December 20, 2023 – Duration 0:17:56

S4 Ep. 6 Take an outside-in perspective on knowledge management

Focusing on knowledge management simply as a way to house and share content misses an opportunity to use it as a more strategic element of the customer and employee experience. When knowledge content is fresh, rich and engaging — created with the customer and employee in mind — it can be a powerful competitive asset. In this episode of Tech Talks in 20, Harshali Desai, Principal Product Manager, Knowledge, at Genesys, explores how to think about knowledge management from the outside in, balancing the experiences organizations want to deliver through knowledge with the experiences customers and employees expect. She also discusses how reimagining knowledge management can streamline journeys and improve business outcomes.


Listen and Share

Listen on Apple Podcasts
Listen on Google Play Music
Listen on Spotify
Listen on Stitcher
Harshali Desai

Harshali Desai

Principal Product Manager

Harshali Desai is a Principal Product Manager at Genesys, focusing on using the power of AI and Knowledge to help businesses deliver smart and empathetic experiences. Her background in business and functional consulting helps her solve the high level business problems of today with focused technologies of the future.

Conversation Highlights

Here are conversation highlights from this episode, edited and condensed. Go to the timestamps in the recording for the full comments.

Welcome to Tech Talks in 20, Harshali. Tell us about what you do for Genesys.

Harshali Desai (01:14): 

I’m a principal product manager in the conversational AI team and I focus on knowledge. What that means is anything to do with creating knowledge, getting that rendered, having it optimized.

One definition of knowledge management is that it's all about managing content in a way that creates a single source of truth. What else should CX practitioners think about when they define what KM should be for their organization?

Harshali Desai (01:53): 

That’s a good definition to start with. CX practitioners today need to think about what knowledge is for them. Most organizations today have their knowledge in different systems across different departments in various forms. For example, the sales department can have information on the website about all the electronic gadgets they sell, whereas the support department might have information for troubleshooting saved inside a content management system.

I call this their knowledge landscape and some questions to ask oneself as a CX practitioner are: Where is all my information that I need? Is it in one place or is it in different places? Do I have the resources, the time and the need to bring it all together in one place? And if yes, what is that one place? There can be situations where there is a strong need to keep the knowledge where it is. If so, what should I expect from a knowledge-surfacing solution?

What else would a knowledge management solution need to do outside of just being a repository for all this information?

Harshali Desai (03:07): 

It is more than just housing content. I would say that when someone says knowledge management, it means — or should mean — a trifecta of, first, content management and strategy, the thing we just spoke about. Then second, search and rendering. Now that your content is out there, how to make sure it is searched effectively and that it is rendered or presented in the right way. And then third, take all that content, all the search experiences, and consistently make them better by filling gaps.

Knowledge management about answering customer questions and providing employees with insight, but often it's created from the perspective of the experience the organization wants to deliver. How can knowledge leaders flip that to be even more customer and employee eccentric?

Harshali Desai (04:14): 

It is important to think about what content an organization has and how to present it. Although I do feel that there is tremendous value in looking at knowledge from an outside-in perspective.

The way I look at it is knowledge is a medium or a way to deliver answers to questions that customers ask, and this medium needs to adapt to the needs of the person asking that question. For example, the expectation on the answer to “I want to buy a new coffee machine,” is that there are pictures and videos of coffee machines on a company website.

But now when you think about this for a bot, you are looking for an interactive experience. Maybe the bot can ask a few follow-up questions, “What kind of coffee do you like? Where do you reside? What’s your budget?”

And that same expectation is entirely different when this question is posed to an agent because you’re expecting that agent to be an expert not just on the coffee machine, but on coffee grinds and a whole lot of everything offered in the world of coffee.

When we start thinking about it from the point of view of person asking the question, we’ll realize that a knowledge solution does not just need to be adapting to the needs of the customer. We’re also working with other components in the conversational AI stack. So, if someone wants to buy a coffee machine, it’ll be great to know if they already bought one or were they looking for a specific machine in the knowledge portal or are completely new and maybe better off with an interactive bot experience or a human in the loop. It’s just flipping that conversation a little bit differently.

What are some of the use cases for more customer- and employee-centric knowledge?

Harshali Desai (06:15): 

If we think about the different kinds of use cases we could have for knowledge, we might think about if it is self-service versus an employee experience or an internal versus an external experience.

If we think about self-service — and I’m going to expand it a little bit to the examples that we just spoke about — knowledge can cover a broad breadth of questions from customers and can be the first line of information to customers. With the right tools and AI, it is relatively simple to cover a large breadth of questions for a lot of topics. For example, a question of buying a coffee machine can lead to another question on which grind to use or what coffee machine attachment we need to put in and so on.

