual Report | 2026-05-05 | 20,000 knowledge workers who use AI at work, 10 countries, fielded by Edelman Data x Intelligence from 18 February 2026 to 7 April 2026 per the methodology section (two chart captions on the same page say to 20 April 2026), plus Microsoft 365 telemetry | 'AI lifts the ceiling on individual potential' | “They represent a small but disproportionately valuable group: 16% of the AI users we surveyed” |
| 26 | Agents in the Microsoft 365 ecosystem increased 15 times year over year, and 18 times in large enterprises. | Microsoft WorkLab, 2026 Work Trend Index Annual Report (key takeaways) | 2026-05-05 | Microsoft 365 telemetry | key takeaway 5 | “Agents in the Microsoft 365 ecosystem have increased 15x year over year and 18x in large enterprises” |
| 27 | Customer service is the most common primary agent use case (26.5%), then research and data analysis (24.4%) and internal workflow automation (18%). | LangChain, State of Agent Engineering | 2026-06-12 | over 1,300 professionals building AI agents: engineers, product managers, business leaders and executives | 'Leading agent use cases' | “Customer service emerged as the most common agent use case (26.5%), with research & data analysis close behind (24.4%)” |
| 28 | Among enterprises running agents, the most common uses are customer service and experience (49%), marketing (46%), security operations (46%) and tech support (45%). | Google Cloud (with National Research Group), The ROI of AI 2025 (press release) | 2025-09-04 | 3,466 senior leaders of enterprises across 24 countries, all with generative AI already deployed | 'Applications of Agentic AI' | “The most common cross-industry applications for AI agents reported in the study were customer service and experience (49%), marketing (46%), security operations and cybersecurity (46%), and tech support (45%)” |
| 29 | Agent use is most commonly reported in IT and knowledge management, for service-desk work and deep research. | McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation | 2025-11-05 | 1,993 respondents in 105 nations, fielded 25 June to 29 July 2025, weighted by each nation's share of global GDP | section 'Many organizations are already experimenting with AI agents', Exhibit 3 | “agent use is most commonly reported in IT and knowledge management, where agentic use cases such as service-desk management in IT and deep research in knowledge management have quickly developed” |
| 30 | Among U.S. firms that use AI, the most common functions are sales and marketing (52%), strategy and business development (45%) and IT (41%); 57% use AI in three or fewer functions. | U.S. Census Bureau, The Microstructure of AI Diffusion (working paper CES-WP-26-25) | 2026-04-15 | BTOS AI supplement, nationally representative, reference period Nov 2025 to Jan 2026 (working paper CES-WP-26-25) | abstract | “57% of users integrate AI in three or fewer business functions, most commonly Sales and Marketing (52%), Strategy and Business Development (45%), and IT (41%)” |
| 31 | Two-thirds of U.S. firms using AI (66%) use it only to help people with tasks, not to replace them. | U.S. Census Bureau, The Microstructure of AI Diffusion (working paper CES-WP-26-25) | 2026-04-15 | BTOS AI supplement, nationally representative, reference period Nov 2025 to Jan 2026 (working paper CES-WP-26-25) | abstract | “Most users (66%) rely on AI solely to augment tasks” |
| 32 | On Claude.ai, 52% of conversations are collaborative (the person learns, iterates or asks for feedback) and 45% are handed over entirely; through the API, business use is mostly handed over. | Anthropic, Anthropic Economic Index, January 2026 report (Economic primitives) | 2026-01-15 | privacy-preserving samples of Claude.ai conversations and first-party API traffic; usage data, not a survey | 'What has changed since our last report', item 2 | “The share of conversations classified as augmented jumped 5pp to 52% and the share deemed automated fell 4pp to 45%” |
