AI and BPM: ‘The Best of Both Worlds’ - Alan Trefler on the Future of Agentic Workflows and Enterprise Efficiency”

As we like to say on The CRM Playaz, it's an event unlike any other.  When it gets close to PegaWorld, Pega's annual user conference, company founder and CEO Alan Trefler joins us for a sneek peek. It's also a great time to pick his brain and get his take on some of the important topics facing the industry.

The below clip is a taste of our full-blown conversation with Trefler, including an edited transcript. Some of the highlights include the following:

• How Pega's core principles have evolved from 2006 to the present AI-driven era.
• Rather than replacing business process management (BPM) with AI, Pegasystems integrates the two.
• AI enhances reasoning and opportunity identification, while BPM ensures the reliability and consistency of execution.
• How two foundational philosophies core to Pega  (layer-cake architecture built from the center-out) are a perfect fit in the Agentic Era.
• The clear distinction between how workflows and large language models (LLMs) consume power within enterprise systems. 
• Pega's shift away from user-based pricing to a model based on the number of work items processed and how that removes the increased usage penalty with which some vendors are struggling when it comes to AI.

Also included below is an edited transcript from the clip, as well as a summary of our conversation from Perplexity. You can also check out the full conversation at https://www.youtube.com/live/PB8zEj9laro?si=5U8dKoOGK4c_ofnh.

Edited Transcript

Brent Leary: I found a pretty interesting clip of Alan giving his  closing statement at Pegaworld 2006:

Alan Trefler from 2006: BPM can fast-forward an organization. It can make an organization far more effective, far more able to get done what it needs to get done in ways that will differentiate, in ways that will kick the butt of the competition, and in ways that will save money. And that's a wonderful and rare opportunity.

Brent Leary: So I want to go back to 2006 [Alan]. In that clip I played, you were giving your closing argument for the 2006 PegaWorld. First of all, it's hard to believe it was 19 years ago. I've mentioned that everything you said, it seems like it could apply today, with the exception of replacing BPM with AI.

Alan Trefler: I'm going to contest that. I know you hypothesize that you could just take the term BPM and replace it with the buzzword du jour, which would have to be AI agents, and say the exact same thing. But we've been able to  take the AI and make it AI AND BPM, right, as opposed to flipping one out for the other.

I think that's super, super-critical because the AI helps you reason and identify opportunities to replace a legacy system with something that's going to be a lot better. And the AI gives you that intellectual thought process that's really very, very invigorating; that you rethink what you're doing. But the BPM which the AI creates then gives you the reliability of execution that businesses need.

Leary: Two things that I always think about when I think about Pega. Layer-cake and center-out. How does, layer-cake and center-out manifest itself in the agentic era?

Trefler: I think it's perfect for the agentic era. And one of the things we're super-excited about is that every workflow in a Pega system is inherently agentic.

And what I mean by that, and this is the center-out concept, the workflows run at the center of your business. The workflows and the decisions. Those really are what your business offers to your clients. The AI has made it so much easier to create all the variations you need of those, and you're able to curate them and check them out.

But those workflows at the heart of your business, every single one of them is agentic, which means I can actually have a conversation with my application. And Blueprint actually changes literally every one to two weeks. And some of the advances in it recently have included the ability to have a agentic conversation with an app that you just created in Blueprint so you create the app and you can begin talking to it.

And what's amazing about it is it uses the power of the language models to be able to do a brilliant job conversing. But then it uses the predictability of the workflows to make sure that the things you go through are going to be consistent from client to client, are going to be provable, are going to be, well, fair. All the things you really need to be able to run an enterprise center-out architecture is a perfect architecture for agents to be able to operate, because the agents use the power of the center to do their work.

And of course, the layer-cake. That's, where we put all of our assets so that when we're designing these workflows, to say, hey, here's some stuff that goes across the entire organization as opposed to things that might apply to just one division or another division or one region or another region. The ability to layer those elements together provides reuse and provides visibility and stability to what's going on.

Leary: I think there has been some industry-wide struggle with how to price this stuff. They're grappling with the cost of power consumption and trying to get all the infrastructure set. And being able to figure that out before they feel comfortable with how much it's going to cost these folks [customers]. How do you guys look at that aspect of this?

Trefler: We don't have the problem. Most of our competitors do because about five years ago we stopped pricing by user. And the reason we did that is we found that our systems were massively reducing the numbers of users. So we instead began pricing based on the work; What are the number of work items done in a month or a year? At a price based on quantity. So we don't have a pricing structure problem.

The way that we do this, where we do the design using the reasoning and the power of the LLM, when Blueprint is running, that's chewing through tokens, right? That's just really doing logic, thinking of what the best way to do stuff is, putting it all together. But at runtime, which is, you know, a factor of a thousand more usage than design time, we're running workflows. We're using the LLM just for language interpretation. But the efficiency of actually the runtime environment...

Brent, that was a good question. People don't appreciate how non-green some of this stuff is here. We love that the workflows are literally 100 times more efficient than running the LLMs. And since they've been authored by the LLMs, you get the best of both worlds.

Perplexity's take on the conversation

The transcript features a conversation between Brent Leary and Alan Trefler, founder and CEO of Pegasystems, reflecting on the evolution of PegaWorld and the company's technology vision from 2006 to 2025. The discussion centers on how core concepts and architectures at Pega have adapted to the rise of AI and agentic systems.

**Key Points Discussed**

**1. Enduring Principles: BPM and AI Integration**

- Brent Leary notes that Alan Trefler’s closing remarks from PegaWorld 2006 remain relevant today, except for the shift from BPM (business process management) to AI.

