Pegasystems’ Alan Trefler Warns Against Mass AI Agents

Alan Trefler, founder and CEO of Pegasystems, warned recently that mainstream enterprise software suppliers are pushing companies to deploy large numbers of autonomous AI agents and that doing so risks unpredictable failures in mission‑critical systems. He described the idea of running thousands of agentic processes as ‘madness.’

Trefler said many vendors expect to operate thousands, potentially more than 10,000, of AI agents and are building ‘control towers’ to monitor them. He argued those monitoring tools do not address the core issue: agents can behave unpredictably and, when given authority over critical decisions, might treat customers inconsistently or violate laws and regulations. He pointed to instances where AI-generated marketing output provoked public backlash as examples of how automated systems can fail at scale.

The Pegasystems founder also raised economic concerns. He warned that rapid investment in data centres and AI infrastructure, together with rising costs for model usage, is not sustainable. He described the current market as overheated and said many smaller AI vendors may not survive an industry shakeout, comparing the situation to past infrastructure bubbles. ‘There was billboard after billboard of AI companies. I’m in the business, and I didn’t know 80% of them,’ he recalled.

Trefler said Pegasystems supports selective AI use. The company uses AI to help build deterministic workflows in its low‑code business process platform, which serves regulated sectors including telecoms, banking and automotive. He argued enterprises that need consistent process outcomes should rely on rules‑based systems rather than fleets of loosely coordinated agents. ‘You don’t want these disaggregated, disassociated initiatives trying to run important things in the business where it might treat customers differently in ways that it shouldn’t,’ he added.

Industry analysts agree caution is warranted for high‑volume, mission‑critical workflows such as claims eligibility checks and regulatory processes. Neil Ward‑Dutton, research vice‑president for agentic automation and AI technologies at IDC, warned that replacing deterministic systems in those contexts with agent fleets would be ‘an absolute disaster.’ At the same time, he noted many vendors are adopting hybrid approaches that combine AI agents with defined workflows and guardrails to reduce risk for less critical tasks.

Several large software suppliers now offer templates, workflow engines and curated plug‑ins that limit agent behaviour. Vendor platforms are being positioned to let agent assistants handle routine desktop work, marketing campaign assembly, translation or catalogue checks, while core transactional systems remain governed by rules and controls.

Trefler described how AI and low‑code tools are changing IT work. Pegasystems’ low‑code products allow people without deep technical skills to design business applications, shifting some development away from specialist programmers. Large language models are generating higher volumes of code, which is compressing some traditional IT roles. He cautioned that producing more code does not necessarily make systems more reliable. ‘The trouble is, we’ve all learned that having more and more code in the business doesn’t make that business more reliable,’ he observed.

On measuring AI’s business impact, Trefler urged companies to focus on concrete financial outcomes rather than internal usage metrics such as token consumption. He said the only credible metric is whether AI helps a company save or make money.

Pegasystems, founded in 1983, provides a low‑code business process platform to large enterprises and counts customers such as Deutsche Telekom, Lloyds Banking Group and Daimler Trucks. The company is on track to approach $2 billion in revenue. Trefler’s comments contribute to a broader industry debate over the appropriate role and scale of agentic AI in enterprise IT.