| 33 | 77% of business use of Claude through the API follows automation patterns, mostly full delegation of a task, against about 50% for people using Claude.ai. | Anthropic, Anthropic Economic Index, September 2025 report | 2025-09-15 | privacy-preserving samples of Claude.ai conversations and first-party API traffic; usage data, not a survey | 'Systematic enterprise deployment of AI' | “1P API usage is automation dominant: 77% of business uses involve automation usage patterns, compared to about 50% for Claude.ai users” |
| 34 | The share of Claude.ai conversations where the person hands over a whole task rose from 27% to 39% in eight months of 2025. | Anthropic, Anthropic Economic Index, September 2025 report | 2025-09-15 | privacy-preserving samples of Claude.ai conversations and first-party API traffic; usage data, not a survey | 'Changing patterns of usage on Claude.ai over time' | “"Directive" conversations, where users delegate complete tasks to Claude, jumped from 27% to 39%” |
| 35 | About 49% of occupations have seen at least a quarter of their tasks performed with Claude. | Anthropic, Anthropic Economic Index, March 2026 report (Learning curves) | 2026-03-24 | privacy-preserving samples of Claude.ai conversations and first-party API traffic; usage data, not a survey | 'What has changed since our last report' | “About 49% of jobs have seen at least a quarter of their tasks performed using Claude” |
| 36 | 49% of Microsoft 365 Copilot conversations support cognitive work: analyzing, solving problems, evaluating, thinking creatively. | Microsoft WorkLab, 2026 Work Trend Index Annual Report | 2026-05-05 | 20,000 knowledge workers who use AI at work, 10 countries, fielded by Edelman Data x Intelligence from 18 February 2026 to 7 April 2026 per the methodology section (two chart captions on the same page say to 20 April 2026), plus Microsoft 365 telemetry | 'AI lifts the ceiling on individual potential' | “49% of all conversations support cognitive work—helping workers analyze information, solve problems, evaluate, and think creatively” |
| 37 | 39% of organizations attribute any earnings (EBIT) impact to AI, and most of those say it is under 5% of earnings. | McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation | 2025-11-05 | 1,993 respondents in 105 nations, fielded 25 June to 29 July 2025, weighted by each nation's share of global GDP | section 'AI as a catalyst for innovation', Exhibit 6 | “Thirty-nine percent of respondents attribute any level of EBIT impact to AI, and most of those respondents say that less than 5 percent of their organization's EBIT is attributable to AI use” |
| 38 | About 6% of organizations are 'AI high performers': they attribute at least 5% of earnings to AI and call the value significant. | McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation | 2025-11-05 | 1,993 respondents in 105 nations, fielded 25 June to 29 July 2025, weighted by each nation's share of global GDP | section 'Organizations with ambitious AI agendas are seeing the most benefit' | “our definition of AI high performers, representing about 6 percent of respondents” |
| 39 | CEOs say only 25% of their AI initiatives delivered the expected return over the last few years, and only 16% scaled company-wide. | IBM Institute for Business Value, 2025 CEO Study (press release) | 2025-05-06 | 2,000 CEOs from 33 countries and 24 industries, February to April 2025, with Oxford Economics | 'CEOs face competing pressures of short-term ROI and long-term innovation' | “Surveyed CEOs report that only 25% of AI initiatives have delivered expected ROI over the last few years, and only 16% have scaled enterprise wide” |
| 40 | Only 7% of senior leaders worldwide report an established return on their AI investment; 24% face pressure to prove value to investors. | KPMG, Global AI Pulse: Q2 2026 (press release) | 2026-06-24 | 2,145 C-suite and senior business leaders across 20 countries, 28 April to 25 May 2026, at organizations with annual revenue above 50 million dollars (the press release's About section sets a 100 million floor for its ten largest markets) | 'AI momentum builds as leaders prioritize people to unlock value' | “turning that momentum into measurable returns remains concentrated to only 7 percent of leaders who report establishing ROI, even as nearly one in four (24 percent) face pressure to prove value to investors” |
