Joining me is Ronit Ghose, Industry Chair of the Future of Finance and AI Group. Ronit, thank you for coming and welcome to Wall Street to MENA. What exactly is agentic AI, and why is it different from the AI we use in our everyday lives?
Sure, and thank you for having me. Agentic AI is part of the "do it for me" economy. Generative AI, which came on the scene about three or four years ago, is creative AI — it creates words and text. It's the "show it to me" AI. When we explain it to our grandparents, they say, isn't that just Google? So that's generative AI. Agentic AI goes further — it does things for you. So you say, I want to go on holiday in July or August to Italy. Generative AI gives you all the options. Agentic AI says, here are the options — can I book it for you? And fully autonomous AI would say, I'll book it all for you. Just tell me you want somewhere by a beach or by a castle, and it'll do it.
So it's like having an agent — it books the tickets and does it all for you.
Exactly, like a human assistant would.
Is that safe, though?
When it comes to finance, both banks — with their internal compliance and governance — and regulators will make sure that the path to truly self-driving finance is measured. Think of the analogy of self-driving cars. Have you ever been in a Tesla on autopilot, or in a Waymo in San Francisco? Whenever I go to San Francisco I love getting in a Waymo — there's no driver, you just get in the back and it drives itself. Agentic AI could be the same for everything. But when it comes to finance, you want the same careful approach. Just like self-driving cars went through ten years of trials before going mainstream, finance will go through that same process. We've seen it in the Marina here, and we'll see it more across the UAE.
And now the self-driving cabs here in Dubai too.
Right. But imagine doing that in the Marina where traffic moves slowly — then imagine doing it on Sheikh Zayed Road. You want a lot of training before you put something on a highway, because things can go wrong very fast. Finance is the same.
And with AI managing finances, it has all your information — credit cards, everything. You wrote a book about the future of finance — how close are we to AI actually managing our finances autonomously?
Thank you for mentioning the book. I wrote it in 2023 and 2024, and when I read it now, so much has already changed. Things are moving really, really fast — faster, probably, than humans' ability to manage and control them. When it comes to high-risk areas like finance or healthcare, you can have strong controls — compliance, regulation, humans in the loop — because you don't want your credit card to just decide, I think Ronit wants to go shopping, and you wake up with your balance wiped out. So it won't be fully autonomous — it'll be semi-autonomous. Think of it less as self-driving finance and more as assisted finance. An agent that gives advice, suggests where to go in Italy, but doesn't book it straight away — because you might change your mind.
You said things are moving quickly. For banks in this region specifically, what does agentic AI change first?
Banks in Dubai, Abu Dhabi, Saudi, Qatar — the big ones are already quite digitally advanced, on cloud computing, with very good apps. The question now is where to deploy AI best. One clear area is wealth management. If you have very rich friends, you know they get full-service banking. The other 99.9% of us — we email or call the bank, and if we're lucky they reply quickly. A lot of clients get frustrated just trying to find basic information. Agentic and generative AI should improve that client experience at the front end and make it more enjoyable — because for most of us, talking to our bank is a bit like going to the dentist.
Something you postpone.
Exactly. And inside the bank, there are a lot of processes and workflows that can be improved through agentic AI. Banks are white-collar factories — lots of people in local and international banks doing processing work. Over time, a lot of those jobs will migrate to agentic AI and software.
Banks are going digital here, we have neobanks, and AI infrastructure is being heavily invested in across the Gulf. Does that give the region an edge in deploying agentic AI?
I think the region has three things going for it. Number one, the region is young — the average person in the UAE or Saudi is in their late 20s or early 30s. That means they're technologically savvy. If you look at YouTube or TikTok usage per capita, this region ranks among the highest in the world. People here are digitally native. Number two, the governments and leadership in the UAE and increasingly other countries in the region are very tech-forward — partly because they themselves are young and reflect their populations. They're bold in their plans, and we know what those plans look like in Dubai and Abu Dhabi — very AI and technology focused. And third, the capital expenditure on AI and data centers in the UAE is very impressive. They have capital, and they have stable, relatively low-cost energy — which is exactly what you need to run AI at scale. So you have the people, the leadership, and the spending all coming together.
What is the biggest risk when it comes to financial decisions being made by agentic AI?
The biggest risk is this: humans have flaws, but our flaws are relatively differentiated. My biases and prejudices won't completely overlap with yours. AI models, on the other hand, are often trained in the same places and built in the same places. The UAE has its own highly regarded Falcon model, but most models being used day-to-day — Claude, ChatGPT, and others — are concentrated in the same places. In finance, that creates a risk of herding. We've seen herd mentality before — everyone rushes to buy the same stock or the same real estate, prices go up, a bubble forms, and then it bursts. With AI, that risk could be amplified, because our biases won't just be our own — they'll reflect the model's biases. And those model biases will be more similar to each other than we are as individuals. Here in Dubai we have so many nationalities, so many different ways of thinking — the models don't have that level of differentiation.
Very interesting insights — and this is a conversation we'll keep having. Thank you so much for being with us.
Thanks for having us. Thank you.
