As one of the UK’s leading artificial intelligence speakers, Professor Birju Shah brings a rare blend of strategic insight, technical depth, and real-world experience to the global stage.
With a career spanning leadership roles at Uber, Meta, and Google Health, he has delivered transformational impact across sectors — from transportation and technology to business and healthcare.
Renowned for turning complex AI concepts into practical strategies, Birju has advised world-class organisations on scaling innovation, unlocking value from data, and futureproofing operations.
In this exclusive interview with The Champions Speakers Agency, the sought-after technology speaker shares actionable insights into AI deployment, cross-sector lessons, and the hidden risks businesses must navigate when adopting emerging technologies.
Q: When it comes to deploying AI systems at scale, what are some of the critical yet overlooked risks businesses must be prepared for?
Birju Shah: “The biggest risk, I think, is this notion of being trapped in your own current system of how you’ve always done things. We call this institutional memory. And so there’s a book by Marshall Goldsmith called What Got You Here Won’t Get You There, and I think that’s one of the biggest risks with artificial intelligence — companies trying to fit this new tool into an old way of operating.
“So instead of rethinking their entire process and workflow around a new capability that AI provides, they just try to insert AI into their legacy model, hoping it improves it a little bit. But often, what happens is you get minimal value or worse, you build something you can’t scale or trust. You might get a result you don’t understand and say, “AI isn’t working for us,” when in fact it’s the old structure that’s incompatible with how AI thinks and learns.
“You’ve got to be willing to break some glass, to rethink your business model or operational model from scratch, and that’s culturally very hard. But if you’re not willing to do that, I think you’re really going to miss the upside of what AI can offer.”
Q: For organisations at the beginning of their AI journey, what practical guidance would you offer on how to identify opportunities and build scalable value with emerging technologies?
Birju Shah: “Yeah, it’s a fantastic question. It’s one that a lot of businesses are going through right now in the various form of transforming their profit and loss of their company. So the last two years, a lot of companies that have used artificial intelligence have used it to get efficiency or productivity, and to potentially do what I think is a very simple thing: reduce labour.
“Right, so if you have customer service people, you have 100 people, you start fragmenting those calls and you say 90% of those calls are not that valuable to us. 10% are really valuable. So if only 10% are really valuable, you only need 10 humans for that call. But 90% are just — we need to get an efficacy or accuracy of 60 or 65%, we can start automating that, and then we can reduce headcount, right?
“So that’s been the last couple of years. The bigger thing has been finding value creation now. So what’s the one hero use case where you could unlock new dollar value with artificial intelligence technology, right? And AI, you know, you can’t mistake it — it’s a hammer. It’s a tool.
“You have to have a problem that you’re attacking it with. And so we’re seeing the best companies look for R&D and automating R&D. For example, if you’re in pharmaceutical industry, you’re using artificial intelligence to upload all your past data to find new drug molecules, and to go a lot faster and simulate things, right?
“So I always tell companies, you know, the best place to start is holistically look at your profit and loss, your revenue, your operational cost, your different operating activities. Figure out where you can apply that technology to get efficiency, right?
“Superpower your people to be 3 to 4x more productive. An example there is in coding. If you have coders, they could be eight to eleven times more productive with GitHub Copilot or Replit — these two technologies. And showing companies that they can do this in very minor ways as they execute is incredibly valuable just to open their aperture of what’s actually possible.
“But then from there, you quickly move into a more strategic, “let’s find one way to create value and actually supercharge our business in a unique way.” And so I always tell companies that it’s time to definitely experiment and prototype on efficiency, but it’s also time to experiment on value creation.
“And then moving to scale is actually still incredibly hard. And so this is why you see AI being used for marketing and customer service and lower efficacy and accuracy type of industries, but actually moving towards a scaled solution requires a significant cost.
“And so this is why we always kind of paint the picture of pilot a lot, but when you’re talking about scale, we have to consider the cost–benefit return of artificial intelligence, which is actually quite loaded in terms of data centre chips, personalisation data to use.
“And so you have to have a significant use case to be able to implement this at scale. And so the only significant use cases we’ve seen are cannibalising current businesses, like cars with self-driving cars, right? And so there’s kind of things like that.
“But it’s a great way to just start — start thinking about this. And then we have a little bit of a framework that I’ll talk about. I call it: Find it, Bottle it, Scale it. So in the “find it” row, you’re finding the use case, the problem, the job to be done that you want to solve, right? In the “bottle it” row — this is a little bit harder for people — what data do you have that you can train AI on like you would train a 19-year-old kid? Right? AI right now is like a 19-year-old human. Okay?
“So as you train a 19-year-old human, you give it experience, you give it examples, and as it returns a result, hopefully you don’t fire the 19-year-old kid right away. Right? You actually give it a little bit more experience.
“You give it iterations, right? Or give him or her or they iterations. AI is the same boat. You actually have to train it. It’s going to get things wrong, it’s going to lie to you a little bit, but as you give it more data and more experience, it could accelerate to a 35-year-old experienced mid-level executive pretty fast.
“And that “bottle it” row is so important — to understand what data you have, data that you could train it for, then AI to answer your questions. So like a human, you ask questions and they return you answers. That’s what in the AI vocabulary we call prompting. And so how you would talk to a human on how to get answers — that’s what you call prompting, and prompting consciousness that you have to develop.
“And so that’s the “bottle it” row. And as you do that, you start thinking of quick ways to prototype this to replace humans with AI. And that’s a kind of good framework to start thinking about executing pilots: Find it. Bottle it. Scale it.”