| 41 | The share of organizations at the 'driving adoption' stage rose from 13% to 22% in one quarter of 2026, the biggest move on KPMG's maturity curve. | KPMG, Global AI Pulse: Q2 2026 (press release) | 2026-06-24 | 2,145 C-suite and senior business leaders across 20 countries, 28 April to 25 May 2026, at organizations with annual revenue above 50 million dollars (the press release's About section sets a 100 million floor for its ten largest markets) | highlights | “More organizations have reached the stage where AI is part of everyday work — rising to 22 percent, up from 13 percent in Q1” |
| 42 | 74% of executives at companies running generative AI report a return within the first year; among 'agentic early adopters' 88% see a return on at least one use case. | Google Cloud (with National Research Group), The ROI of AI 2025 (press release) | 2025-09-04 | 3,466 senior leaders of enterprises across 24 countries, all with generative AI already deployed | 'ROI and Investment Remains Strong' | “74% of executives report achieving ROI within the first year” |
| 43 | Among companies adopting agents, 66% report higher productivity, 57% cost savings, 55% faster decisions and 54% better customer experience. | PwC, PwC's AI Agent Survey | 2025-05 | 308 U.S. executives (C-suite 33%, VP 13%, director 54%), surveyed 22 to 28 April 2025 | '2. Measurable value is here' | “Of those adopting AI agents, nearly two-thirds (66%) report increased productivity. Over half (57%) report cost savings, faster decision-making (55%) or improved customer experience (54%)” |
| 44 | 65% of employees at organizations that implemented AI say it has improved their productivity; only 14% strongly agree it has transformed how work gets done. | Gallup, Global Indicator: Artificial Intelligence (workplace AI adoption tracker) | 2026-05 | Gallup Panel, probability-based random sample of U.S. adults employed full or part time; figures as of May 2026 | 'Does AI Improve Productivity at Work?' and 'Is AI Really Changing the Way People Work?' | “As of May 2026, 65% of employees working in organizations that have implemented AI say it has had a positive effect on their productivity and efficiency at work” |
| 45 | 75% of workers at OpenAI's enterprise customers say AI improved the speed or quality of their output; they report saving 40 to 60 minutes a day. | OpenAI, The state of enterprise AI (2025 report) | 2025-12-08 | usage data from OpenAI's business customers plus a survey of 9,000 workers across almost 100 enterprises, all paying customers | 'Workers report measurable value from using AI' | “Across surveyed enterprises, 75% of workers report that using AI at work has improved either the speed or quality of their output. Workers report saving 40-60 minutes per day” |
| 46 | Controlled studies find productivity gains of 14% to 15% in customer support, 26% in software development and 50% in marketing output, and smaller gains in work that needs deeper reasoning. | Stanford Institute for Human-Centered AI, AI Index Report 2026, chapter 4 (Economy) | 2026-04 | studies summarized in the chapter highlights (Brynjolfsson et al., Cui et al., Ju and Aral) | Chapter highlights, item 9 | “Studies report gains of 14% to 15% in customer support, 26% in software development, and 50% in marketing output. Gains are smaller in tasks requiring deeper reasoning” |
| 47 | 95% of organizations are getting zero return from generative AI, measured as no measurable profit-and-loss impact within about six months of a pilot. | MIT Project NANDA, The GenAI Divide: State of AI in Business 2025 | 2025-07 | 52 organizations interviewed, 153 senior leaders surveyed at four conferences, 300 public AI initiatives reviewed, January to June 2025; preliminary findings | Executive summary, page 3 | “this report uncovers a surprising result in that 95% of organizations are getting zero return” |
| 48 | 60% of organizations evaluated enterprise-grade AI tools, 20% reached a pilot and 5% reached production; general tools like ChatGPT were piloted by over 80% and deployed by nearly 40%. | MIT Project NANDA, The GenAI Divide: State of AI in Business 2025 | 2025-07 | 52 organizations interviewed, 153 senior leaders surveyed at four conferences, 300 public AI initiatives reviewed, January to June 2025; preliminary findings | Executive summary, page 3 | “Sixty percent of organizations evaluated such tools, but only 20 percent reached pilot stage and just 5 percent reached production” |
