Why AI Agents Need Better Operating Models to Scale with Sumit Taneja | EXL

Obviously, we have been deploying agentic AI solutions for the last 15, 18 months.

And we have seen a variety of clients trying to leverage AI in their workflows, but not always thinking of a business outcome which they want to achieve or manage.

I'll give an example.

If your underwriting today takes 30, 40 days, what should be your new target if you want to really transform it? And a conventional mechanism would be, let's just reduce by half.

put AI in parts of workflow where we can automate and that should be great.

But I think sometimes we are not thinking that, let's imagine we were not doing underwriting the way we are doing today and we can completely reimagine it.

Can we do underwriting online? Maybe that's too aggressive, maybe in a few hours, maybe in a few days, right? But having an ambition which is transformative is probably the first thing in my view.

But the second, even more interesting is, Once you start looking at AI deployments, have you looked at your operating model, how role will change, how accountabilities will change? How will you assure agents are working in production on an ongoing basis? What's your reliability metrics? And some of that is not fully thought through in the industry.

We're trying to work with the clients.

Obviously, technology is also evolving very fast, so we can understand if it's not very clear to every client.

But I think it's very important that we go back to the basics of transformation.

Let's focus on our business transformation objective and use technology and AI and agentic AI to make sure that you reach to that business objectives.

It sounds like, you know, when you're so focused on getting some small piece to work, it's almost the forest for the trees from what you're saying that you're missing maybe the opportunity to really rethink the entire process.

And it's, you know, you're bolting on a piece and are you getting a little bit of effectiveness maybe, but you're saying that you really have to think broadly and, you know, completely reinvent the entire thought process as you go in.

Agreed.

So that sounds really hard, right? You know, that's really hard.

And I'm interested to see how you guide people to think differently.

And where do you begin?