Video: Human in the loop

Hi everyone, this is Steven Ramirez.

I’m the CEO of Beyond the Arc.

I just wrapped a webinar presentation with BAI by ProSight, a team I’ve been excited to work with on topics related to compliance, bank operations, and bank innovation for the last several years.

This webinar was something I really look forward to, and I’ll share the link when it’s available.

You can check out the full transcript and the video, but in the meantime, here are three key ideas that got me really energized.

First, thinking about AI in a regulated environment.

There are a couple of key ways you introduce and manage the idea of human-in-the-loop.

The first comes down to your prompting strategy.

A lot of people work with AI interactively on a chat platform like ChatGPT or Gemini or Claude.

But beyond the actual prompt is the importance of setting context.

That context can include additional documents.

It could be regulations.

It could be policy manuals.

These are the types of resources you introduce to make the AI smarter and grounded in the information, to make it fact-based as part of your prompting strategy.

I think this idea of managing context is critical to reduce the possibility of hallucination—something you have to avoid in a regulatory context.

That’s idea number one.

Very exciting stuff.

Number two is how you interact with the AI.

If you take a platform like ChatGPT, for example, you have the ability—before the AI executes your plan, your context plus your prompt—to have the AI explain how it’s going to approach the task.

You can ask what the AI thinks the task is, and how it will interpret your instructions to carry it out.

And you can ask the AI to present that plan back to you for your approval.

In the example I gave, it was an analysis of the three most important aspects of Regulation E because I wanted the AI to create an infographic based on that set of regulations.

It gave me its first take on the three ideas.

We had some back and forth.

For example, I wanted it to specify that Regulation E does not apply to small businesses.

Those inputs were part of a dialogue to further refine the action plan the AI was going to follow before it generated that infographic.

And I’ll share that infographic so you get a sense of what I’m talking about.

It’s very powerful that AI can, out of the box, create such immediate value.

The third idea is the kinds of processes and procedures you need to follow within your team to document your workflow when it comes to AI.

I laid out a five-step process that progresses from identifying your use case to documenting your inputs, including your prompt history and your context.

Then downstream, it includes the types of monitoring and governance you apply to that.

It’s essentially a step-by-step process to require humans-in-the-loop to manage AI in a regulated environment.

I think this is a topic that’s going to be hot for the rest of this year.

I’m really excited to share it with you.

This is just a little teaser for what’s to come.

If you want to find out more, check my feed next week.

Again, I’m Steven Ramirez.

I’m the CEO of Beyond the Arc.