| 49 | 66% of organizations report productivity and efficiency gains from AI; 20% report increased revenue, while 74% hope to. | Deloitte AI Institute, The State of AI in the Enterprise 2026: The Untapped Edge (press release and report page) | 2026-01-21 | 3,235 business and IT leaders across 24 countries and six industries, August to September 2025 | report page, 'What does AI do for business?' | “Improving productivity and efficiency top the list of benefits achieved from enterprise AI adoption so far, with two-thirds (66%) of organizations reporting gains” |
| 50 | Only 25% of organizations have moved 40% or more of their AI pilots into production; 34% say they are using AI to deeply transform the business, 37% only at a surface level. | Deloitte AI Institute, The State of AI in the Enterprise 2026: The Untapped Edge (press release and report page) | 2026-01-21 | 3,235 business and IT leaders across 24 countries and six industries, August to September 2025 | press release, 'Bridging the pilot-production gap' | “only 25% of respondents have moved 40% or more of their AI pilots into production” |
| 51 | 76% of enterprise AI use cases are bought rather than built in-house, up from 53% a year earlier; AI purchases reach production at 47%, against 25% for ordinary software. | Menlo Ventures, 2025: The State of Generative AI in the Enterprise | 2025-12-09 | 495 U.S. enterprise AI decision-makers at companies actively using AI tools, 7 to 25 November 2025 | 'Enterprises Are Buying More Than Building' and 'AI Buyers Convert at Higher Rates' | “Today, 76% of AI use cases are purchased rather than built internally” |
| 52 | Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027, citing cost, unclear value and weak risk controls. | Gartner, Gartner press release, 25 June 2025 | 2025-06-25 | Gartner analysis; the poll behind the release is the 3,412-attendee webinar poll | headline and first paragraph | “Over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls” |
| 53 | Gartner estimates only about 130 of the thousands of vendors calling their products agentic AI are real; the rest are 'agent washing' assistants, RPA and chatbots. | Gartner, Gartner press release, 25 June 2025 | 2025-06-25 | Gartner analysis | fourth paragraph | “Gartner estimates only about 130 of the thousands of agentic AI vendors are real” |
| 54 | Gartner predicts 40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% in 2025. | Gartner, Gartner press release, 26 August 2025 | 2025-08-26 | Gartner analysis | first paragraph and 'Stage 2' | “Forty percent of enterprise applications will be integrated with task-specific AI agents by the end of 2026, up from less than 5% today” |
| 55 | Gartner predicts at least 15% of day-to-day work decisions will be made autonomously by agentic AI by 2028, from 0% in 2024, and 33% of enterprise software will include agentic AI by then. | Gartner, Gartner press release, 25 June 2025 | 2025-06-25 | Gartner analysis | 'Realizing Business Value' | “Gartner predicts at least 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028, up from 0% in 2024” |
| 56 | Trust in fully autonomous AI agents fell from 43% to 27% of organizations in one year. | Capgemini Research Institute, Rise of agentic AI: How trust is the key to human-AI collaboration | 2025-07-16 | 1,500 leaders at director level and above, 14 countries, large organizations; 900 re-confirmed their adoption answer | Executive summary | “Only 27% of organizations express trust in fully autonomous AI agents, from 43% 12 months ago” |
| 57 | Only 15% of business processes are expected to run semi- or fully autonomously within 12 months, rising to 25% by 2028. | Capgemini Research Institute, Rise of agentic AI: How trust is the key to human-AI collaboration | 2025-07-16 | 1,500 leaders at director level and above, 14 countries, large organizations; 900 re-confirmed their adoption answer | Executive summary | “in 12 months, only 15% of all business processes are expected to operate at Level 3 (semi-autonomous) to Level 5 (fully autonomous). This share is expected to grow to 25% by 2028” |