And knowledge is great at covering this first level of broad breadth questions.

Then to make them even more relevant, they can be personalized and optimized for a wide range of touchpoint for bots, website portals, knowledge apps, as well as for different channels like web chat or a social app like Instagram. All these can then be thought of from a voice as well as a chat perspective.

Now, with an employee experience, an agent can benefit from knowledge surfacing, and this can be manual while they’re talking. They can type in those questions and get more information, or AI can either listen in or read through the voice and chat conversation and provide answers as the customers ask them. And all these use cases can be internal, where employees are the ones who are looking for the answer.

The strength of a knowledge solution when we look at all these different use cases is to be able to provide personalized experiences while adapting to these touchpoints and making sure that knowledge experiences are consistent and accurate across channels.

Talk about how to use tools like generative AI and large language models in addressing knowledge gaps or improving the knowledge experience.

Harshali Desai (08:51): 

When customers ask us about LLMs and generative AI or AI in general, my first question to them is, “What are the problems you’re trying to solve?” If the problem is they want to find the right answer to questions quickly, LLMs and generative AI can be trained and prompted to help find the exact answer, if it exists, from a large document. For example, a question on what grind to use for a French press does not need a large article on coffees and grinds, just the answer: You need medium course for a French press.

If the problem is solving for time to value, the fact that there’s so much content in there, I want to bring it to the customers fast without working too much on it. In those situations, generative AI can help knowledge authors by creating first drafts of knowledge answers that the knowledge authors can then edit and verify. And if the problem is, let’s say, to deliver or look for knowledge gaps and fill them, then AI and analytics can play an important role in highlighting those major areas where knowledge is missing and can even provide suggestions on how to fix the knowledge gap. AI is important, but what’s also important is to really know what is it that we are trying to solve through the use case and use AI to find those solutions.

Are organizations making it easy enough to access information?

Harshali Desai (10:30): 

We live in a world where there is so much information everywhere. Accessing that knowledge then means matching the question accurately with the right answer for that person. And this is where I think the search experience comes in. I spoke about how important it is to find the right answer and making the experience optimized for the touchpoint. What is also important is two more things. First, knowing when we don’t have the right answer and acknowledging it and presenting a next step in this situation — knowing we tried our best to find that answer, and this is probably not the best place for the customer to get that information.

And then the second is maintain that experience of the customer as we move them on to the next channel or the next touchpoint so those questions can be accessed and answered. For example, a customer can search for a knowledge on a website and a good search experience is when you know that now you’ve reached the limit of all the knowledge you can provide there and it is time to now pass it on to, say, a bot for a more interactive experience — more than, “What are the different types of coffee machines,” but also, “What were you looking for”?

And if that question is specific and nuanced, you pass this customer to an agent where, again, bots and knowledge are helping the agent pick up where the customer left off and get that question answered in a deeper way that only a human in the loop can do. The goal is to get the customer an answer to their question in the smoothest way to get it answered.

We should build that knowledge and a conversational AI ecosystem around this larger goal.

How should organizations approach knowledge optimization to keep their content fresh and keep it growing where there's information they could add?

Harshali Desai (13:07): 

For me, keeping content fresh means you keep it relevant and personalized. It’s important to connect the consumers of knowledge with the creators or managers of knowledge. And this in my opinion, is knowledge optimization.

Analytics and AI, if used right, can let you know what knowledge answers worked well and the answers that need work. Where is it that you need work? Is it where the agent provided an answer that there was a problem or was it how the bot went through that conversation? Was it during self-service or an agent facing experience? AI can let knowledge managers know which are the most talked about topics, which of them were answered or unanswered, and provide capabilities to create those answers.

Another way to keep that content fresh is by making sure it is personalized additional information, or in some cases different information can be presented to customers who are coming from a different journey. For example, is this customer who’s asking for a coffee machine authenticated? Did they buy a coffee machine before? Were they looking for some specific coffee machine? Are they coming from a specific location? Are they using a specific device or are they preferred customers making this entire experience fresh in terms of personalization or optimizing for what information is missing is the key to keeping it all fresh.

What's next for knowledge optimization?

Harshali Desai (14:56): 

There’s a lot of excitement in this space. There are so many interesting use cases that can use the technologies of today and they’re changing so fast.

In terms of looking at what were the questions that were asked by the customers and trying to generate what could be the best answer, we can look at conversations that agents have been having without a knowledge solution or without a bot. If this is an expertise they are bringing to the table, can we leverage that expertise and help agents who are new? Those are the spaces where there are a lot of use cases.

We are working at adding some of the processes and technologies to build in more collaboration and build in more trust in the knowledge that is presented to customers.