Rohit Kedia, chief executive officer of mid-tier IT company Xoriant, says AI is the first technology that is untrustworthy by nature which impedes its rapid adoption. In an interview with Avik Das, he discusses the evolved role of system integrators and the AI deflationary pressure impacting topline revenue.
Why does enterprise AI adoption still lag expectations?
AI is the first digital technology that is untrustworthy by nature. It is probabilistic, it doesn't work perfectly every time, and it has been built by titans of industry who haven't inspired much trust either. For people at large, that creates a trust barrier that shows up in adoption. Even a good solution may not get adopted because people aren't sure it will always work. Next, business processes haven't been redesigned for a probabilistic approach. Every business process in the world was designed to be deterministic, and that is a problem in itself. But even where good solutions have been built, adoption still hasn't happened because of trust barriers. Even when the solution works well enough, the trust barrier holds back adoption.
Last part is simply: 'Does it work well?' Even when it does, people hold back. Human-centricity has been missing from the approach, and together these things create adoption barriers. Adoption was never easy, even in past digital transformations. This time it is even harder.
How do you view the scepticism and doubts on AI that have gained some momentum since the middle of the year?
A number of ideas have kept developing and have now come together into a new approach. We are going to promote it in the market, with our clients and with our people. It is a human-centric approach to driving AI value. AI innovation is still moving far ahead of adoption and value realisation. In a few domains value realisation is starting to pick up, but the general pattern is that it is still far behind. We are now in the third or fourth year of this, so we have to ask more seriously why that is.
You can see this in the S&P 500. If AI adoption were widespread, value realisation should by now show up in the stock performance of companies broadly. That hasn't happened. The gains are concentrated in what I call the AI maker economy, as opposed to the AI user economy. That is evidence that adoption has not yet scaled enough to show up in the bottom line of companies broadly.
Then there is the extreme commentary of AI becoming super smart or all of us dying by 2030. We start with the conviction that AI will never surpass human ingenuity. As AI improves, human ingenuity doesn't go away. It rises further. The idea is to create a flywheel that keeps compounding, and we call it the HI (human ingenuity)-AI flywheel. So our entire approach to the market is now human-centric AI adoption.
How is this approach resonating with your customers?
In early conversations it has resonated very powerfully. One customer said that AI makes good software engineers 100 times better and bad software engineers 100 times worse. That's because software engineering is moving from coding to systems thinking. You don't want to type code by hand any more. That doesn't mean the role of engineering goes away.
Another example is contact centres, one of the fields that has adopted AI the most. Adoption so far looks like this: I have 500 human agents, AI can now handle fairly human-like voice interactions, so can I cut the centre from 500 people to 200? That is one opportunity, taking cost out. But you may have left a lot of value on the table.
The reason contact centre exists is because it is a customer loyalty function. And loyalty is built with human touch. The higher purpose could be turning the contact centre from a query-response function into a customer loyalty engine.
With this approach, does the role of system integrators (SIs) become more important?
I think it does. The conversation of the last two years assumed that SIs do coding. Now AI will do the coding, so SIs will matter less. We made the same assumption about SaaS companies in a slightly different context but people realised they may have overthought it.
An SI's value goes well beyond coding. It is about reimagining business processes, helping clients find the best intersection of desirability, feasibility and viability in a transformation. The other important area is feasibility. Sometimes people use AI to automate simple deterministic processes, which is the wrong use. Working through desirability, feasibility and viability is how a human-centric SI drives adoption.
Is AI putting deflationary pressure on your revenue and margins?
The deflationary pressure is in the mid-single digits in some large-scale engineering engagements. But there are also clients where I've seen that deflation and also seen expansion on the other side. We are winning a fair share of short tenured AI deals, so much so that it more than makes up for the deflation. As for AI deal sizes, they will grow when large business processes are transformed and adopted. But for this early stage of the cycle, our AI deals have become quite sizeable.