| 58 | Fewer than one in five organizations report high data readiness for agents, and over 80% lack mature AI infrastructure. | Capgemini Research Institute, Rise of agentic AI: How trust is the key to human-AI collaboration | 2025-07-16 | 1,500 leaders at director level and above, 14 countries, large organizations; 900 re-confirmed their adoption answer | Executive summary | “fewer than one in five organizations report high levels of data-readiness, and over 80% lack mature AI infrastructure” |
| 59 | Only 21% of companies planning agentic AI have a mature model for governing agents. | Deloitte AI Institute, The State of AI in the Enterprise 2026: The Untapped Edge (press release and report page) | 2026-01-21 | 3,235 business and IT leaders across 24 countries and six industries, August to September 2025 | press release | “only 21% of those companies report having a mature model for agent governance” |
| 60 | The most chosen way to manage agent risk over the next year is a human-in-the-loop model, picked by 52% of large U.S. companies: a person checks outputs but does not watch every action. | KPMG, AI Quarterly Pulse Survey, Q2 2026 (U.S.) | 2026-06 | 204 U.S. C-suite and business leaders at organizations with annual revenue of 1 billion dollars or more, 28 April to 25 May 2026 | 'Accountability and cost visibility' | “The top risk mitigation approach to AI agents in the next 6-12 months is 52% a "human-in-the-loop" model where a person validates outputs but does not oversee each agentic action or decision” |
| 61 | One-third of senior leaders (33%) cite limited understanding of usage costs as a key challenge in deploying agents; leaders with strong cost visibility are five times as likely to report a return (15% versus 3%). | KPMG, Global AI Pulse: Q2 2026 (press release) | 2026-06-24 | 2,145 C-suite and senior business leaders across 20 countries, 28 April to 25 May 2026, at organizations with annual revenue above 50 million dollars (the press release's About section sets a 100 million floor for its ten largest markets) | highlights | “One-third of leaders (33 percent) cite limited understanding of usage costs, as a key deployment challenge for AI agents” |
| 62 | U.S. executives rank cybersecurity and cost as the top agent challenges (34% each); trust in agents for financial transactions is 20%, for data analysis 38%. | PwC, PwC's AI Agent Survey | 2025-05 | 308 U.S. executives (C-suite 33%, VP 13%, director 54%), surveyed 22 to 28 April 2025 | '3. People are the problem' and the trust paragraph | “respondents expressed the highest levels of trust in areas like data analysis (38%), performance improvement (35%) and daily collaboration with human team members (31%). But trust dropped sharply for higher-stakes activities like financial transactions (20%)” |
| 63 | Quality is the top barrier to putting agents in production, cited by 32% of builders; 89% have monitoring in place but only 52% run evaluations. | LangChain, State of Agent Engineering | 2026-06-12 | over 1,300 professionals building AI agents: engineers, product managers, business leaders and executives | 'Biggest barriers to production' and 'Observability for agents' | “Quality is the production killer, with 32% citing it as a top barrier” |
| 64 | CEOs say 25% of operational decisions are already made by AI without human intervention where rules can be written down, and expect 48% by 2030. | IBM Institute for Business Value, 2026 CEO Study: Rewiring the C-suite (press release) | 2026-05-04 | 2,000 CEOs and equivalent leaders across 33 geographies and 21 industries, February to April 2026, with Oxford Economics | 'As CEOs turn to AI-driven decisions, governance and controls become more critical' | “By 2030, surveyed CEOs expect 48% of operational decisions where consistency and guardrails can be codified will be made by AI without human intervention, compared to 25% today” |
| 65 | 64% of CEOs are comfortable making major strategic decisions based on AI-generated input; 76% of their organizations now have a Chief AI Officer, up from 26% a year earlier. | IBM Institute for Business Value, 2026 CEO Study: Rewiring the C-suite (press release) | 2026-05-04 | 2,000 CEOs and equivalent leaders across 33 geographies and 21 industries, February to April 2026, with Oxford Economics | key findings | “76% of surveyed organizations have a Chief AI Officer (CAIO) in 2026, up from just 26% in 2025” |
