Business Process Automation 2026: Adoption Wide, AI Agents Thin

Automation is widespread but shallow

The 2026 numbers on business process automation say two things at once. Automation is now normal at large companies. The newer layer of autonomous AI agents is mostly still in pilots, and the market forecasts built on top of all this disagree so much that they are hard to trust.

The broadest figures look strong. In Camunda's 2026 survey of 1,150 decision-makers at large organizations, respondents said 48% of their processes are automated on average and that 64% could be.22 McKinsey's data, as compiled by industry trackers, shows 66% of organizations have automated at least one business function, up from 57% the year before.25 The Business Research Company puts the business process automation market at $18.83 billion in 2026, up from $16.32 billion in 2025, and expects it to reach $33.43 billion by 2030 at a 15.4% compound annual growth rate.2428

A closer look gives a different picture. The same compilation that cites Camunda's 48% also notes that only 27% of an organization's applications are connected to each other, according to MuleSoft. It also reports that just 11% of agentic AI use cases reached production last year.22 Across the whole US economy, the Census Bureau found only about 17% to 20% of businesses used AI between December 2025 and May 2026.22 Other trackers report that only 4% of businesses run fully hands-free automated operations and only 21% run AI workflows at enterprise scale.29

The reading that best fits this data is that automation in 2026 covers a lot of ground but doesn't go deep. Companies have automated many separate tasks. Far fewer have connected those tasks into end-to-end workflows that run without people stepping in.

AI agents: high expectations, few deployments

AI agents are the main reason enterprise software vendors keep saying "workflow automation," and they show the gap between plans and practice most clearly. In its 2026 Hype Cycle for Agentic AI, Gartner placed the technology at the Peak of Inflated Expectations. It cited its CIO survey, which found that only 17% of organizations had deployed AI agents while more than 60% expected to within two years. Gartner called that the steepest adoption curve of any emerging technology in the survey.11 The analyst firm also warned that most deployments are narrowly scoped and that fully autonomous agents are not ready for most enterprise use cases.11

The adoption figures vary widely depending on who is asked and how. A Gartner preview released on August 31, 2026 reported that 37% of respondents had already deployed agents and another 34% planned to within 12 months. Its public summary gave no sample size.14 KPMG's US AI Pulse for Q3 2026, which surveyed 314 leaders at companies with more than $1 billion in revenue, found 62% building, deploying or developing agents. It found 25% working on multi-agent systems, up from 6% in each of the two previous quarters.14 At the high end, EY found 91% of senior AI executives at large public companies using agentic AI in pilots or production. PwC's 2025 survey reported 79% adoption.14

These figures don't contradict each other so much as measure different things. One analysis pointed out that McKinsey asked who was scaling agents, Gartner asked who had deployed them, and IBM asked who felt fully ready. It found IBM's figure was 11% of tech leaders in June 2026.16 A survey that counts pilots and self-reported use will always give a bigger number than one that asks about production systems. Some marketing material has blurred that line, for example by presenting PwC's 79% as proof that agents are already "in production."19

One forecast is now due. In August 2025, Gartner predicted that 40% of enterprise applications would include task-specific AI agents by the end of 2026, up from less than 5% in 2025.12 That prediction is about vendors adding agent features to their software, not about customers actually using them. Vendors adding agent features to SaaS products may well hit the target while real use of those features stays low. Gartner has also forecast that more than 40% of agentic AI projects will be canceled by the end of 2027 because of rising costs, unclear business value or weak risk controls.1714

Two other risks stand out. Gartner's Hype Cycle named "agent-washing," meaning RPA and automation tools rebranded as agents, as a market problem. One estimate cited by the same tracker found that only 16% of deployments qualify as agents in the planning-and-acting sense.16 A Berkeley study found that seven open-source multi-agent frameworks failed between 41% and 86.7% of their tasks.13

RPA: the forecasts disagree by billions

Robotic process automation is the older, rules-based technology underneath much of this work, and its market data shows how unreliable the forecasts are. Grand View Research estimates the global RPA market at $6.0 billion in 2026, growing to $35.8 billion by 2033.1 Future Market Insights puts it at $5.7 billion in 2026.4 Mordor Intelligence says $8.12 billion.5 Fortune Business Insights says $27.22 billion.6 Precedence Research says $35.27 billion and projects $247.34 billion by 2035.27

The highest 2026 estimate is about six times the lowest. The firms define the market differently, especially in whether they count services. Fortune Business Insights says services make up 79.35% of the market in 2026.6 Precedence says services held a 77.21% share in 2025.7 A firm that counts consulting and managed services will report a much larger market than one that counts only software licenses. One fact-checking compilation made the same point about workflow automation, noting that published 2024 estimates differ several-fold because each firm defines the market its own way.22 The useful lesson is to compare a firm's forecast only against that firm's own earlier figures, never against another firm's.

The forecasts do agree on direction. All of them expect fast growth, with projected annual rates from 18.2% to about 29%.419 They also agree on why: RPA is being combined with AI, OCR and natural language processing, and moving to cloud subscriptions.75 Mordor Intelligence describes companies combining RPA, AI and process mining in a hyperautomation market it expects to grow from $18.64 billion in 2026 to $45.17 billion by 2031.25 RPA is not being replaced by agents. It is becoming the reliable layer that agents call on to act.

No-code: who builds automation is changing

The change in who builds automations matters as much as the agent story. Gartner has projected that 80% of low-code users will work outside IT departments by 2026, up from 60% in 2021.3631 Vendors are responding by making generative AI a standard feature. Gartner's definition of the enterprise low-code platform category now explicitly includes generative AI capabilities.31 Tools such as Zapier, Make, n8n and Power Automate now include copilots that build workflows from plain-English descriptions, along with agent components for LLM-driven workflows.35 UiPath, the best-known RPA company, sells a shared environment where citizen developers build low-code workflows that IT staff can then review.38

Much of the no-code coverage leans on older predictions that are now past due. A widely cited Gartner forecast said 70% of new applications would use low-code or no-code by 2025. Gartner never published a follow-up measurement.22 The same compilation found that the popular claim that 69% of managerial work would be fully automated by 2024 came from a 2020 prediction that was never checked.22 That claim still appears in 2026 roundups.30 Readers should treat any automation figure without a recent, named survey behind it as marketing.

What the 2026 data shows

Put together, the coverage supports a fairly clear conclusion. Basic process automation is a mature market that is still growing. Task-level automation is common at large companies, and workers still have plenty of manual work that could be automated. One widely cited estimate says about a quarter of the work week still goes to repetitive tasks.25 The weak spots are integration and governance. With only 27% of applications connected, it's hard to build automation that runs end to end.22 Another tracker found that only 33% of organizations report mature governance frameworks for AI automation.29

That is why 2026 looks like a transition year for agentic workflows. Vendors are adding agent features to their software faster than customers can safely put them into production. If Gartner's cancellation forecast holds, many of this year's pilots will be dropped. The companies likely to benefit are the ones that first connected their systems, kept rules-based RPA handling the predictable steps, and put approval checkpoints in front of agents before giving them more autonomy. The gains are real, but they will come from that groundwork rather than from rebranding older automation as agents.