CallMiner CEO talks generative AI, conversational intelligence

More than half (55%) of companies now have an AI board to steer technology purchases toward those that serve business needs and pass risk assessments. But only a little more than half of those have a leader akin to a chief AI officer, according to a survey of 1,800 executives conducted by Gartner last June.

That is where contact centers are right now, said Jeff Gallino, CEO and co-founder of CallMiner, a conversational intelligence platform that began in 2002 as a company that analyzed calls for the sake of improving customer service operations. In recent years, the company expanded its cloud services to sync customer service insights with other parts of its users' businesses, such as sales and marketing, and vice versa.

Gallino said he sees generative AI as a huge advancement, but he also sees a huge barrier to adoption. It's not so simple to overlay onto technologies already in place. Vendors that make generative AI easiest to use will win the competitive battle, he said. As Gartner's survey proves, companies are just beginning to evaluate and implement generative AI and need all the help they can get. We talk about this and much more in the following Q&A.

Editor's note: The following was edited for length and clarity.

What is generative AI doing for conversational intelligence and the contact center market?

Jeff Gallino: It is a transformer at this point -- has been for probably two years. Our approach to GenAI is not, "Here's a product, a GPT with some things on the front of it."

Jeff Gallino

We do have something like that in the product, CallMiner GPT, to query your own data using that as an interface. But primarily our view to the AI craze -- and I really will use the word 'craze,' because people acted irrationally for a while -- is that we see it as yet another good user interface tool of automation.

I use the term user interface carefully. I think one of the reasons that this stuff has taken off is because it cracked a nut: How do we get information out of highly trained AIs? Just being able to ask it in any language, in any way, is a huge, huge step forward for people, especially in our industry, who are used to scripting (i.e., 'You must use this API, then you must use this kind of query and a query language').

We have a sophisticated query language that we're not getting rid of because our customers have spoken clearly that they see the need for that kind of prescriptive, 'I want exactly what I want in an answer.' But they also would like the ability to say, 'What else? What don't I know?' So we've rolled out something called semantic search.

What's that?

Gallino: We call it 'the search for meaning,' which connects to other systems such as CRM and data lakes. It has a journey view [map of a particular customer's interactions with a company], so you can just look at the summary in CallMiner instead of having to retrieve it and pull it up in some CRM system. We've also added speaking with CRM, automated CRM updates for basically every major CRM, and that's just included as part of the summary product.

Then we opened up all of the APIs around summarization, so some of our customers have already availed themselves of it. They're pulling that summary data out and getting that summary data straight through our APIs instead of through the user interface because they're operating on it in some other place, maybe a data lake.

In a changing market where many large and small companies claim to be 'CX platforms,' who does CallMiner see as competition?

Gallino: The hyperscalers -- specifically Google. We [compete against] not a lot -- I'd say one in 20 deals. We'll see those guys and Microsoft. We haven't lost yet to Amazon, Google or Microsoft head-to-head. That said, there's probably a ton of deals that we're just not seeing because they're already talking to a hyperscaler. We have a pretty great relationship with Microsoft. We are in the midst of building a relationship with Amazon around a large shared customer. Outside of that, believe it or not, Genesys is actually causing me the most headaches. Nice continues to be just a pain in the butt; we lose to them here and there.

Everybody's analyzing every conversation. Do we have enough data and have hit saturation, or are you of the mind that there's not enough data and that more is better?

Gallino: That's a good question. Early in this industry, I advocated for collecting 100%, everything, primarily because at the time it was a differentiator for CallMiner.

But now you've got it.

Gallino: But now we have it. We're seeing people raise their hand and go, 'I want more data.' Few customers feel like they've hit saturation. Compliance use cases are where 100% becomes easy to talk about.

However, the economics of AI are going to force us to reconsider sampling beyond compliance cases. What I mean by that is I think it'd be a lot more use case-oriented. The stark reality is that a majority of CX companies are selling tools that sample data not because of the complexity of scaling AI to that size but more because of the cost. Customers are asking, 'What do you have to pay for this?' You're like, 'Yeah, this isn't free.' Guess what? It's going to cost more, not less, as we commoditize generative AI. They're going to push it into features that differentiate and then it's going to get expensive. We're going to go through expansion and contraction of pricing that we always see. From that perspective, I think we're getting close to saturation on customer service, typical helpdesk calls.

Did you ever imagine CallMiner would get into customer journey mapping?

We're seeing people raise their hand and go, 'I want more data.'

Gallino: We do journey mapping; we don't do journey orchestration. The difference is important. We dipped our toe into the water of orchestration through acquisition and then decided not to. In fact, we were looking at Pointillist, which Genesys finally acquired, and we looked at Kitewheel [acquired by CSG, a revenue management, customer engagement and payments tech vendor] as well. What we discovered when we got deep with those companies is they had spent all of their time and IP digging into tracking usage across digital platforms.

They were very, very good at knowing that somebody pressed this button at this time, and they laid that out in the journey for us. 'Did they click the 'call me' button?' I'm like, 'We already know that.'

I think the big thing that we do is: Every contact that we see gets the full treatment. We don't ignore any contacts. What I mean by 'the full treatment' is that a contact goes through a classification and categorization step, a [satisfaction] scoring step, and then rolls that data up across the contact center hierarchy. You not only know how each and every contact is done, but you can also roll it up and say, 'Here's how satisfied that group is. Here's how satisfied through the hierarchy.'

You can say, 'Hey, agent, did you know you've talked to this customer four times?' or, 'We've talked to this customer four times. Here's what they said before. Here's their scores across those particular parts of the journey.'

You might find out that somebody scored very, very low on knowledge on a chat the customer had prior to them getting frustrated and calling. You should know from the get-go exactly what you're dealing with. One of the uses of that data is we send it to the workforce management system. We use it to affect scheduling -- but more importantly, those kinds of immediate routing of calls -- and we try to get the person we know is calling in an escalation. There's no way to know that except by looking at their previous contact, and we use that to help them route it correctly.

Don Fluckinger is a senior news writer for TechTarget Editorial. He covers customer experience, digital experience management and end-user computing. Got a tip? Email him.