Pega CEO Alan Trefler
Pega CEO Alan Trefler says competitors asking users to write prompts are “setting themselves up for slews of challenges” as the vendor unveils Agentic Process Fabric to orchestrate predictable AI automation
If you’ve been following the enterprise AI agent hype train, you’ll have noticed a common theme – vendors pushing prompt studios, asking users to craft the perfect instructions to make their AI agents particularly useful. There has been an ongoing discussion in the industry about how to effectively up-skill employees for effective Artificial Intelligence (AI) prompting, to get the most out of AI investments. Pega CEO Alan Trefler thinks that’s fundamentally the wrong approach. Speaking ahead of PegaWorld in Las Vegas this week, he offered a sharp critique of where the market is heading.
Anybody who’s asking people to write prompts where those prompts are going to run their business is setting themselves up for slews of challenges.
These prompts can respond very differently to even subtle changes in the data they receive. Whereas we know from our extensive experience that workflows are extremely reliable and already capable of running businesses.
It’s a stance that positions Trefler and Pega somewhat uniquely in the market, given that most other platforms and vendors see prompts and natural language instructions as a unique advantage to the future of enterprise business. But Pega is zigging where others zag, unveiling Pega Agentic Process Fabric – a service that orchestrates AI agents through pre-designed workflows rather than runtime prompting.
The distinction matters more than you might think. While the rest of the market talks about thousands of agents needing control towers to manage them, Pega says your existing workflows can become intrinsically agentic. It’s the difference between hoping an AI agent figures out what to do versus knowing it will follow a predictable path – and Pega hopes that this will be compelling for nervous buyers.
The problem with prompt engineering at scale
Instead of focusing on a future whereby employees can ask their army of agents to carry out a myriad of tasks using natural language prompts, Pega argues that agents could be used to solve a more immediate challenge for buyers: technical debt. Alongside its product updates this week, Pega has released some research which states that 68 percent of IT decision makers say legacy systems are preventing their organizations from fully embracing modern technologies like AI. Nearly half (48 percent) can’t stop supporting their legacy applications because they’re still business critical, according to the report. Some organizations run applications that are 20-30 years old.
Against this backdrop, Pega CTO Don Schuerman frames the challenge:
The big gap for a lot of organizations implementing this in a real way is sort of this black box problem. If I deploy agents at runtime, I don’t have a huge amount of predictability and explainability and consistency in how they operate. And if I have agents making recommendations to customers where there’s no predictability, or where I can’t trace where that recommendation came from or why they chose a particular process, that’s not going to fly for an organization that’s trying to sell products into a regulated industry.
This isn’t theoretical hand-wringing. The research shows 88 percent of respondents are concerned about how technical debt impacts their ability to keep pace with more agile competitors. More than half acknowledge their legacy systems likely cause customers to defect due to poor experiences. Adding unpredictable AI agents on top of this mess isn’t a solution – it’s a recipe for chaos, according to Pega.
Enter the Agentic Process Fabric
So what’s the alternative being offered? The company today has announced Pega Agentic Process Fabric, which extends its existing Process Fabric architecture to create what it calls “a network of predictable agents that automate work across the enterprise.”
The key point here is timing. Pega uses AI reasoning during workflow design – where it sees creativity and exploration as assets – but switches to semantic AI execution at runtime when predictability matters. As Schuerman explains:
What we’ve taken is an approach that combines the power of agents to reason and think and act with the predictability that you get from workflows, from knowing that there are certain things you want to do in the business where you actually want to follow a predetermined path.
It starts with Pega Blueprint, which the company calls its “design agents.” These agents can ingest documentation, pull in best practices, research process options, and suggest new workflows. But – and this is crucial – humans validate and approve these workflows before they go live. The AI creativity happens at design time, not runtime. Schuerman notes:
By shifting that agentic reasoning to the design time conversation, we actually make the fact that agents are unpredictable and creative a good thing, because the agents can now suggest new ideas for a process. They can suggest new ways of working the business might not have thought of but the business still gets to control what is finalized.
How it actually works
Matt Healy demonstrated the system during the briefing, showing how the Agentic Process Fabric creates a registry of workflows, systems, AI agents (both Pega and non-Pega), and data across the enterprise.
In his demo, he showed a Customer Relationship Management (CRM) scenario where the fabric connected CRM workflows and data, account management and escalation processes, and AI agents from both Pega and Salesforce. The system can then process deals, create incidents, and track marketing efforts. Healy says:
I then have a sort of overarching AI agent which can help me stitch everything together into a unified agentic experience. It’s going to traverse my library of connected AI agents and available workflows, figure out which is the optimal based on my request, which one should be called, and then actually walk me through that process of kicking that off and actually getting that work done.
According to Pega, this isn’t just about making existing processes prettier with conversational interfaces. It’s about changing how work gets orchestrated across systems. The fabric also supports emerging standards like Model Context Protocol and Agent-to-Agent communication, meaning it can orchestrate agents across multiple platforms.
