Following his appointment as CEO of Evalueserve, Pallab Deb shares his perspective on the evolving AI landscape and what it takes to transform AI from a promising technology into measurable business outcomes.

You’ve spent years at the intersection of technology and enterprise adoption. What’s the most common mistake you see companies making with AI right now?

Pallab: They’re bolting it on. That’s the honest answer. They see a capability, they get excited, they run a pilot, and the pilot works. Then nothing happens at scale. The reason is almost never the technology. The technology is extraordinary right now, and it’s only getting better. The reason is that they haven’t figured out how to get past the friction that’s built into every enterprise: data that isn’t ready, security concerns, and the hard work of actually transforming the business.

Enterprise environments are brownfield, not greenfield. You’ve got processes that have existed for 15 years, technology investments that are deeply embedded, and a culture that won’t change overnight just because a new tool shows up. If you walk in with AI but you don’t understand that context, you’ll produce a very impressive demo and very little else. Change management is critical here; without a deliberate plan for how people, processes, and incentives adapt, even the best technology will stall out against the weight of existing habits and systems.

So what does real operationalization look like?

Pallab: It starts with the domain. Before you even talk about technology, you have to understand what the business is actually trying to do. What does an end-to-end workflow look like? Where does it break down? Where is human judgment irreplaceable, and where is it just friction? You can’t answer those questions from a distance. You have to be in it.

The companies that are winning right now are the ones that combine deep domain expertise with the ability to bring the best technology to bear on very specific problems. And then, critically, they stay. They’re not delivering a solution and walking out the door. They’re sticking around to make sure the value actually materializes for their customers, and then they’re building from there.

That domain expertise matters most in the edge cases. The clean, happy path is easy to automate. It’s the strange exception, the account that doesn’t reconcile, the file stuck in dispute, the situation nobody mapped, that breaks things. If you’re not running the process every day, you don’t know where it actually breaks. We do, because we’ve been doing this work across hundreds of companies and thousands of workflows. Those edge cases are the difference between a demo that works and a system that works in production.

The engineering has to match that ambition. It’s not enough to build something that works once. You have to build agentic workflows that can scale, adapt as the underlying models evolve, and keep driving the process forward rather than just responding to it. The infrastructure has to be supported over its lifetime. And increasingly, the commercial model has to reflect that, priced on the outcomes the workflow actually delivers, not just on the build. That’s a different kind of commitment. It’s also the only model that makes sense if you’re serious about the value materializing.

But here’s where it really takes off: when you can do this in a repeatable way. The breakthrough isn’t solving one workflow once. It’s solving it faster, better, and more cost-effectively the next time, because you bring assets to the table that already address the domain challenge. Take spreading in commercial lending. Rather than rebuilding the solution for every client, we bring something like Spreadsmart that accelerates how spreading gets done from day one. That’s the difference between a one-off project and a capability that compounds.

You’ve talked about a three-legged stool. Can you walk through that?

Pallab: The first leg is domain expertise. You need a deep understanding of the customer’s industry, their business model, their processes, and what they care about. Without that, you’re going to show up with a hammer looking for nails.

The second leg is technological depth. You need to be current, genuinely fluent in what’s available and what’s coming, and you need to have built that into platforms and solutions so you can move fast. Time to value-realization matters. No one wants a twelve-month engagement before they see results.

The third leg is trusted relationships. The transformation decisions are being made by lines of business and CXOs. If you don’t already have a relationship there, if they don’t already trust you, you’re starting from behind. That trust takes years to build. It’s one of the most durable competitive advantages a company can have.

Most companies have one or two of those legs. Very few have all three in the same place.

There’s a lot of pressure on CFOs and finance teams right now to justify AI investment. How should they be thinking about it?

Pallab: First, let’s name the conundrum they’re sitting in. There’s a belief taking hold that if you’re not spending heavily on tokens, you’re not doing enough. Jensen Huang has made a version of this argument: if an engineer costs $500,000 and you’re only spending $250,000 on tokens for them, you’re underinvesting. I don’t think that’s the right way to look at it. Token-maxing is not a proxy for value creation. It’s pulling the conversation toward tokenomics and away from business value, which is the only thing a CFO actually cares about.

So reframe the question. Stop asking, “What is this AI tool going to cost?” and start asking, “What is this workflow costing us right now?” The business case for AI transformation
almost always lives in that second question.

When you map out what it costs to run a process at the current level of quality, speed, and error rate, the math usually becomes very clear, very fast.

What the CFO really wants to know is simple: when I bring in AI, and knowing what AI costs, is the combined cost below what I’m paying for the human-only approach today? And at scale, is the token-plus-human model producing more throughput than humans alone? The real question is where those cost curves intersect as you scale agents alongside people. That’s the analysis worth doing.

But it requires tech, business, and finance to be in the same room. The finance team needs to understand what the workflow looks like before and after. The business team needs to understand what’s actually feasible. And the technology team needs to understand what outcome they’re being held accountable for. When those three aren’t aligned, you end up with a proof of concept that never becomes a product.

Why should a company come to Evalueserve as their partner for this work?

Pallab: The companies that get this right don’t do it alone. They find a partner who can close the gap between what the technology can do and what the business needs it to do. That’s a very specific capability, and it’s rarer than the market would have you believe.
What makes Evalueserve different is that we come in with all three legs already built. The domain expertise is real and deep, earned across 25 years of work inside specific industries and functions. The client relationships are long-standing, which means we’re not spending months getting oriented. We already know the processes, the pressure points, the history. And we’re building out the technology muscle to make sure we’re bringing the best of what’s available to bear on the right problems.

Because we understand the business process, we take a long-term view. We’re not just solving for today; we’re building something that still holds up five years from now. And critically, we know where the guardrails are. In knowledge-intensive work, the cost of getting it wrong is enormous. For example, a patent filed incorrectly can cost millions. So our people always surround the technology we deploy. We can underwrite the output and account for the edge cases, because we’ve operated these workflows ourselves. That’s the difference between a tool that’s confident and a system you can actually trust.

The way I think about it: a great partner doesn’t just deliver a solution. They help you figure out what the right problem is, they move fast to show you value, and then they stay to make sure it compounds. That’s the kind of partnership we’re here to build.