I built an AI fitness coach, then realized AI was only solving half my funnel
I spent 8 years working in tech, as a Product Manager and Designer. Today I'm a certified personal trainer running two things at once: FitDots, an AI-powered training app, and a 1:1 coaching business on top of it. This post is about the part nobody warns you about when you build an AI product: figuring out exactly where the AI's job ends and where a human has to take over — and building your funnel around that line instead of pretending it doesn't exist.
The problem I actually had:
When I started building FitDots, my first instinct (like most of us building AI products) was "automate everything." Generate the workout, adapt it to time and fitness level, ship it, done.
That part worked. Give the system 15 to 45 minutes, a fitness level, and feedback from the last session, and it produces a structurally solid workout, every time, at zero marginal cost. That's a genuinely good use of AI: bounded inputs, bounded outputs, no ambiguity.
But I kept seeing the same failure mode with users: people completing every workout in the app and still quietly losing motivation, or drifting toward burnout, weeks before it showed up in any metric I was logging. The AI had no signal for that, because the thing it needed to detect wasn't in the data. It was in tone, in the gap between what someone reports and what's actually going on, in change over time that only a human tracking someone across weeks would notice.
That's when I stopped thinking of "add more AI" as the roadmap, and started thinking about the product as two separate businesses stitched together.
Splitting the business, not just the product:
Here's the model I landed on, and the reasoning behind each piece:
Free tier (FitDots + a nutrition guide as lead magnet): this is the part that scales for $0 marginal cost. Adaptive workouts, and a free resource I built called "The Lean Nutrition System - Foundations" that covers the four levers that actually drive body composition (protein intake, treating carbs as fuel, adjusting for training/rest/travel days, not spiraling after one bad meal). None of this requires judgment calls — it's a framework, so it's exactly what you should be handing to an algorithm.
Paid 1:1 coaching: this is where the actual margin is, and it's deliberately NOT automatable. It's positioned for people who've already got structure from the free tier and are still stuck — the ones where the bottleneck is diagnostic (is this a protein number, a stress response, or a pattern they can't see because they're the one living it?) or behavioral (getting someone to be honest about their week, or to not abandon the whole plan after one missed session).
The free tier isn't a watered-down trial of the paid tier. It's solving a genuinely different problem (logistics) than the paid tier (diagnosis + accountability). That distinction turned out to matter a lot for how I write copy, price things, and decide what to build next in the app vs. what stays human-only forever.
The four things I stopped trying to automate:
Useful to spell out, because "AI can't replace human connection" is such a vague claim it's basically useless as a product decision. Concretely, here's what doesn't scale, and why:
Assessment — telling training fatigue apart from early burnout. Requires context on the person, not just their inputs.
Edge-case nutrition — disordered eating patterns, medical history affecting metabolism, the gap between reported and actual intake. A framework can't catch what a person isn't reporting.
Behavioral coaching — getting someone to reframe a setback instead of spiraling into all-or-nothing thinking. This is closer to therapy than to logistics.
Accountability — compliance changes when there's a real person on the other side who'd notice if something's off. This one might be the most economically important of the four; it's most of why people pay for coaching instead of just following a plan.
Why this matters if you're building an AI product?
If you're building something AI-powered right now, the question I'd push on is: what's the actual shape of the problem you're solving, and does it match what AI is structurally good at (bounded inputs → bounded outputs, applied at scale) or what it's bad at (judgment calls under incomplete information)?
Most AI products I see get pitched as "AI does everything now," which is a fine hook but a bad roadmap. The version that's actually defensible, at least for me, was admitting AI covers maybe half the value chain — and building a real (paid) business on the half it doesn't.
Happy to go deeper on the funnel numbers (free → paid conversion, what actually gets people to book a call) if that's useful to anyone building something similar.
I like the distinction between logistics and judgment. Feels like a lot of AI products try to automate the entire workflow instead of asking where humans actually create the most value.
This maps almost exactly onto something I've been thinking about in compliance tooling. Sanctions/AML screening has the same two halves — the "logistics" part (does this name match a list, structurally) is exactly what a model is good at: bounded input, bounded output, zero marginal cost. But the moment there's a fuzzy match or an edge case, the actual decision (flag this, clear it, escalate it) legally has to stay a human call — not because AI can't pattern-match well enough, but because "the algorithm decided" isn't a defensible answer to a regulator. Curious whether you've run into the version of this where the human half isn't just better UX, it's the part that has to be a human for accountability reasons, not capability ones.
That's a classic indie maker trap - you build something AI-powered and assume it'll unlock the whole business, but then you hit the wall on the completely different problem of actually getting people in the door. What did you find was the other half of the funnel that AI couldn't touch?
Certifications, trust (since it's about health), and mostly motivation and coaching. Those parts are still out of reach for current AI, and for the legal framing around it too.
Great insight.