Legacy transformation through AI design
The most compelling part of Pega’s announcement is how it addresses the legacy modernization challenge facing buyers. During the pre-brief with media and analysts, it felt like Pega was pointing its product towards a very specific use case: technical debt (which is welcome in a world where it feels like AI tools are released in search of a problem, rather than working the product back from the use case). Kerim Akgonul, Senior Vice President and Chief Product Officer at Pega, demonstrated how organizations can record their legacy applications in action, upload these recordings to Blueprint, and have AI agents analyze and recreate modernized workflows. He explains:
We’re basically saying you can take these applications and do a recording of it. Sit down in front of this application, start your screen recording, run through the functionality, narrate it and capture a video that shows how this application essentially works.
The process represents a shift from traditional legacy modernization approaches. As Akgonul notes:
“What Blueprint will do is go through and essentially what a human would do in months, it will just basically analyze all the content and actually digest it and basically provide guidance on what this application should do.
In his demonstration, Blueprint analyzed a recorded banking application and quickly identified it as a credit card management solution. The system determined the modernized version would handle four key workflows: credit limit adjustments, account maintenance, transactions, and billings. Akgonul explains:
Using all the information that we have from that context that we saw, as well as definitions that we found in the video and in the documentation that came from AWS Transform, we determined that the modernized version of that application will handle these four workflows.”
The competitive landscape
Trefler didn’t hold back when discussing how Pega’s approach differs from competitors. I asked him how he thinks Pega’s strategy differs from ServiceNow’s approach, which similarly seeks to use agents to automate processes across the enterprise – and in true Trefler fashion, he didn’t stutter on his point:
ServiceNow talks about the right things, but actually look at how it’s implemented, see how radically different the Pega architecture is from the ServiceNow architecture. They just went out and bought Moveworks to try to bring a prompt-based approach to running agents. I’ve heard Bill [McDermott, CEO of ServiceNow] talk about thousands and thousands of agents that need a control tower to manage them.
In contrast, Pega takes what Trefler calls “a completely different approach”:
We’re basically saying that your workflows, including for Pega customers, workflows that they’ve already built, are now all intrinsically agentic, so that you’re going to be able to use them as part of a repeatable, predictable way to run your business.
The key differentiator, according to Trefler, is the absence of prompt engineering in Pega’s approach:
A lot of these words sound the same, but I think when you peel it back just a little, you don’t see any prompt authoring in what Kerim did. Not at all. It’s because it’s workflow authoring, and those workflows are now being authored by agents.
What this means for enterprise buyers
For organizations struggling with technical debt and legacy systems, Pega’s approach certainly offers a pragmatic path to AI adoption. Instead of asking IT teams to become prompt engineering experts overnight whilst in search of tangible use cases, it uses AI to accelerate the journey to modern architectures. Technical debt is an ongoing challenge for buyers and there’s definitely returns to be sought in modernization.
Pega’s research backs up this need – 74 percent of respondents admit their businesses prioritize investments that improve profitability over customer experience improvements. This echoes earlier Pega research showing 69 percent of consumers feel businesses prioritize profits over positive experiences. The result? One in three organizations saw average customer query resolution times increase 26-50 percent in the last year.
Schuerman articulates why this matters for competitiveness:
There’s so much hype in the AI agent space, but not a ton of adoption in the enterprise. And the reasons are risk and compliance number one, and that’s where that predictable AI agent approach comes into play, and then the second is integration of AI agents and core operations, core systems within an enterprise.
Pega also opens up Blueprint to partners, allowing them to embed their industry expertise and best practices directly into the platform. Schuerman sees this as important:
Blueprint really now becomes a showcase, not just to the power of workflow, automation and orchestration, but more importantly, of our partners, unique IP and unique knowledge in certain areas. We think this is a really compelling way for our partners, both to showcase their unique skills, to up level some of the conversations that they’re having to more of a transformation and a consultative conversation, and to make that vision instantly real for our joint clients.
My take
Pega’s approach to AI agents feels grounded in the reality of enterprise IT. Pega uses AI to solve a more fundamental problem – how to modernize legacy systems and create reliable, explainable automation.
The distinction between design-time reasoning and runtime predictability is clever. It gives organizations the benefits of AI creativity and adoption without the risks of agents going rogue in production. For regulated industries or any business that needs to explain its decisions, this approach makes sense.
What’s particularly interesting is how Pega turns the unpredictability of AI into an asset at design time while maintaining the predictability enterprises need at runtime. It’s a nuanced approach that suggests Pega senses an inherent distrust of running too soon with agentic AI. It gives buyers the opportunity to say ‘yes, we are using agentic AI’, whilst also minimizing any risk that they perceive to be involved. And as noted above, it also tackles a decades-long problem for buyers (legacy).
That being said, there is a risk here for Pega too. As consumers and the public get more familiar with prompt-based problem solving, this consumer-led approach may well push quickly into employee expectations (remember mobile and Software-as-a-Service (SaaS)?). Other organizations that adopt prompt-based platforms, using natural language to carry out tasks or solve problems – where they push creative control into the user’s hands, across their companies – may be adopting a riskier strategy, but if successful, could deliver new or interesting results that Pega customers don’t have access to. There’s a risk Pega falls behind with its ‘behind the scenes’ approach, but as ever, customers will be the ones that decide that.
