Start with the workflow, not the AI

Banking workflow automation works best when you start with the work, not the technology. Find the process that takes too long, frustrates employees, creates unnecessary handoffs, or makes an account holder work harder than they should. Then ask whether AI can make that experience meaningfully better.

That sounds simple, but with AI moving as quickly as it is, it is surprisingly easy to do the opposite.

Why is it so easy to start with the AI?

There is a lot of pressure to do something with it.

A new capability launches. Someone sees a compelling demo. The board asks about AI strategy. Before long, teams are looking around the institution asking, “Where can we use this?”

We understand why. But it is also how technology can end up looking for a problem.

Bank Director's 2026 Technology Survey found that 76% of bank executives and directors primarily view technology as a way to improve efficiency. The same research found that 72% said their bank had already implemented generative AI.

So the appetite is clearly there. What matters more is whether the technology is improving work that actually needs to be improved.

What should you look at before choosing an AI solution?

We think one of the best places to start is with the moments where friction is already obvious.

Where are employees spending time searching rather than helping? Where does the same question get answered over and over? Where is information being copied, summarized, or re-entered? Where do account holders get passed from one person or channel to another?

You probably do not need an AI workshop to find these problems. Your employees already know where many of them are.

Ask them what slows them down. Ask your contact center what people call about repeatedly. Look at where conversations stall, follow-up gets lost, or something that should feel simple takes five steps.

That gives you something much more useful than a list of AI capabilities. It gives you a real problem to solve.

Does every frustrating workflow need AI?

No. And we think that is an important part of a good AI strategy.

Sometimes the answer is a clearer process. Sometimes it is better access to information. Sometimes a simple automation is enough. And sometimes AI really can change the experience.

McKinsey has cautioned against treating AI applications like a “hammer looking for a nail”, noting that AI initiatives often struggle when they are disconnected from business value or deployed as narrow point solutions.

That distinction matters because the goal should not be to get AI into as many workflows as possible. The goal is to make the work better.

Where can AI make a practical difference?

We tend to think about this less in terms of the technology itself and more in terms of the burden it can remove.

Maybe an employee spends several minutes searching through policies and procedures before answering a question. An employee AI assistant can help surface approved information faster.

Maybe an account holder has a straightforward question at 9 p.m. AI-powered self-service can help resolve it without making them wait until the next morning.

Or perhaps an employee starts every conversation by piecing together what already happened. AI can help summarize context so the conversation starts further ahead.

Those are very different use cases, but they share something important: there was a clear problem before AI showed up.

That is usually a good sign that you are solving the right thing.

Where should people stay involved?

This is another reason we like starting with the workflow.

It forces you to look at where speed matters and where judgment matters, because those are not always the same places.

The American Bankers Association found that banks are prioritizing lower-risk AI applications that support human judgment, with institutions reporting productivity gains particularly in information-heavy workflows.

That makes sense to us.

Let AI remove unnecessary work. Let it find information, prepare context, handle appropriate routine interactions, or help employees move faster. But when the situation becomes sensitive, complex, or simply more human, make it easy for a person to step in.

The point is not to remove people from the process. It is to give them more time for the parts of the work where they add the most value.

How do you know if a workflow is worth changing?

Before you automate anything, understand what happens today.

This does not need to become a six-month process study. A few practical questions can tell you a lot:

• How much employee time does this consume?
• How many steps or handoffs does it take?
• How often does the same work happen?
• Where do people wait?
• Where do mistakes or repeat contacts occur?
• What would a noticeably better experience look like?

Then decide what outcome you actually want to improve.

Maybe employees find answers faster. Maybe fewer routine questions require a live interaction. Maybe conversations get resolved with fewer handoffs. Or maybe employees simply have more capacity for work that requires expertise and judgment.

Those are outcomes worth measuring. “We launched AI” is not.

How does this fit into a broader AI strategy?

Your AI strategy sets the boundaries. The workflow gives you somewhere useful to begin.

Last week, we talked about why an AI strategy should help an institution move rather than become another reason to wait. We think this is the natural next step.

Once you know what you are comfortable allowing AI to do, look at the work happening across the institution and find the friction worth solving first.

You do not need to know what AI will look like three years from now. None of us do.

You need enough clarity to know what you want to make better today, and enough discipline to start there.

We think the best place to start with AI is often not with AI at all. Start with the work that needs to get better.

Frequently asked questions