Michael was my friend and martial arts teacher for 8 years before we ever talked about building anything together. In 2024 we both opened a new chapter in our careers within weeks of each other. I left nine years at SAP, and he left a coaching agency to start his own practice. Michael came to me with an idea: what if coaches had an AI that could show them who they actually are in their sessions, the patterns and blind spots that are invisible from the inside?

That idea became JourneyLoop. A little over a year later, we're closing it down.

I want to write about this honestly, because I learned more about building a business in this one year than in my entire corporate career. We were right about the idea and wrong about the business.

JourneyLoop recorded coaching sessions, analyzed the transcripts, and gave each coach a personal AI companion that learned their practice over time. Coaches named their companions. Kate's was called Chaz. Stephen's was Atticus. Michael's was Wisp.

Here are some demo screenshots of what the platform looked like:

By the time we wound down, the platform had analyzed 364 coaching sessions. One of Michael's coaching relationships is tracked across 25 sessions and two years, and reading it back shows the client's thinking slowly shift from tactical to strategic. No single session would have revealed that.

The numbers were never big. What happened to the people using it is what I'm proudest of. Kate was working toward her LPCC credential while using her companion every day. After she passed, she told our community:

Michael saw it change his clients, not only his own coaching:

And on our very last community call, Stephen said:

If that's what a failed startup leaves behind, I'll take it.

While building JourneyLoop, Michael and I actively shaped the conversation about AI’s role in coaching by hosting workshops, talking about our ideas on podcasts, and engaging with communities like Hub4Coaches.

In addition to these public events, we also hosted 13 community calls, 8 Bootcamps and 5 “Coaching with AI Roundtable” events where we invited members of the coaching community to learn and grow together.

Launching JourneyLoop was my first serious attempt of bringing a product to market. Every day I made mistakes, and every day I learned something new.

I'm an engineer, and I did what engineers do: I built. A web platform, then a client portal, then an AI companion living in Telegram, then a community platform with tiered pricing. Every time demand didn't show up, I changed the product or rewrote the pitch. Over 14 months we repositioned JourneyLoop around ten times.

In August we finally tried something different. We stopped selling software and started offering our help directly as a service. Within weeks we had a paying client. The product had taken a full year to reach $188 a month.

The most valuable moments in JourneyLoop were always the manual ones: the one-on-one calls where I mapped a coach's goals to exactly how their companion would help, which made their eyes light up, or the hands-on onboarding coaches got from me personally. A service would have taught us every lesson of this past year in a few weeks, at a fraction of the cost. From now on, I only build a product once I know the problem is real and people are already waiting to pay.

When your hair is on fire, what’s more valuable to you – A Rolex or a bucket of water? This analogy is often used to illustrate how customers decide what (not) to buy. No matter how nice or useful the things is I’m trying to sell you, if your hair is on fire all you care about is fixing that first. This is the trap we kept falling into.

Coaches loved talking to us. Pilot clients said "Wow! This is so cool!" and called the analysis "wild." Almost everyone we demoed to called it useful and gave us encouraging feedback. We spent a year in that comfortable middle ground where people find what you're building interesting, and we mistook their enthusiasm for demand. It turned out that even our most committed users would have been fine if I took away JourneyLoop from them over night. On one of our last calls, one customer put it plainly: "I don't think it's essential to my practice." – Ouch…

Reflecting on their coaching was always nice to have, but never critical to running their business. It made them better at what they do, and they enjoyed it, but their practice didn't depend on it. It turned out that the most urgent problem for the coaches we served was finding paying clients, and we were selling them on something they could live without.

I listen differently now. Compliments and even daily use are weak signals. What I look for now is whether the problem is tied to someone's revenue or survival, whether it's something they would have to solve with or without me.

In June, before we launched paid plans, Michael and I agreed on one number: 10 coaches paying full price by the end of September. If we hit it, we'd keep going. If we didn't, we'd wind down gracefully, help everyone take their data with them, and be honest with our members.

That number did more for us than any strategy session. July was hard. We were both close to burnout, changing direction almost every week, with no clear payoff in sight. Without the gate, I'm sure we would have kept going on hope. With it, "should we keep going?" became a question we'd already answered.

September came, and two coaches were paying. We knew what to do, and we had planned from the start how to tell them.

We're not switching JourneyLoop off overnight. I'm personally building our remaining users a replacement setup on Claude and Google Drive that keeps the parts of JourneyLoop they valued most, and we're aiming to finish by the end of the year. Most of them have already been using Claude and ChatGPT with their coaching data, which is part of why we stopped: the big players caught up faster than Michael and I could keep up with.

Michael keeps coaching and building community, including a monthly conversation on AI and coaching. I've returned to engineering consulting, and I'm looking for my next niche with everything this year taught me about choosing problems.

On our final call, Michael called JourneyLoop an experiment. He also said it was "a beautiful idea" – and I agree. Most people still haven't realized just how much of who they are shows up in the language they use, and how much AI can really help them understand themselves better. Stephen called the concept revolutionary and said people will eventually catch on, "because it's just too obvious not to." I believe that too.

To Michael, my friend and co-founder: thank you for the trust, the patience, and the courage and integrity you brought to this (ad)venture. To every coach who gave us their time and their sessions: thank you for building this with us.

The loop is closed.