June 09, 2026
Agentic AI Is Outpacing the Proof It Works
Companies that succeed with agentic AI tend to share one habit. They rebuilt how they work before letting the technology run.
That finding comes from research by enterprise software firm Pegasystems, which surveyed more than 500 business and IT decision makers worldwide, all of whom had already put agentic AI projects into practice.
Almost every successful implementer (96%) had rethought existing processes, and just over half said they had done so to a significant degree, rebuilding much of how their organisation operates to get the most from agentic tools. The majority said business and IT teams were willing to try new technology and ideas, a sign that companies were dropping old habits in favour of fresh ways to work together.
Metrics Agreed Before the Agents Arrived
Success has less to do with the software and more to do with the discipline around it. Most of the successful companies (95%) reported a defined corporate strategy and execution plan, and roughly two thirds had agreed success metrics tied to business outcomes that they reviewed regularly. Their main aim before deployment was to automate and simplify complex processes so they run the same way every time, across systems and platforms. More than half said this approach had already delivered predictable outcomes, less complexity, and better customer experiences.
Asked what holds projects back, many pointed to a lack of resources, and a similar number named limited knowledge of what agentic AI can do as the single biggest obstacle to getting value from it.
Don Schuerman, Pega’s Chief Technology Officer, said the technology has hit a tipping point where “adoption is high within organisations, but maturity is not.” He argued that value comes from rethinking how people work and aligning company culture around what the technology makes possible. The culture point echoes a separate study in which executives admitted their customer experience was broken and blamed internal dynamics rather than missing technology.
Where the Proof Runs Thin
If agents are being rolled out everywhere, why is so little of it surviving past the pilot stage? According to Gartner, more than 40% of agentic AI projects will be cancelled by the end of 2027, blaming rising costs, uncertain value, and weak risk controls, and it reckons only around 130 of the thousands of self-described agentic vendors are the real thing. MIT Sloan Management Review and Boston Consulting Group found agentic AI spread to about a third of organisations in two years, quicker than any earlier wave of AI, while most of those deployments never left the pilot stage. Deloitte counted the share actually running in production in the low double digits.
Pega’s central claim places the burden of success on the buyer, on culture, discipline and process, while leaving aside a harder question of whether the technology is yet ready to do the work. With so many projects stalling before they reach production, that is the question truly worth pressing.
None of which means agentic AI is a dead end. Enthusiasm for it keeps climbing, but appetite is not the same as evidence, and the two are still moving in different directions. The companies getting value are doing the slow work of redesigning how they operate around agents. The rest are buying the promise and waiting for the proof, which, for now, is the part still missing.
