Case study: how fixing the process before AI cut churn by about 60%

A client asked us for "AI." We gave them a ticketing system first.

That sounds like a bait and switch. It wasn't. It was the fastest way to get them the result they actually wanted, and it's the clearest example I have of why I don't start projects with AI.

The situation

The client was a health and wellness company. They had a great product and customers who loved it, but growth had stalled. Their plan was to use AI to bring in new customers.

That's a reasonable plan on paper. It's also a plan built on an assumption: that the problem was acquisition.

The discovery

Before building anything, we did a three-hour deep dive into their mission, their sales funnel and how the business actually ran day to day.

That's where the real problem showed up. They didn't have an acquisition problem. They had a churn problem. Customers were leaving, and when we looked at why, it came down to one thing: they didn't feel heard when something went wrong.

That changes everything about what to build. If you want more customers, you build marketing. If you're losing the customers you have, every new one is filling a leaky bucket. It was more valuable to stop customers leaving than to find new ones.

If you want a quick way to find this kind of hidden problem in your own business, try the fill-in-the-blank test in why most AI projects fail before anyone writes code.

Step 1: a process that works without AI

The first fix was deliberately boring.

We built a manual intake and ticketing flow, so every customer problem got logged in one place. Every customer got an automatic "we're on it" email telling them someone was looking at their issue.

No model, no prompts. A deterministic process: the same input produces the same result every time.

It sounds too simple to matter. But a customer who hears "we're on it" within minutes feels very differently from one who hears nothing for days. The process addressed the real complaint, which was silence, before any AI touched it.

Step 2: AI triage, once the process had proven itself

Once the manual process was running well, we had something AI never has on its own: real examples of what "working well" looked like.

That's when AI came in. Its job was triage: sorting every incoming issue by severity, high, medium or low, so the urgent ones never waited behind the routine ones.

Because the process already worked, the AI had a clear job and a clear quality bar. It wasn't inventing a workflow. It was speeding up one that people had already proven.

Step 3: governance from day one

The part nobody had thought about was governance. We set the rules before the system scaled, not after:

• No personal information in the model. Customers' personal data stays out of it.
• No medical advice. For a health and wellness company, that one isn't optional.
• Every AI suggestion is clearly a suggestion, with legal language to match.

Governance is a product decision. It's much easier to make before launch than after something forces you to.

The result

Over six months, churn fell by about 60%.

The AI helped. But the result came from the order we built things in: name the real problem, fix the process, then let AI make the working process faster, with rules in place from the start.

What to take from this

The fastest path to AI results usually runs through a process that works without AI first. Here's the short version of the order I now use on every project:

1. Name the real problem. It's rarely the one you walked in with.
2. Build a process that works without AI.
3. Run it until you've collected examples of it working well.
4. Teach AI to copy that quality.
5. Add governance before scale.

Most teams start at step 4. I explain each step, and why skipping the first three causes so many AI projects to demo well and fail in real use, in Stop starting with AI. And if you're not sure what your own step 1 is, the three questions to ask before any AI build are the place to start.

Wondering whether AI is the right fix for your bottleneck? Book a free consultation. In 30 minutes we'll look for the real problem, and tell you honestly if the first fix doesn't need AI at all. Not ready to talk yet? Take the AI readiness quiz first.