Main content

Pega CEO prioritizes practical AI use cases, says most enterprise AI agents are “garbage”

Derek du Preez Profile picture for user ddpreez June 10, 2024
Summary:
Workflow and case management vendor Pega is taking somewhat of a different approach to its competitors when it comes to AI - by using AI to get customers to its desired center-out architecture.

An image of Pega CEO Alan Trefler on stage in Vegas
(Image taken by diginomica)

Speaking with Pega CEO Alan Trefler is never dull. Where other technology leaders remain diplomatic or quiet, Trefler has a reputation for being outspoken and frank about how he sees technology trends playing out for enterprise buyers. And during Pega's annual user event in Las Vegas this week, he did not disappoint - giving a candid assessment of the current hype surrounding AI technologies. 

It’s been an interesting event so far. Whilst Pega’s competitors this conference season have argued the case for AI being the new UX, and have said that AI agents will soon be carrying out much of the heavy work of an enterprise, Trefler has taken a more pragmatic approach (albeit still using his entertaining language). Rather than showcasing a future AI state that perhaps feels intangible for many buyers that are likely still in the midst of other digital transformation projects, Pega has instead focused on AI use cases that aim to speed up the move to what the vendor sees as a preferred operational architecture for the enterprise - ‘a center-out’ approach. 

Last year Pega laid the foundation for this idea, positioning case management, workflows, processes and knowledge as a ‘situational layer cake’ that sits in between front-end and backend systems. This layer cake essentially is the ‘process fabric’ that holds an enterprise together - it’s the understanding of how work gets done - with technology systems and front-end channels supporting the core. 

Rather than showcasing an ‘art of the possible’ for how work could be carried out using AI, the AI announcements made this week instead help customers get to this center-out architecture more quickly - at which point, Trefler argues, more sophisticated AI adoption could be pursued. On the approach itself, Trefler said during his keynote: 

One of the things that I'm super excited about is that a vision we've talked about for the last five years, has really been taking hold. That vision is the idea of a center-out business architecture. Everyone talks about wanting to do a great job for their customers - they make promises, they envision, they spend money, they do all sorts of things - but when they begin to apply the technology, they apply it through channels. 

And when they apply it through channels, inevitably, they end up putting business rules, and business processes in the channel. And all of a sudden, they’re doomed - this is the birth of the legacy system. They end up having things that maybe work great in your Salesforce system, but when they want to have it work on the website, they have to change it there too. And then the mobile app. And then the contact center, or the back office. 

Trefler argues that this approach has historically led to silos for customers and a lack of ‘true channel independence’. Equally, trying to fix things on the backend, pushing work and data towards the transactional systems of record, has resulted in complexity and an inability to change. Or as Trefler put it: 

They'll take all that complexity that makes businesses hard to change, and difficult to even understand, thinking they can encapsulate that in the back end - it's not encapsulated. It's vomited all over everywhere. 

The reality is you can't put this in the front. And it doesn't work putting it in the backend. Where do they belong? By the process of elimination, you’ve got to put it in the center. You have to have engines in the center that are capable of making AI driven decisions, that are capable of then initiating workflows, and having those workflows use decisions, and all of it requires record keeping - what we call case management. We keep track of what you did and what you're doing, so you can do machine learning on it. So that you can make yourself better. 

Pega CTO and VP of Product Marketing, Don Scheurman, reiterated this point during a briefing with media and analysts. He said that whilst many vendors in the market at the moment are focused on introducing generative AI tools that improve productivity for users (in areas such as content creation or case summarization), he doesn’t see this as bolstering buyers’ competitive advantage in the long term: 

The important thing with productivity is that I think there will be a certain point, I believe, that some of this generative AI stuff is going to be like Microsoft Excel - we all have it, we are all more productive from it, but me having Microsoft Excel doesn’t give my corporation a competitive edge. 

But organizations that understand the need to move to this center-out approach will be able to take greater advantage of sophisticated AI down the line: 

We think that organizations need to organize themselves around the work itself - a center-out architecture. What we see is two patterns in organizations. 

One is that the organization organizes itself around the legacy systems and the data. The other thing we see is organization around the channels - so, you have a mobile team, a web team, a chat bot team. 

And what that basically led to was the customer gets on the mobile app, starts something, and then calls the contact center, and the contact center says they don’t know what you’re doing. It’s a horrible customer experience. 

But if you organize around the work, and you have a team that’s going to own, for example, the credit card dispute process - and owns that experience end-to-end - both pushing it into the backend transactional systems, whilst also making sure that process works across the various front end experiences…that’s where the power comes. 

Making this easier

If we consider Pega’s pitch a good one (I do think it is logical, but inevitably this assessment will vary between buyers and their priorities), we need to know how to progress towards this center-out architecture. And this is where Pega’s AI announcements and product updates largely focused on this week. During his keynote this morning, Trefler acknowledged that Pega hasn’t always made this easy, but added that this is where AI could be useful:

We’ve heard that as people have wanted to use Pega, there are things that were difficult. There were problems for our customers. We’ve heard that Pega can be hard to learn, hard to use, the resources can be scarce and expensive. A year ago I said ‘we’ve heard this’. With this new introduction of generative AI, we think there’s a way to change this - that was completely inconceivable before. 

