AI, but make it useful: A conversation with Sedgwick’s Chief AI Officer

As head of brand and content at Sedgwick, I get the opportunity to talk with many of our leaders and subject matter experts. Usually, I come to those conversations with a topic, a few questions and a pretty good idea of where I think things might go.

My conversation with Balaji Viswanath, Sedgwick’s Chief AI Officer, was no different.

I had an idea – let’s talk about AI. More specifically, where AI can drive efficiency, where human expertise matters most, and how we should think about the balance between the two.

Before I could get very far, Balaji had a question.

Balaji: Is somebody pontificating about AI?

Julia: Well, technically, that was going to be you.

Balaji: Who wants to hear one more person pontificating about AI?

Fair point.

AI is here. We’re all using it. The mystery is largely gone. And there are already plenty of people willing to tell us what AI is going to mean for business five or ten years from now.

Balaji was much more interested in talking about what’s real.

Sedgwick is heavily invested in technology and AI, and we’re steadily integrating it into our operations. But as Balaji explained, the starting point isn’t finding places to put AI. It’s finding places where AI can create meaningful business value.

Julia: Okay, no pontificating. So how are you thinking about AI at Sedgwick?

Balaji: We’re approaching AI with a laser focus on adding top-line or bottom-line value. It has to have a top-line or bottom-line argument for us to pick up a project or product.

We’re not going to do AI for the sake of AI. We’re not going to do AI because it’s a cool new toy, or because we want to be known as “AI whatever.”

We’ll do AI because it really adds value.

The starting point for anything we pick is business value, as opposed to, “Hey, that’s a cool AI use case.”

Julia: Which is perhaps not the answer people expect from the Chief AI Officer.

Balaji: Maybe not. But it should be.

Julia: So where are you seeing that value today?

Balaji: One example is conversational AI for first notice of loss, or FNOL, for some of our clients in workforce absence and casualty.

The idea is that conversational AI can do the clean intake, so our care professionals are able to spend their time on much more human factors, like empathy.

That matters because people tend to reach out to us at arguably some of the most difficult times in their lives. And I think the empathy factor is often underrated.

There are parts of an intake that are very objective. What’s your name? What happened? What information do we need? AI can do a lot of that heavy lifting.

But if there is an emotional component, or the conversation reaches a point where the AI cannot handle the intent or the facts, we need to be able to bring the conversation over to a human.

Julia: Which gets us back to the question I originally wanted to ask you. Where does AI end and human expertise begin?

Balaji: It depends on the work. 

The goal is to have AI handle more of the heavy lifting around data collection so our care professionals can truly focus on the value-added work.

Julia: So AI isn’t necessarily replacing the human interaction. In some cases, it’s actually creating more room for it.

Balaji: Yes.

That may be one of the less flashy, but more important, ways to think about AI.

If technology can handle the repetitive collection and processing of information, people have more capacity for the things technology is considerably less suited to provide: judgment, expertise, reassurance and empathy.

Julia: What’s another example where you think AI is doing something genuinely useful?

Balaji: Workers’ compensation wage calculation.

Calculating wages can be a hugely manual process. You’re dealing with hundreds of documents, unstructured data, spreadsheets and state-by-state nuances. It can take weeks of work to get one wage calculation right.

Julia: So how does AI help here?

Balaji: Think about all of the places the information might be sitting. A document here. A wage statement there. A PDF. Something handwritten. A spreadsheet. Rules on a state website.

Instead of someone having to find and consolidate all of that information manually, the AI agent can bring those unstructured sources together. That’s where AI becomes particularly useful.

Julia: So again, this isn’t about AI doing someone’s job. It’s about taking a process that can consume an enormous amount of someone’s time and making it considerably more efficient.

Balaji: Right.

Julia: Are most of the opportunities you’re looking at directly related to claims?

Balaji: No. There are a lot of opportunities across the company.

We’ve built an AI agent to help with the RFP process, which involves reading documents, finding information from previous deals and developing responses. That’s a very manual, effort-heavy process, and AI can help the team turn RFPs around much faster.

We’ve also built an agent for finance to automate monthly journal entries, and an employee self-service agent where employees can ask questions about things like benefits and policies.

There’s another one we’re piloting around onboarding.

A new employee has a lot of questions. How do I use this tool? What’s the process? Where do I find something? What are the policies?

Historically, many of those questions go to an onboarding buddy or mentor who also has a day job. We’ve built an AI agent so a new employee can get answers to many of those questions without always needing to ask another person.

Julia: So the onboarding buddy gets to do their actual job.

Balaji: That’s the idea.

Julia: You’ve only been here for about six months. How quickly are you trying to move all of this?

Balaji: Very, very fast.

It will take time for everything to be adopted. But we want to have things standing at the toll booth so that when the businesses are ready and we’re able to lift the toll gates, adoption can accelerate.

Julia: I like that. Build the capability, prove it works, and have it ready when the business is ready to use it.

Balaji: Exactly.

There’s another part of that equation that matters too: not everyone is going to adopt AI in the same way or at the same speed.

Industry data suggests that younger populations tend to have a greater propensity to go to AI first when they need an answer, while other populations may still turn first to websites or more traditional sources. We don’t have enough Sedgwick data yet to draw conclusions about our own population because these tools have been deployed so recently.

Julia: So how much more is on your roadmap?

Balaji: There is a lot more.

Julia: Define “a lot.”

Balaji: I won’t give an exact number, but it’s not small. And we’re deep in organizing and prioritizing mode.

Julia: Great! Let’s talk about those.

Balaji: No, not yet…

Julia: Come on. What are we investing in now? What’s next?

Balaji: You can stop recording now 😊

And there it was.

After starting our conversation by accusing me of trying to make Sedgwick’s Chief AI Officer pontificate about AI, I had my answer.

The more interesting story isn’t that Sedgwick is investing in AI – of course we are.

It’s about how we’re deciding where to use it. 

Start with the business problem. Look for the places where technology can remove friction, process information faster or take repetitive work off someone’s plate. Measure whether it actually works. And keep people focused on the things people are uniquely good at.

As for the other 90-plus things Balaji and his team are working on?

Apparently, I’ll need another meeting.

Just don’t tell him we’re going to talk about AI.