If you are integrating AI into your business, ask these questions first

AI has moved quickly from a talking point to a serious commercial priority for brokers and lenders. Across financial services, firms are no longer asking whether they should pay attention to it. The question now is where to start, how to apply it properly, and how to avoid wasting time on the wrong use cases.

That is where many businesses get stuck.

Too often, firms begin by looking at tools rather than operational problems. They jump to demos, features and vendors before they have properly defined what they actually need AI to do. In my view, that’s the wrong way round.

If you are considering integrating AI into your business, the first step is not choosing a platform. It is asking the right questions.

The first is simple: what problem are we really trying to solve?

A vague ambition to “do something with AI” is not a strategy. It is not a use case, and it is certainly not a business case. The starting point has to be operational. Where is work getting stuck? Where is time being wasted? Where are people tied up in repetitive admin? Where are service levels being affected by delays, incomplete information or excessive back-and-forth?

In lending and broking, the same pattern appears time and again. Too much effort is still spent gathering information, checking information and chasing for missing information. That gather, check, chase cycle is often where the biggest inefficiencies sit, and therefore where the clearest AI opportunities are.

The next question is whether the business is thinking about AI in the right way.

A lot of firms still approach it as if it were traditional software. They think in terms of isolated tasks and rigid process steps. That is understandable, but it is limiting. A more useful way to think about modern AI, particularly agentic AI, is as though you are hiring a new person into the business.

If you brought in a new team member, you would not simply tell them to press a button. You would explain the outcome you need, the standards you expect, the context they need to understand, and how they should improve over time. That is much closer to the right mental model.

The real opportunity is not just task automation. It is using AI to complete jobs properly and consistently, within the rules and standards of the business.

That leads to another important question: how mature are we in our understanding of AI?

Not every business starts from the same place. Some are still learning. Some have experimented with ChatGPT. Some have tested early use cases but not operationalised them. Others already know AI will form part of their future operating model and are now trying to identify the right path to implementation.

The key is honesty. Do you understand what AI can realistically do today? Have you moved beyond surface-level experimentation? Do you understand the difference between a basic chatbot and something more outcome-driven? Have you thought through governance, controls and deployment in a live environment?

The answers are important because they shape what a sensible next step looks like.

Another question I would always ask is: are we starting in the right place?

There is often a temptation to start with the most exciting or high-profile use case. In practice, the best starting point is often much less glamorous. It is usually found in the operational admin that slows teams down and consumes resource without adding much strategic value.

That is why administrative workflows are so often the right place to begin. They are frequent, measurable and often highly manual. They create friction for teams and delays for customers. More importantly, they usually offer a very clear route to demonstrating value early.

A related point is that firms need to think carefully about outputs, not just process steps.

Traditional process design tends to focus on sequences: do A, then B, then C. With AI, it is often more effective to think output first. What does a completed case need to look like? What information must be present? What standard must it meet before it can move on?

That shift sounds small, but it is important. Rather than asking the technology to mimic every individual action, you define the result you need and the standards around it. In my experience, that is a much better way to frame AI implementation.

Of course, none of this matters unless the commercial case stacks up.

That means asking practical questions. How many people are involved in the process today? How many cases are being handled? Where are the delays? How much time is being spent on manual work? What is that costing the business in real terms? What improves if that process becomes faster, more accurate and less resource-intensive?

When firms get clear on those questions, the business case usually becomes much easier to build. In many cases, the value is not just cost saving. It is also improved turnaround times, more consistent service, better customer experience and the ability to free up skilled people for higher-value work.

The final question is one of pace: how quickly do we want to move?

Some firms are ready to invest in a defined project. Others want to begin with a lower-risk proof of concept. Some are still at the point of understanding what is possible. There is no single route that fits everyone.

What matters is taking a structured approach. The firms that will get the most from AI are not the ones making the most noise about it. They are the ones asking better questions at the beginning and applying the technology where it can make a genuine operational difference.

AI is moving quickly, and businesses do need to start thinking seriously about it now. But successful adoption will not come from chasing hype. It will come from clearly understanding the problem, defining the right use case and implementing with purpose.