Generative AI tools - Pega’s Knowledge Buddy & its implications

Brian Sommer Profile picture for user brianssommer February 26, 2024
Summary:
Thoughtful generative AI applications are starting to become available. Pega’s new tools have a couple of pleasant surprises to note. Here’s a quick look at its Knowledge Buddy technology.



knowledge

Interestingly, few vendors offer up their own firm as a case study in their own technology. Makes you wonder why their tech is good enough for other firms but not their own?  

With the explosion of new generative (and other types of) AI this last year, you’d expect to hear from more firms using these new tools internally, but it’s been actually quiet on that front. Recently, I got a ping from process automation vendor Pega (nee Pegasystems) about its new generative AI tool Knowledge Buddy and the uptake it has gotten internally.  Pega’s products are used by a number of large and mid-sized organizations to “connect workflows and interactions across the customer lifecycle”

According to Pega, Knowledge Buddy is a:

generative AI-powered assistant that will quickly and easily enable customers and employees to get specific answers synthesized by generative AI from content scattered across enterprise knowledge bases.  

These tools use:

natural language processing and machine learning to understand the user’s query and provide relevant knowledge articles from the company’s knowledge base.

While that sounds like the description of many chatbots and other gen AI powered tools, Pega has deployed 13 different buddies within its own firm in areas like HR, sales, events, etc.  

It's different talking to a vendor about how they are using their own tech. Why? Well, vendors place different priorities on different functions within their firm than the average client of mine. For example, software development and support are two extremely important and expensive aspects of a software company while in many firms they are barely a top 10 focus area. 

Pega provided two use cases of their internal buddy technology: 

  • Pega Buddy helps its People Team manage the slew of employee inquiries about benefits, review cycles and more – especially during peak periods like open enrollment. By empowering more than a quarter of its staff with the tools to locate information quickly and easily, tickets are reduced by 25% and Pega Buddy has answered more than 6,000 questions.   
  • Support Buddy is a vital tool for Pega's support engineers, aiding in quickly retrieving pertinent information from case histories and product documents. Used by over 250 employees daily, it optimizes the triage process by directing requests to the appropriate support teams. Additionally, it serves as a moderation tool for client and partner inquiries, saving moderators up to 30% of their response time.

To put the above in perspective, Pega has around 6000 employees

Pega seems to understand the risks associated with newer AI/gen AI tools. When you look closely at one of these buddy tools, you’ll see how Pega is recommending that users double-check any advice/guidance from these tools with the reference materials it also presents to the users.  (see the yellow highlighted text in the attached graphic)

bs1

This disclaimer is a key best practice that all vendors should emulate. Why? The popular press is already full of stories of chatbots returning erroneous recommendations that people or companies did not check. In one example, a car dealership placed a shopping recommendation bot on its website. Unfortunately, it recommended potential buyers should buy vehicles from a different manufacturer than the one this dealership sold. In another situation, an airline chatbot provided bad bereavement fare advice that cost the flyer hundreds of dollars in higher fares. The PR blowback from this last example has been significant and more examples will likely hit the news in the days and months to come.  

And just today, there’s a piece detailing a number of bad behaviors from ChatGPT. The article summarized these thusly:

In recent hours, the Artificial Intelligence tool appears to be answering queries with long and nonsensical messages, talking Spanglish without prompting – as well as worrying users, by suggesting that it is in the room with them.

The importance of these tools

There are countless points in business processes where a constituent (e.g., employee, customer, supplier, regulator, alumni, etc.) has a question or needs some reference material. They usually want an answer immediately so they can complete the task at hand. Often these answers cannot be found in large public search engines (e.g., Google) as these don’t have access to the internal data, content, business practices, policies, etc. that are unique to each firm. Smart chat tools are often a great response to these needs. 

Businesses love these tools as they can improve worker productivity. According to a study by consultancy McKinsey:

Employees spend approximately 19% of working hours trying to track down information needed to complete tasks.

Great chat tools must be trained on a company’s proprietary data. After training, many of these tools are ready to aid in customer service, underwriting, claims processing, invoice processing, supplier payment inquiries and other functions. The tool, great source data and the training are all needed to deliver great outcomes.  

But these tools are not infallible. They are only as good as the training data and programming within them. 

Conversation with a Pega executive

I recently spoke with David Vidoni, VP of Information Technology at Pega, about these new capabilities. We discussed how these new tools:

  • provide an audit trail that points to the source materials where specific guidance was referenced
  • not only offer guidance but also determine what the next best action should be. That’s something a process automation vendor like Pega should understand well.
  • can incorporate and/or learn from document attachments, prior inquiries, texts and other sources
  • have a feedback mechanism built-in. Users are encouraged to make content changes before data is shared with the end-user. User and customer feedback can be summarized and reported for appropriate follow up activities.
  • Use Azure AI Services (see https://azure.microsoft.com/en-us/products/ai-services/ ). This ensures each installation has security protocols in place. Internal customer data remains within the firm. Pega wants to ensure that private, internal data (e.g., sales forecasts, customer purchases, etc.) is not leaked or available to others outside of the firm.
  • Have reduced ticket volumes by 65% in some situations

Software development

Pega is also using generative AI to create application software code. It is being married to Pega’s existing low/no code capabilities. While many AI tools can create code, the real announcement here has to do with a recent capability Pega is rolling out called Pega GenAI Blueprint.

The blueprint idea is a step change improvement over a simple code development request fed into an AI tool. Users are leveraging other customers’ development efforts to help make their application ideas become more fully-featured and contain more appropriate workflows, controls, etc. The blueprint tool informs the user of other capabilities they may want to include in their application based on the experience of other firms in their industry and/or application development activities. These more robust blueprints are what development teams can share with future system users, functional experts and others to round out the application design. This makes for a more complete and hopefully more valuable solution the first time. 

According to Pega:

Pega GenAI Blueprint enables teams to accelerate the journey from idea to implementation. Users briefly describe the app’s purpose, and the intuitive interface guides them by laying out the workflow elements and captures all feedback while facilitating collaboration. The Pega GenAI engine leverages Pega’s decades of industry knowledge to design optimized application blueprints, which incorporates best practice workflow processes, data models, and integrations. Once the design is finalized, the blueprint seamlessly feeds into the Pega Platform, transforming it into an enterprise-grade, cloud-architected workflow application that is optimized for their business.

The blueprint tool should speed development efforts, reduce the number of iterations required and lower development costs. It might also help developers create better solutions, too. 

My take

Pega’s annual user conference will be in a few months. I expect there will be considerably more content and discussion regarding advanced technologies (e.g., generative AI) at that time. These announcements re: buddies and blueprints are probably just a taste of what will be covered then. 

It will be interesting to hear what mainline application software vendors think about new tools like blueprint. This tool might help customers create their own industry vertical apps. Those applications could be built at great speed with little development and maintenance cost. This could also change the economics of the application software space significantly. That thought is really intriguing as many old-school application software firms feel that generative AI capabilities should help them raise prices when, in fact, price hikes might drive more software users to build their industry apps, cheaply and quickly, themselves. In other words, tools like the blueprint solution could trigger big dislocations in the traditional application software space.

Loading
A grey colored placeholder image