| 66 | 32% of organizations expect AI to cut their total workforce by 3% or more in the coming year; 43% expect little or no change; 13% expect growth. | McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation | 2025-11-05 | 1,993 respondents in 105 nations, fielded 25 June to 29 July 2025, weighted by each nation's share of global GDP | 'Expectations vary on AI's effect on workforce size', Exhibits 16 and 17 | “32 percent predict an overall reduction of 3 percent or more, and 13 percent predict an increase of that magnitude” |
| 67 | AI-related employment decreases occurred in only 2% of U.S. firms between November 2025 and January 2026. | U.S. Census Bureau, The Microstructure of AI Diffusion (working paper CES-WP-26-25) | 2026-04-15 | BTOS AI supplement, nationally representative, reference period Nov 2025 to Jan 2026 (working paper CES-WP-26-25) | abstract | “AI-related employment decreases are rare, occurring in only 2% of firms” |
| 68 | 82% of small businesses that use AI increased their workforce over the past year. | U.S. Chamber of Commerce (with Teneo Research), Empowering Small Business: The Impact of Technology on U.S. Small Business, fourth edition | 2025-08-18 | 3,870 U.S. small businesses with fewer than 250 employees, surveyed online 6 to 26 June 2025 | Introduction, key findings list | “82% of small businesses using AI increased their workforce over the past year” |
| 69 | CEOs expect 29% of employees to need reskilling for a different role and 53% to need upskilling for their current one by 2028. | IBM Institute for Business Value, 2026 CEO Study: Rewiring the C-suite (press release) | 2026-05-04 | 2,000 CEOs and equivalent leaders across 33 geographies and 21 industries, February to April 2026, with Oxford Economics | 'Organizations are betting on people to drive AI success' | “respondents expect 29% of employees to require reskilling for a different role and 53% to need upskilling to perform their current role more effectively” |
| 70 | CEOs say only 25% of their workforce uses AI regularly, although 86% believe employees have the skills to work with it; 83% say AI success depends more on people's adoption than on technology. | IBM Institute for Business Value, 2026 CEO Study: Rewiring the C-suite (press release) | 2026-05-04 | 2,000 CEOs and equivalent leaders across 33 geographies and 21 industries, February to April 2026, with Oxford Economics | key findings | “Surveyed CEOs say only 25% of the workforce is using AI regularly as part of their job, despite 86% believing their employees have the skills to collaborate with AI” |
| 71 | 61% of organizations report rising employee anxiety about the effect of AI agents on their jobs, and over half believe agents will displace more jobs than they create. | Capgemini Research Institute, Rise of agentic AI: How trust is the key to human-AI collaboration | 2025-07-16 | 1,500 leaders at director level and above, 14 countries, large organizations; 900 re-confirmed their adoption answer | Executive summary | “61% of organizations report rising employee anxiety about the impact of AI agents on their employment prospects, and over half believe AI agents will displace more jobs than they create” |
| 72 | Organizational factors such as culture, manager support and talent practices account for 67% of reported AI impact, individual factors 32%. | Microsoft WorkLab, 2026 Work Trend Index Annual Report | 2026-05-05 | 20,000 knowledge workers who use AI at work, 10 countries, fielded by Edelman Data x Intelligence from 18 February 2026 to 7 April 2026 per the methodology section (two chart captions on the same page say to 20 April 2026), plus Microsoft 365 telemetry | 'The job of every leader is to rearchitect work' | “organizational factors 9 like culture, manager support, and talent practices account for more than 2x the reported AI impact of individual factors like mindset and behavior (67% vs. 32%)” |
