Hexaware Technologies Global CTO Satyajith Mundakkal (left) says AI now contributes to more than half of Hexaware's revenue and highlighted its work across sectors including financial services, healthcare and manufacturing.
Most firms still at first stage of AI adoption: Hexaware Global CTO
Cultural resistance is a big reason companies are going slow on AI implementation, since people see it as a replacement, when it isn't, says Satyajith Mundakkal
Edited excerpts:
You were recently named Top Business Transformation Partner of the Year specifically for agentic automation. In practice, what's the difference between a client using a generative AI tool and one deploying autonomous AI agents, and how many clients have actually made that jump?
Generative AI is just a term used for what's essentially a large language model. It helps you communicate back what you ask for. You ask a question, it responds. You give it a problem statement, it helps resolve it. But it doesn't take any action on its own.
With agentic AI, you can actually perform actions. Instead of just talking back to you, it does things on your behalf. That's the big jump between what generative AI initially was and what agentic AI actually does.
Also read: If AI can translate instantly, why learn another language?
There are three stages of agentic AI. Stage one is understanding and giving you an output that you can use to execute. Stage two is where the agent executes, but you have to approve it first. Stage three is going fully autonomous, where the agent takes decisions and executes on its own.
Most of our clients are at stage one. Some have matured to stage two. Stage three is very rare, mostly used in security areas, but there are a few such cases as well.
Hexaware has specifically highlighted financial services, healthcare and manufacturing as sectors where AI adoption is moving faster. What's different about these sectors?
These three verticals are seeing a lot of AI traction simply because of the technical debt they carried in the past. Take healthcare, for instance. The time taken to process drugs through compliance and regulatory approval used to be long because of manual documentation. With AI, that's been accelerated drastically, which has cut down the time needed and sped up how quickly drugs reach the market.
In banking, today's Gen Z and younger customers want banking to be faster and more interactive on mobile phones. They don't want a traditional experience. To deliver that, AI is being infused at a pace far beyond what we saw earlier. It's the same story with fraud detection and other banking functions. A lot of AI is being used to identify and eliminate fraud and other forms of wastage that banks normally deal with.
Where has AI actually reduced reporting work or compliance overhead for your clients?
Let me take insurance as an example. If you go to an insurer to submit a claim, the documents have to be verified manually. It's the same with loan documents at a bank, a lot of manual verification is involved.
Also read: Anthropic warns AI could one day build its own successors, challenging human control
With AI, this can be accelerated at least 30 to 40 times faster than normal, at a fraction of the time and cost. That's how drastically AI has accelerated documentation processing in these two industries compared to others.
What's the single biggest reason an enterprise AI rollout stalls or fails, in your experience? Is it technical, is it data readiness, or is it organisational resistance?
Honestly, it's all three, depending on the industry you're targeting. Some industries face bigger versions of some of these issues than others.
Cultural resistance is a big one, because people see AI as a replacement, when it isn't. It's more a complementary function. There are a lot of rumours going around about AI replacing jobs, but that isn't really the reality. If you're skilled, or if you reskill, you won't face a challenge. You actually evolve and get better by using AI, and your value goes up.
Second, there's poor selection of use cases, where the understanding isn't there and the choice is assumption-based. Third, there are use cases where the technology itself isn't mature enough yet, and going into those isn't the right decision. So it really is all three factors.
Also read: AI may use more water than world's annual drinking water needs by 2030: UN Report
You've expanded your partnership with upGrad for enterprise AI training. Is the bigger blocker to AI adoption right now the technology itself, or a shortage of people who know how to deploy and manage it?
If you don't learn things in a systematic way, you'll face roadblocks down the line. Our ability to train our employees systematically, so they don't run into those roadblocks later, is critical for us. That's the reason we tied up with upGrad.
