How Businesses Can Get AI Right: Governance, ROI & Human Accountability | Dr. Shakti Goel

Shakti Goel (0:00)

Many companies don't have a strong governance framework. Data governance is needed to curate data and to make sure that the AI that you build on top of such data is reliable, because the data itself has to be reliable. So that is data governance. AI is mainly about customer, I'm like, if you are a customer-facing company, customer solution, so you need to have that strong platform. The other thing is the use case. Why are we implementing AI in the company? What are those KPIs? Do we have enough backing from the business? Do we have subject matter expertise to build it? So a lot of things are needed before you go ahead and deploy AI into production.

Sanjay Puri (0:43)

Hello and welcome. We are here live in Delhi at the Universal AI Awards launch with Knowledge Networks. And we have Mr. Shakti here. How are you doing sir today? Thank you for being here.

Shakti Goel (0:55)

Sandhary, I'm doing very well and hope the same for you. Thank you for having me at the Universal AI Awards.

This is a really prestigious event. So I feel honored to be a part of this event.

Sanjay Puri (1:05)

Absolutely sir. Our pleasure. So I just have a few questions for you today.

So let's begin with, as AI moves from experimentation to the core of businesses, what's one thing companies are getting dangerously wrong about building AI?

Shakti Goel (1:22)

See, that's a very, very interesting question. You know, AI is a very new technology. And when we talk about AI, I'm referring to the LLM-based, large language model-based AI.

So the question to ask is, are companies ready to even try AI? Is AI being incorporated into the workflow, into the operational workflow of the company, or is it an isolated IT project? Yeah. Right. That is important. So where things are going wrong are, you know, the companies, many companies don't have a strong governance framework. Right? So there's not much data governance, not much AI governance. Data governance is needed to curate data and to make sure that the AI that you build on top of such data is reliable, because the data itself has to be reliable. So that is data governance. Then they should have large amounts of data with themselves, so that they build a customer data platform. AI is mainly about customer. You know, I'm like, if you are a customer facing company, customer solution, so you need to have that strong platform. The other thing is the use case. You know, why are we implementing AI in the company? What are those KPIs?

Do we have enough backing from the business? You know, do we have subject matter expertise to build it? So a lot of things are needed before you go ahead and deploy AI into production.

Sanjay Puri (2:42)

But sir, there are so many use cases always. You know, I think that trying to find out which one to invest in is always a challenge. So is there a particular formula that you follow to make sure that we are investing in the right use cases?

Shakti Goel (2:56)

Yeah. See, based so if since I am from a commercial organization, we have to worry about the top line and the bottom line. So I take care of projects which are not academic in nature, but which have a well-defined ROI or well-defined business outcomes. So for example, we have customer facing applications which are AI based. Another question that we look for is, have I reduced the resolution time for the customers? We do a lot of development also. Have we increased the productivity in development? We do hyper-personalization. So if I'm sending a hyper-personalization message, has that increased the number of bookings we have? So essentially, how is my bottom line and the top line getting impacted? If there are any mundane work, repetitive work, and if I implement AI, just for example, we do a lot of call analysis. Earlier, there was a team of four to five people. Now, there's only one person, and AI actually analyzes calls and is able to do analysis of many more calls than a human would do.

So these are the strong use cases, whichever noticeable business outcome that we look for.

Sanjay Puri (4:00)

Absolutely. So I'm going to move on to Yatra as a sector of travel.

So where AI increasingly makes decisions that directly affect people. Now, where do we draw the line between automation and human accountability? Because that is soon blurring, right?

Shakti Goel (4:19)

Yes, that's right. See, where can you use AI in travel? And where you should not? See, if it's a simple use case like making a booking, you could use AI, you could say, hey, I want to book flight from Delhi to Bombay for 5 PM, show me the options, you see the options, and you say, okay, so now book me flight 6C1234. And it can go and make the booking for you. That's a simple use case. However, if you are dealing with a complex itinerary, so let's say you are planning a trip to Europe, and you are going with your family, you have got kids, right?