//
(Alan Trefler, CEO, Pega, during the opening general session of PegaWorld 2025. Source: Pega.)
Alan Trefler, CEO of Pega (Cambridge, Mass.), used the “Ask Alan Anything” session at PegaWorld iNspire 2025 to reinforce a recurring theme at the show: that generative AI should be used only at design time—never at runtime—if businesses are to build systems that are predictable, reliable, and legally defensible.
Trefler criticized “prompt-driven” AI architectures for relying too heavily on models that are inherently non-deterministic, calling them “expedient, but short-sighted.” He said that prompting alone is not a sound foundation for business-critical workflows. “Generative AI should be paired with mathematical, statistical AI,” he said. “They both have tremendous value, but knowing when and where to use each is critical.”
Predictability, Efficiency, and CO₂
A central theme of Trefler’s remarks was the energy cost of large language models (LLMs). Referencing a chess-related experiment from his keynote, he compared the high energy consumption of generative AI to the ultra-efficient Stockfish chess engine, which achieves superior results without massive GPU usage. “When you conduct a workflow through pre-considered paths, you are just hundreds of times more efficient,” he said.
To the objection that chess was a closed system, however complex, compared with the kind of open-ended problems often addressed by generative AI, Trefler noted that many business processes are also closed systems. “Chess is a limited system without a question—though it’s been considered a pretty complicated limited system for millennia,” commented. “But then so is the decision to issue an insurance policy or pay a claim.”
“You don’t have to reason on things you already know about,” he added. “And that’s where you save time, money—and CO₂.”
This efficiency, he concluded, not only improves speed and cost—but also mitigates the environmental impact of AI deployment. “All this reasoning is where the expense—but also the CO₂—comes from.”
Blueprint-Driven Transformation
Trefler emphasized that Blueprint, Pega’s generative design tool, democratizes system architecture by allowing users to describe outcomes in natural language. The system generates working prototypes complete with workflows, data models, personas, and interface elements. “It shows you what the screens look like. It can even conduct the process as a conversation,” he said.
He shared a recent example in which a senior banking executive asked to model a flavored water business. Within minutes, Blueprint generated a functioning prototype. “It understood supply chain needs, marketing strategy, the digital presence—it stunned me,” Trefler said.
Partner Strategy and Distribution Expansion
Trefler also described a major change to Pega’s go-to-market model. System integrators such as Accenture and EY can now embed proprietary IP into vector databases that pair with Pega agents. This allows partners to deploy customized solutions through Blueprint blueprints that carry their own branding. “We’re not insisting that Pega be in the middle of every transaction anymore,” Trefler said.
The company has also made its software available on AWS Marketplace, with Google Cloud support to follow.
Job Displacement and Workforce Evolution
Addressing concerns about AI-driven job loss, Trefler said change is inevitable, citing as an example how the effectiveness of AI diminishes the need for interpreters and translators. But he emphasized that productivity and communication will improve. “Suddenly, language becomes much less of an impediment,” he said.
He predicted that software development teams will look very different in the next few years, not because AI will write all the code, but because the process of designing systems will be transformed.
Cybersecurity by Design
In response to a question about cyber threats, Trefler said Pega’s security model is baked into its workflows. As agents interact with applications, they inherit the same data visibility rules as human users. “What we built last year is the same infrastructure our agents use today,” he said. “That’s the kind of symmetry you want.”
Cultural Change and the Power of Persistence
Trefler closed with reflections on organizational culture. He emphasized the importance of what he calls an “external lens”—constantly measuring success against the broader market, not internal baselines. “If you compare yourself to yourself, you’re not going to win,” he said.
When asked for his favorite English word, Trefler didn’t hesitate: “Persistence.”
Pega Expands Blueprint AI Tools for Legacy System Modernization
Pega Launches Agentic Process Fabric for AI Workflow Orchestration
Pega CEO Alan Trefler Champions ‘Predictable AI’ at Scale
How Pega Tames AI for Safe Business Impact
Source link
