What I Learned from Speaking About Data and AI Governance at BFSI Nexus 2026

Author: Prashant Saurabh Singh

Date: 05/29/2026

Last month I had the privilege of speaking at the BFSI Nexus Conference. It was one of those experiences that reminds you why these conversations matter so much right now.

My talk was about AI governance and data control. Specifically, about what happens when AI stops just giving us recommendations and starts actually making decisions on our behalf. That shift is already happening in financial services. And most organizations are not ready for it.

Here is what I shared, and what I think every leader in this space needs to hear.

AI Is No Longer Just a Tool

For a long time, AI was the helpful assistant in the room. It would analyze data, suggest options, and then a human would decide. That model is changing fast.

Today, AI systems are approving loans, flagging fraud, routing customer complaints, and adjusting prices. All without a human pressing a button each time. That is not a future scenario. It is happening right now, in banks and insurance companies around the world.

This is not a bad thing by itself. But it does change the rules. When a human makes a decision, we know who is accountable. When AI makes that same decision at scale, across millions of customers, the question of accountability becomes much harder to answer.

That is exactly what governance is for.

The Biggest Mistake Organizations Make

The most common mistake I see is treating AI governance as a compliance exercise. Teams check some boxes, write a policy document, and move on. The real work never happens.

Governance is not about paperwork. It is about building the structures and habits that let you trust your AI systems. And trust, in financial services especially, is everything.

If your AI makes a bad call on a customer loan application and you cannot explain why, you have a problem. If you cannot prove your model was fair, you have a legal problem. If your data was outdated or biased when you trained that model, the problem started long before deployment.

Data Is Where It All Starts

I spent a good portion of my talk on data. Because you cannot govern AI without first governing your data.

Think of it this way. Your AI model is only as good as the information it learned from. If that information was incomplete, biased, or stale, the model will carry those flaws silently into every decision it makes.

The things that matter most are straightforward:

• Source: You need to know where your data comes from.

• Lineage: You need to track how it moves and changes before it reaches your AI system.

• Quality: You need quality checks that catch problems before they cause harm.

• Access: You need to know who has access to what.

None of this is glamorous work. But it is foundational.

Risk Is Different with AI

In the conference session, I walked through a simple way to think about AI risk. There are four broad categories worth paying attention to:

• Data risk: Your training data could be unrepresentative, outdated, or even tampered with.

• Model risk: AI systems can make confident mistakes, drift over time, or be manipulated through clever inputs.

• Operational risk: What happens when your AI makes a decision and there is no way for a human to step in and override it?

• Compliance risk: Regulators are catching up fast. The EU AI Act is now enforceable. The gap between what your AI does and what you can prove it does lawfully is a real exposure.

The key message here is simple. You need to know where your risks are before you deploy, not after something goes wrong.

Trust Has to Be Built Deliberately

The word trust came up many times during the session. And it came up from the audience, not just from me.

Leaders want to trust their AI systems. Customers want to trust that AI is treating them fairly. Regulators want to trust that organizations are in control of what their AI does.

But trust does not appear automatically just because the model performs well in testing. You have to build it deliberately, through three things in particular:

• Explainability: Can you explain why the AI made a specific decision, in plain language, to a customer or a regulator? If you cannot, that is a problem you need to solve.

• Fairness: Are your AI outcomes consistent across different groups of people? This needs to be tested before launch and monitored continuously after.

• Human oversight: There must always be a way for a person to review, challenge, or override what the AI has done. Especially for decisions that significantly affect people lives.

What Leaders Should Do Right Now

The audience at BFSI Nexus was a mix of senior managers and executives. So I kept the practical takeaways focused on what leaders can actually do.

• Know what AI you have: Most organisations I speak to have far more AI systems running than their leadership team realises. Someone needs to own an inventory of every AI in production.

• Classify by risk: Not every AI system carries the same weight. A chatbot that answers FAQs is very different from a model that decides credit limits.

• Assign accountability: Make sure someone is clearly accountable for each system. This sounds obvious but it is frequently skipped.

• Invest in understanding: You do not need every leader to understand the mathematics of machine learning. But they do need to understand the risks, the questions to ask, and the obligations that now come with deploying AI in regulated industries.

The Bigger Picture

What struck me most at the conference was the appetite in the room. People wanted to get this right. They were not asking whether to govern their AI. They were asking how.

That is a genuinely encouraging shift. A couple of years ago, governance felt like the thing that slowed teams down. Now it is being understood for what it actually is: the thing that lets you move faster with confidence, because you know what you are doing is sound.

Financial services, more than almost any other sector, runs on trust. AI can either reinforce that trust or quietly erode it. The difference lies in how seriously we take the work of governance.

I came away from BFSI Nexus feeling genuinely optimistic. The conversations were sharp, the questions were good, and the sense of urgency felt real. If you were there, thank you for being part of that discussion. If you were not, I hope some of these ideas are useful wherever you are in your own AI journey.

Feel free to reach out if you want to talk through any of this further. I am always happy to continue the conversation.