What We've Actually Been Building

I’m often asked what we’ve actually been working on and building.

The problem is that a lot of the work is under NDA, so I can’t always talk about the company, the exact problem, or everything we learned along the way.

But I can talk about the kind of work and, more importantly, why we got involved in the first place.

Most of these projects didn’t start because somebody said, “We need AI.”

They started because something wasn’t working the way they wanted it to.

Credit underwriting

The conversation here usually starts with the same frustration.

A credit analyst is pulling information from too many places, trying to figure out what matters, checking whether they’ve seen the customer before, looking at payment history, reviewing the asset, and working through the company’s underwriting process.

There is a lot of information. The problem is getting the right information in front of the analyst when they need it.

That is what we started working on.

Applications can come in through email, an online application, or another system. The tool reads the information, organizes it, checks for an existing customer, pulls relevant history, validates what was submitted, and then works through the company’s existing underwriting process.

The point is not to have AI make the credit decision.

The point is to help the person making the decision see what matters faster and more clearly.

Data cleanup

This one started with a problem almost every company has but nobody gets excited about talking about.

Bad data.

Duplicates. Missing fields. Different formats. Different names for the same thing. Information sitting in systems that people don’t completely trust.

And then somebody says they want to use AI.

That becomes a problem pretty quickly.

So we built a tool that helps identify those issues, normalize the data, flag inconsistencies, and get the information into a structure that is actually usable.

It doesn’t just start changing things.

It shows you where the problems are first.

Because putting AI on top of bad data doesn’t fix the data.

It usually just gives you a faster way to get a bad answer.

The Vault

The Vault was a tool we used to scrape a 35-year-old company's data, analyze it, and then repurpose it and clean it up with proper data tags, meta tags, and related items.

Turning operating knowledge into workflows

A lot of the time, we get involved because somebody knows a process is too dependent on a few people.

Ask three people how something gets done and you may get three different answers.

The SOP says one thing.

The system says another.

And the person who has been doing the job for eight years knows what actually happens.

That is usually where the work starts.

What triggers the process?

What information is needed?

Who touches it?

Who makes the decision?

What happens when something doesn’t fit the normal process?

When does somebody need to step in?

We’ve been taking that knowledge out of people’s heads and turning it into workflows that can actually be used by the business.

Then you can decide where people belong, where technology can help, and where automation makes sense.

Equipment finance intake and asset setup

This work usually starts because something that looks simple on the front end gets complicated once it starts moving through the company.

A new transaction comes in.

Now sales, operations, credit, and documentation all need different information.

Applications. Equipment descriptions. Titled assets. Fees. Tax and title issues. Collateral information.

If something is missing or unclear early, somebody downstream ends up fixing it.

That creates rework, delays, emails, and people chasing information.

So we’ve been working through how that information comes in, how it gets validated, who needs it, and where decisions actually need to happen.

None of this sounds particularly glamorous.

But this is where a lot of the real work is.

And that has probably been the biggest lesson from the projects we’ve been involved in.

We are rarely brought in because someone needs “AI.”

We are brought in because something in the business is harder, slower, more manual, or more dependent on people than it should be.

That is where we start.

Understand the problem.

Understand the work.

Then decide what actually needs to change.

Sometimes that is AI.

Sometimes it isn’t.