Healthcare billing

Asking the billing data questions directly

90%

Questions self-served without an analyst

Estimate; measured figure to follow.

2

Minutes to an answer, asked in plain language

Estimate; measured figure to follow.

4

Weekly report requests no longer routed through a bottleneck

Estimate; measured figure to follow.

How I worked it

intake volume
handoff points
manual rework
cycle time

Assess

The billing data already held the answers. The blocker was never the data — it was that getting an answer out of it meant knowing how to query it.

Implement

A chatbot over the billing data, so a question can be asked in plain language by the person who actually has it.

Measure

The measure is simple. Does the person with the question get the answer themselves, or does it still route through whoever can run the report?

Challenge, Solution, Impact

Billing data held the answers to most operational questions, but retrieving one meant knowing how to query it — so questions queued behind whoever could run the report, or went unasked.

Read more

The challenge

The billing data held the answers to most of the operational questions anyone wanted to ask. That was never in doubt.

Getting an answer out of it was the problem, because that meant knowing how to query it. So a question took one of two routes. It went to whoever could run the report and waited its turn — or, more often and more expensively, it went unasked. A question you have to file a request for is a question most people quietly drop.

What I built

I put a chatbot interface over the billing data. The people with the questions ask them directly, in plain language, and get the answer back.

There is no query language to learn and no request to file. The skill barrier that stood between a question and its answer is simply gone.

The impact

Billing answers no longer route through a bottleneck. The person with the question gets it themselves.

The quieter win is the class of question that used to go unasked. Those never showed up in anyone's queue, so nobody counted them — and they are the ones this was really built for.

Frequently Asked Questions

If the answer you need isn't here, ask me directly.

Will AI replace my staff?

No, and I'd be sceptical of anyone selling you that. AI is a force multiplier, not an employee replacement. The businesses that get real value from it are the ones that understood their own operations first — the technology amplifies whatever process you already have, including a bad one.

What tools do you build with?

n8n for orchestration, Claude and OpenAI models for the reasoning layer, and whatever systems the business already lives in — EHRs, practice management, CRMs, Microsoft and Google's stacks. The goal is to fit the tools to the operation, not to move the operation onto a tool.

Where does automation actually pay off?

In the work you do hundreds of times a month without thinking about it. Billing, intake, routing, follow-up, reconciliation, reporting. It rarely pays off in the interesting, judgement-heavy work people assume it will — that's where you want your humans.

Do you only work with large companies?

No. Most of the leverage I've seen is at mid-size operators — big enough that the repetitive work is genuinely expensive, small enough to change how they work without a committee.

How do you start an engagement?

By understanding how the business actually runs, which is usually not how the org chart says it runs. I map where the time and the handoffs are going before proposing anything. You cannot automate a process you haven't understood, and most failed AI projects are failures of that step, not of the technology.

Do you speak or come on podcasts?

Yes. Reach out through the form and pick 'Speaking' — it routes straight to me.

Still have a question?Bring the process you think shouldn't need a human doing it, and we'll talk it through.