| 73 | Only 26% of AI users say their leadership is clearly and consistently aligned on AI; 86% treat AI output as a starting point rather than an answer. | Microsoft WorkLab, 2026 Work Trend Index Annual Report | 2026-05-05 | 20,000 knowledge workers who use AI at work, 10 countries, fielded by Edelman Data x Intelligence from 18 February 2026 to 7 April 2026 per the methodology section (two chart captions on the same page say to 20 April 2026), plus Microsoft 365 telemetry | 'The job of every leader is to rearchitect work' and 'AI lifts the ceiling on individual potential' | “Only one in four AI users surveyed (26%) say their leadership is clearly and consistently aligned on AI” |
| 74 | Employees whose manager actively supports AI use are 1.7 times as likely to use it weekly; only 36% strongly agree their manager supports their team's AI use. | Gallup, Global Indicator: Artificial Intelligence (workplace AI adoption tracker) | 2026-05 | Gallup Panel, probability-based random sample of U.S. adults employed full or part time; figures as of May 2026 | 'Do Managers Support Employee Use of AI?' | “As of May 2026, nearly four in 10 employees (36%) in AI-integrating organizations strongly agree that their manager supports their team's use of AI” |
| 75 | Employment of software developers aged 22 to 25 has fallen nearly 20% from 2024. | Stanford Institute for Human-Centered AI, AI Index Report 2026, chapter 4 (Economy) | 2026-04 | labor-market data summarized by the AI Index | Chapter highlights, item 7 | “Employment for software developers ages 22 to 25 has fallen nearly 20% from 2024” |
Questions people ask
How many companies use AI agents in 2026?
It depends on what counts. In McKinsey’s 2025 global survey, 62 percent of organizations were at least experimenting with AI agents, 23 percent had scaled one somewhere in the company, and in any single department no more than 10 percent had. Surveys that count any adoption or intent report higher figures, such as PwC’s 79 percent of U.S. executives saying agents are being adopted. Across all U.S. businesses of every size, the Census Bureau found one in five using AI at all in May 2026.
Is the statistic that 95 percent of AI projects fail true?
It is a real finding that is usually quoted wrongly. MIT Project NANDA’s July 2025 report found that 95 percent of organizations saw no measurable profit-and-loss impact within roughly six months of a generative AI pilot, based on 52 interviews, 153 survey responses at conferences and 300 public deployments (the report), and the authors call the findings preliminary. It measures short-term profit impact in a small sample. It does not say that 95 percent of AI projects fail.
What percentage of AI agent projects will be canceled?
Gartner predicts that over 40 percent of agentic AI projects will be canceled by the end of 2027 because of cost, unclear value and weak risk controls. That is a forecast made in June 2025. Nobody has counted canceled projects yet. The same release estimates that only about 130 of the thousands of vendors selling agentic AI are real.
What is the difference between experimenting with AI agents and scaling them?
Experimenting means a company has tried an agent somewhere, often in a pilot with a few users. Scaling, in McKinsey’s definition, means expanding the deployment within at least one business function so that it runs as part of the work. In 2025, 39 percent of organizations were experimenting and 23 percent were scaling, and most of those scaling did so in only one or two functions.
Do AI agents replace jobs?
So far the recorded numbers are small and the expectations are large. In McKinsey’s 2025 survey, 32 percent of organizations expected AI to cut their workforce by 3 percent or more in the coming year, while a median of 17 percent reported a decline in a function over the past year. The U.S. Census Bureau found AI-related employment decreases in only 2 percent of firms between November 2025 and January 2026. Among small businesses that use AI, 82 percent increased their workforce, according to the U.S. Chamber of Commerce.
Cite this page
bdautomated (2026). AI agent statistics 2026: every number checked at its source. https://bdautomated.com/ai-agent-statistics/ (updated 15 September 2026). Data: CC BY 4.0.
Changelog
- 15 September 2026: first published. 75 figures from 26 documents by 18 publishers, each opened at its source.