- Trefler contests the idea that AI simply replaces BPM, emphasizing instead that Pega integrates both: AI augments BPM, providing reasoning and opportunity identification, while BPM ensures reliable execution. This combination is crucial for modern enterprises[1].

**2. Pega’s Architectural Concepts in the Agentic Era**

- The "layer-cake" and "center-out" approaches are highlighted as foundational to Pega’s architecture.

- Trefler explains that in the agentic era, every workflow in Pega is inherently agentic, meaning applications can be interacted with conversationally, leveraging language models for dynamic, natural interactions.

- The "center-out" model places workflows and decisions at the heart of business operations, enabling consistent, fair, and provable processes across clients.

- The "layer-cake" approach allows for organizational assets to be reused and layered, providing stability and visibility across divisions and regions[1].

**3. AI, Blueprint, and Workflow Efficiency**

- Pega’s Blueprint tool now enables agentic conversations to design and interact with applications, using language models for conversation and workflows for predictable execution.

- Trefler notes that advances allow for rapid iteration, with Blueprints changing every one to two weeks, and the combination of AI reasoning with BPM execution provides both innovation and reliability[1].

**4. Pricing and Efficiency in the Age of AI**

- The industry struggles with pricing AI-powered systems, especially given infrastructure and energy costs.

- Pega moved away from user-based pricing five years ago, now charging based on work items processed, which aligns better with their automation and efficiency gains.

- Trefler highlights that while AI models (LLMs) are resource-intensive during design time, Pega's runtime workflows are far more efficient, up to 100 times less resource-intensive than running LLMs continuously. This results in significant cost and environmental benefits, as workflows authored by LLMs can be executed efficiently at scale[1].

## Conclusion

Alan Trefler emphasizes that Pega's enduring architectural principles (layer-cake and center-out) are well-suited for the agentic, AI-driven era. By integrating AI with BPM rather than replacing it, Pega delivers both innovative reasoning and reliable execution. Their pricing model, based on work rather than users, and focus on workflow efficiency position them well for the future of enterprise automation.

- Alan Trefler emphasizes that at Pegasystems, AI is not simply replacing BPM (business process management); instead, the two are integrated. AI enhances the reasoning and opportunity-identification aspects, while BPM ensures reliable execution for businesses[1].

**2. Pega’s Architectural Principles: Layer-Cake and Center-Out**

- The "center-out" concept means workflows and decisions are at the heart of business operations, enabling consistent and provable processes for clients.

- The "layer-cake" approach allows for organizational assets to be reused and layered, providing visibility, stability, and the ability to apply solutions across different divisions or regions[1].

**3. Agentic Workflows and Blueprint Tool**

- Every workflow in Pega is described as "agentic," meaning users can interact conversationally with applications.

- The Blueprint tool leverages language models for natural conversation during app creation and uses predictable workflows for consistent execution, with frequent updates and improvements[1].

**4. Pricing Model and Efficiency**

- Pega moved away from user-based pricing about five years ago, now charging based on the number of work items processed, which aligns better with automation and efficiency gains.

- The design phase uses LLMs (large language models) for reasoning, but the runtime environment is far more efficient: workflows are about 100 times more efficient than running LLMs continuously, which also addresses concerns about energy consumption and cost[1].

**5. Industry Challenges with AI Cost and Sustainability**

- The industry struggles with pricing AI-powered systems due to infrastructure and energy demands.

- Pega's approach, with efficient runtime workflows authored by LLMs, offers a solution that is both cost-effective and more environmentally sustainable[1].

Sources

[1] Alan-Trefler-Road-to-PegaWorld-25.docx https://ppl-ai-file-upload.s3.amazonaws.com/web/direct-files/attachments/29397469/fab9353f-f32a-4b80-aade-eef4cc70563f/Alan-Trefler-Road-to-PegaWorld-25.docx

## Key Themes Covered in the Transcript

**1. Integration of AI and BPM (Business Process Management)**

- The conversation emphasizes that Pegasystems does not see AI as a replacement for BPM but as a complement. AI is used to enhance reasoning and opportunity identification, while BPM ensures reliable execution in business processes[1].

**2. Agentic Workflows and Conversational Interfaces**

- Pega’s architecture is described as inherently "agentic," meaning workflows can be interacted with conversationally. This leverages language models for natural, dynamic interactions, while maintaining the predictability and consistency required for enterprise operations[1].

**3. Architectural Principles: Center-Out and Layer-Cake**

- The "center-out" approach puts workflows and decisions at the core of business operations, enabling consistency and fairness across clients.

- The "layer-cake" model allows for organizational assets to be reused and layered, providing stability, visibility, and adaptability across divisions and regions[1].

**4. Innovation in Application Design with Blueprint**

- Pega's Blueprint tool uses agentic conversations and language models to design and interact with applications, enabling rapid iteration and development. This supports frequent updates and reflects the shift toward more dynamic, AI-driven application creation[1].

**5. Pricing Model and Efficiency**

- Pega has shifted away from user-based pricing to a model based on the volume of work items processed. This aligns better with the efficiencies gained from automation and reflects the reduced need for manual users as systems become more autonomous[1].

**6. AI Cost, Efficiency, and Sustainability**

- The transcript highlights industry challenges with the cost and energy consumption of AI systems. Pega addresses this by using resource-intensive LLMs (large language models) primarily during the design phase, while runtime workflows are much more efficient: up to 100 times less resource-intensive than running LLMs continuously. This offers both economic and environmental benefits[1].

These themes collectively illustrate how Pegasystems is adapting its foundational principles to the new era of AI and agentic technology, focusing on integration, efficiency, and sustainable innovation.