By now, you’re probably familiar with the various large language model (LLM) chatbots available online. Maybe you lean on one to help with a research project, double-check an important email, or pull key themes out of a long set of notes. But if you’re at a complex organization, the limits of publicly facing LLMs quickly become apparent:
- They don’t contain corporate knowledge, so they’re less helpful on work-specific questions.
- They’re opaque, and you can’t be certain that sensitive information won’t be exposed elsewhere.
- They have a tendency to make up answers and hallucinate.
As a result, a new generation of tools has emerged that takes the best parts of an LLM – like that conversational ease that sometimes feels like magic – and injects some practicality into it in the form of Retrieval Augmented Generation (RAG). This lets you point the LLM at a more specific and authoritative knowledge set, like the thousands of PDFs, product guides, process documentation, HR handbooks, and more that live in folders across your enterprise ecosystem.
Let’s take a closer look at RAG, AI-powered enterprise knowledge, and how Pega Knowledge Buddy can bring straightforward answers to your customers, sellers, service reps, and everyone else throughout your organization.
A smarter way to surface knowledge
RAG is an important development in how organizations can leverage generative AI to deliver on-the-second information to their customers and employees. Instead of relying on an LLM’s training data to generate responses, tools equipped with RAG only source answers from the documents and data you feed into it.
First it uses a process of semantic search and document chunking to recognize the relevant content and then directs the LLM to formulate an answer from that content. That means more accurate, more contextual answers, and nearly zero risk of hallucination.
The importance of good prompts
One element of RAG and generative AI more broadly that’s critical to delivering fast, seamless experiences for end users is prompt engineering. When you stand up an AI tool like Pega Knowledge Buddy, you’ll have the opportunity to define how it interacts with its end users: Is it a cheerful customer chatbot? Is it a dry, informational helper for IT requests? How should it respond to questions it doesn’t know the answer to? How creative or basic should it be with its responses?
These questions are just a handful of what you might consider at the setup. And by getting granular with how a tool like Knowledge Buddy will act, it opens up the possibility to deliver segmentized Buddies that are built for purpose for different areas.
Beyond prompting, other settings also let you dial in the bot’s behavior. You can home in on the appropriate document chunking size, which defines elements of how RAG retrieves information from its sources. You can also set parameters around the content you want the tool to reference, including some and walling off others to ensure you don’t surface sensitive information to the wrong audience.
What sets Knowledge Buddy apart?
We’ve discussed RAG tools as a broad category. But Knowledge Buddy brings other elements to the table.
For starters, as a Pega product, Knowledge Buddy is a low-code tool. It’s easy to stand up Buddy initially and it only takes minutes to deploy additional Buddies for the use cases that are most important to your organization: self-service, sales, onboarding, compliance, RFP response, anything you can think of.
At PegaWorld iNspire 2024, Steven Park of aaseya, a Pega Partner, detailed how his company uses two Buddies to facilitate their compliance and auditing platform.
One is an Inspection Buddy, used to help on-site inspectors make recommendations about how to resolve non-compliance results during a live audit. The other is a Configuration Buddy, which aaseya uses to ingest all the content for audits that can vary greatly by region or industry. One thing that came through clearly at the end of the session was just how similar these two Buddies looked in the Pega admin panel. Using plain language and very little code, it’s possible to create two distinct tools that output results in different formats with different audiences, with just a few small changes between them.
“It took just a couple days to go through the setup process and understand the mechanics of how Knowledge Buddy works,” Steven confirmed. When asked about the option of building their own RAG versus licensing Knowledge Buddy, he responded: “With all the functionality that’s built into Knowledge Buddy already, there’s no reason to go build it yourself.”
Knowledge Buddy comes equipped with RAG straight out of the box, alongside the ability to deploy to your website through chat. It integrates seamlessly with the Pega platform of course, but also with external systems – so long as the application can make an API call, it can interface with Knowledge Buddy.
But the true value of Knowledge Buddy shines through in concert with a set of products that guide end users through journeys and workflows. It’s when Knowledge Buddy lowers overall case or inquiry volume by offering better answers at first point of contact; or delivers immediate answers to a seller or customer (or compliance officer) during the course of their work; or accelerates onboarding and upskilling so that employees can do better work, faster – those are the scenarios where it becomes apparent that this is more than just a fancy chatbot.
An encyclopedic tutor for your customers and employees
Knowledge Buddy also made an appearance at Kerim Agkonul’s keynote on the Age of Transformation.
“With so much documentation and knowledge scattered all over the place,” Kerim asked, “What would it take to be good at your job? What would it take to be great?”
He told the story about Alexander the Great, who had a famous personal tutor to answer all of his questions: Aristotle. But the rest of us don’t have access to legendary Greek philosophers who can guide us in our day-to-day lives and jobs. And frankly, even Aristotle might not be that helpful when you need to find the latest product documentation, or pass along a policy PDF to a customer waiting on the other end of the phone.
Instead, we have Knowledge Buddy to answer your customers' questions on the first try, and make every employee your best employee.
Be sure to watch the full Knowledge Buddy session, along with Kerim’s keynote, from PegaWorld iNspire 2024. And while you’re at it, check out all the other keynotes and breakout sessions you may have missed.