The big push this week has come in the form of Pega’s GenAI Blueprint product. Although not entirely new, there were a number of updates to Blueprint that were announced today. Essentially, Blueprint is a ‘app design-as-a-service’ tool that lets organizations reimagine their mission critical workflows. 

For example, during a demonstration we saw how a full workflow could be designed in a matter of minutes, based on best practice in various industries/use cases. For instance, a financial institution could quickly get a ‘best case scenario’ workflow for processing insurance claims or handling credit card chargebacks, or a healthcare organization could quickly design a workflow for managing patient care. The idea is that not only does Blueprint reduce the time spent with design teams in harshly lit rooms spitballing ideas about what they think is the best design for a workflow, instead the tool bases it on evidence of what works best, whilst also speeding up the time to development of an actual application. 

The hope is of course that this application will then be built and used on Pega, but equally uses generative AI to speed up getting customers to this center-out architecture. Some of the key announcements for Blueprint today include: 

  • Legacy transformation accelerators - Organizations will soon be able to import existing assets, content, and knowledge. By the end of June, users will be able to import legacy business process modeling notion (BPMN) models, and by end of Q3, Blueprint will also support additional legacy inputs like process documentation, app screens, and full blueprint designs via APIs.
  • Live application previews - By the end of June users can instantly preview what their user interface will look like across any channel - including web, mobile, customer service desktop, and customer self-service via Pega’s DX API – at any stage of the design cycle with the ‘preview my app’ button.
  • Data model generation - The tool now automatically generates data model fields behind each case type, with the ability to edit, manually add, and delete existing fields.
  • Generate more ideas - By clicking the ‘generate more’ button anywhere in the process, users can now ask the tool to create more asset ideas – such as case types, fields, or personas.

In addition to Blueprint, which essentially uses generative AI to speed up how organizations achieve a center-out architecture, which at that point will allow for more AI applications, Pega today also launched GenAI Socrates to speed up the learning of Pega skills. 

Socrates is essentially a generative AI tutor that dynamically adapts to users to teach core Pega skills to customers and partners. Adopting the Socratic teaching method, GenAI Socrates educates students through a two-way conversation interspersed with open-ended questions on each learning objective (much like a teacher in a classroom). If Blueprint is aimed at speeding up the ideation and development of a center-out architecture, Socrates is aimed at speeding up the learning and skills adoption required to maintain it. 

Pega stays practical

The generative AI announcements outlined above are vastly different from some of the others we’ve seen coming to market in recent months from some of Pega’s competitors. It’s been interesting to note that whilst vendors in a similar field to Pega are promoting how AI can revolutionize the way we work, with conversational agents becoming a one-stop-shop for carrying out a variety of enterprise tasks, Pega is taking a more foundational view of initially using AI to guide organizations towards a better technology architecture. It’s at that point it sees the application of more sophisticated AI being used.

During a media Q&A with Trefler, diginomica asked the CEO why Pega was not being as ambitious with its messaging as others in the market. And the response was quite clear: 

I think what most of the others are talking about with agents is just garbage. This fantasy of AGI, artificial general intelligence, where agents are going to suddenly mysteriously do things…we operate in regulated businesses, we operate with companies that want to explain how the work they do is automated. Being able to use AI to generate tangible, visible and very precise workflows, which can, by the way, be fully automated, and are as automated as anything these magic agents are doing...I just think we're on a whole different plane. 

It’s much more practical to tell you the truth. We've all seen all the ‘magic’ before. This isn't magic. If you try it, I think you'll understand how easy it is to create all the variations of the workflow you need to truly automate everything, without claiming some agent is going to figure it out.

My take

I must admit, during the keynote this morning I was perhaps a bit surprised with the tempered messaging coming from Pega. But then I reminded myself that at previous conferences this year I’d been frustrated with how vendors were selling a vision that was so far from reality for many enterprises. A lot of buyers are still in the weeds with establishing and maturing their cloud architectures - amongst other data and distributed work challenges. Many, many organizations are not yet ready to think about AI agents that can carry out work autonomously on behalf of users. 

There is plenty of work ‘in the middle’ that needs to be done to take advantage of AI effectively. Implementing ‘meta’ platforms that understand how work is carried out, sitting in between backend and frontend systems is something I’ve been suggesting is crucial for some time. The ‘knowledge’ of an enterprise and the work it carries out is going to be essential for establishing effective AI use going forward. Pega is trying to give its customers the tools to get there more quickly. 

Now, of course, this is somewhat of a self-fulfilling agenda for Pega. It needs its customers to get to this center-out architecture for it to further deliver AI success for buyers. However, having spoken to a few customers at the event today, they’re clear about how beneficial Blueprint and Socrates will be for them in the interim. It may not be the sexiest pitch, but it does feel like a realistic and practical one for many buyers at this moment in time. 

Loading
A grey colored placeholder image