Revisiting Pega in 2026: GenAI, Constellation, and Blueprint

A journal for 2026

I’ve been building Pega applications since 2010. It has been a long journey.

Over the past few weeks, I spent some time catching up through Pega's CSA, CSSA and CLSA courses. The experience was not tedious, and it was not deeply technical either. It was more of a reset: a way to understand what has changed, what has stayed the same, and where Pega seems to be heading now.

What made me seriously take a second look was Pega Launchpad.

Launchpad feels like an intentional move by Pega to create room for building applications for clients that are not quite in the Fortune 500 tier Pega has traditionally been known for. For years, Pega’s reputation has been tied to large enterprises, complex operations, and equally large price tags. Launchpad looks like an attempt to open another lane: one where partners and builders can deliver workflow-driven applications to mid-sized organizations with a more productized entry point.

At least, that is how it looks from the outside.

What still feels unclear is the pricing. Pega talks about charging by “outcomes,” and that raises more questions than it answers. What exactly counts as an outcome? A completed case? A business event? A customer transaction? It is the kind of term that sounds modern and attractive in a sales deck, but vague in practice. I filled out their contact sales form hoping for clarity. So far, nothing. Just silence.

Even so, the idea itself is interesting. It reminds me of how Pega has repeatedly tried to accelerate the early stages of application building.

Back in the 5.5 days, the Application Wizard helped lay down the application’s foundation: class structure, operators, rulesets, and the basic scaffolding. Over time, that idea evolved. From Pega Express to Blueprint, Pega kept pushing farther upstream, trying to define more of the application earlier and with more automation.

Today, Blueprint does much more than create a shell. It tries to infer flows, data structures, personas, case types, and user experiences from prompts and business requirements. That is a meaningful expansion compared to the old days. It is encouraging to see, even if it is not perfect.

Like most AI-assisted generation tools, Blueprint appears to generate what it thinks best fits the requirements. How reliable that is remains something of a black box. It can also update flows through chat, but from what I have seen, that capability still feels limited. Helpful, yes. Fully trustworthy without review, no.

It also feels like Pega wants to keep more of its accumulated implementation “wisdom” inside its own ecosystem. Blueprint is not just a helper feature. It increasingly looks like a strategic front door to the platform.

On the UI side, Constellation is clearly the present and future. “Legacy” UI is now exactly that, and Cosmos sits in the middle as a kind of transitional generation. A friend joked that we are dinosaurs now if we still think in the older Pega UI paradigms. There is probably some truth to that.

My opinion here may be polarizing, but I think it is fair.

Pega’s UI has never really been its strongest point. It has always been more about process, orchestration, and enterprise behavior than interface elegance. Cosmos felt like a familiar upgrade. It expanded UI features, events, and controls in a way that still felt natural to experienced Pega developers.

Constellation is different. It comes from a more prescriptive, data-model-first philosophy. In that sense, it is a more mature and opinionated framework. I can see why Pega went in that direction. But coming from Cosmos, it can also feel constrained. Sometimes it feels less like a step forward in flexibility and more like a trade-off: consistency and speed in exchange for freedom and control.

Security also seems more developed now. The two additions that stood out to me were CBAC and BAC.

• CBAC (Client-Based Access Control) appears to be aimed at privacy-oriented handling of personal data, helping address regulations and compliance requirements that used to require a lot of custom implementation.
• BAC (Basic Access Control) is focused on protecting against unauthorized requests coming from the UI layer, including sections, controls, and harnesses. In plain terms, it is there to help prevent broken access control issues where users try to bypass the intended experience through crafted requests or URLs.

That part stood out to me because it addresses real pain that teams used to solve the hard way.

Reporting also caught my attention. Insights is the first time I have seen Pega present reporting in a way that feels more intentionally business-facing, rather than just another technical reporting capability. I have not gone deep enough to judge it fully, but it stood out.

On the server side, another shift seems sensible: capabilities like search and stream processing are increasingly treated as external infrastructure concerns rather than things Pega should fully own. That makes architectural sense. If those capabilities depend on technologies like Elasticsearch or Kafka, it is cleaner to let specialized infrastructure do that work rather than hide it under the platform.

I was also hopeful about Pega’s AI direction.

From what I have seen, Pega now has more explicit concepts around GenAI, including Connect Generative AI, Agents, and Tools. That signals a platform trying to formalize how LLM-driven behavior fits into case processing and automation. Conceptually, that is promising.

But I still have to complain about one thing: the lack of detailed developer how-to guidance.

This is where Pega still frustrates me.

There is enough material to show that the features exist. There are pages, overviews, learning modules, and product messaging. But when it comes to practical implementation details, especially for scenarios slightly outside the happy path, the guidance still feels thin.

A good example is external LLM integration. I was trying to understand how to connect an LLM provider outside the standard, Pega-supported list. Something like Grok, for example, instead of the usual major providers. I found hints and community discussion, but not the kind of direct, grounded documentation that gives confidence. For a platform that increasingly wants to talk about AI, that gap matters.

So overall, yes, things have changed.

Pega has moved forward in meaningful ways. Launchpad is an interesting strategic move. Blueprint is far more ambitious than the setup tools Pega used to offer. Constellation is now the center of gravity whether older developers like it or not. Security is more layered and more explicit. The AI direction is clearer than before.

At the same time, some familiar frustrations remain: vague pricing language, patchy implementation guidance, and the sense that some important details are still harder to find than they should be.

Maybe that is the most honest conclusion.

Pega still feels like an old friend. Familiar, capable, opinionated, useful, and occasionally maddening.

I guess I will check again sometime in the future.

I’ll see you around, Pega.