You've clearly defined where AI stops and human expertise begins. I'd make sure the website communicates that just as clearly it's often the first place where people decide whether to trust the product or book coaching.
Clear messaging usually converts better than adding more features.
Great insight, OlaWebDesigner. Thanks for sharing.
Feel free to take a look and tell me if you think we accomplished it:
1:1 Coaching: https://productfitcoach.com/
AI SaaS: https://www.fitdots.ai/
Really resonates, especially the point about the failure mode not showing up in your metrics until it's too late. I'm working on a much smaller AI tool right now, and I hit a version of this early: tracked my funnel closely (visitors → clicks → signups), and the drop-off wasn't where I expected at all. Turned out people were engaging with the "automatable" part just fine, the real friction was somewhere I hadn't even measured.
Your framing of "what's the actual shape of the problem" vs. just defaulting to "automate everything" is a good gut check. Curious how you decided the free tier boundary specifically, did you test giving away more before landing on this split, or was it clear from the start?
The free tier has always been on the table since the start. The idea is to learn from both angles: the 1:1 side gives us insight into what can actually be automated, and the AI side is already validating certain parts on its own. Both work in a complementary way, solving toward the same direction.
In your case, where and how did you end up finding the drop-off once you dug in?
"Earning the right to sell the human part" is a sharp way to frame it, AI proving it understands someone's situation well enough that paying for a real coach feels like the obvious next step instead of a cold upsell. The upgrade moment question is the real one though, is it triggered by a plateau, a missed goal, or just raw usage volume, because each of those probably needs a different nudge to convert.
Good framing, but isn't the real question whether "accountability" stays non-automatable forever? I'd argue the gap is shrinking fast — async coaching with AI-generated nudges is already eating into 1:1 coaching margins. What's your moat if someone automates the behavioral layer in 12 months?
That's the real tension in any AI plus human hybrid model, the automatable part keeps expanding every year. I'd guess the actual moat isn't the coaching itself staying non-automatable forever, it's whether the relationship and trust built through 1:1 sessions makes someone stick around even after the automated version gets good enough, that's harder to replicate than the behavioral layer alone.
Hey dongchao, nice point. Curious what your take is: do you think an AI avatar could actually keep someone motivated to stick with a health and fitness plan?
Yes, exactly. I want to be clear that one product isn't "better" than the other, it's that for some clients, scaling into 1:1 coaching turns out to be the natural next step. For others, the SaaS, self-knowledge, and free resources are more than enough on their own. We aimed to cover the full spectrum, not push everyone toward the same.
Fitness, nutrition and workouts can be far more complicated than automations and following what's written in the plan. Emotions, motivation, framing, and the human aspect of it are always in play.
That distinction makes sense, some people genuinely just need the structure and self-serve tools, others need the human accountability layer because the emotional and motivational side is where things actually break down. Curious if you've noticed any early signal for which type a new user is, before they even tell you, or is it something that only becomes clear after they've used the free tier for a while?
I am carefully aligning the content with the user's request, ensuring clarity and relevance.
Drawing from the context of establishing an e-commerce presence, I focus on key elements such as brand identity, target audience, product range, and unique value proposition.
My goal is to deliver a coherent, professional overview that supports strategic decision-making.
This is a really good lesson. AI can handle the structured part, but it doesn’t always solve the messy human part.
I’m seeing the same pattern with my own SaaS. “Add more AI” sounds like progress, but sometimes the real work is knowing where AI should stop and where trust, context, or human judgment still matters.
Exactly! And there's so much noise around this lately. I bet it's becoming harder every time to stay focused.
Are you trying anything specific to keep yourself on track?
The logistics vs judgment split is the part a lot of AI products miss. Automation shines when the work is bounded, but once the job turns into interpretation, motivation, or reading tone over time, the human is the product. That's basically how I think about DictaFlow too: capture and clean up the words fast, but don't pretend the model should replace the person's actual judgment. Products get a lot clearer when you decide exactly where the machine stops.
Right, and "the machine stops" is doing a lot of work in that sentence. That's usually the decision some people avoid making explicitly, and then the product ends up mushy on both ends.
Where does that line land for DictaFlow though? Is it the judgment part or the rephrasing of the words?
what I liked most is that you didn't treat AI as the goal. You treated it as the tool that gets someone moving, and saved the harder part - the interpretation, the accountability, and the course correction - for a real person.
Exactly. That was the whole unlock for me. The moment I stopped asking "what else can the AI do" and started asking "what does the human actually need to be free to focus on," the roadmap got a lot clearer. Most AI feature requests I get now actually go through this filter.
You offered to go into the free → paid numbers, so I'll bite, because that's the actual hard part of this model. If the free tier genuinely solves logistics and isn't a crippled trial, you've removed the very friction that usually makes someone go looking for a coach. So what's the moment a satisfied free user realises they've hit the judgment wall and books a call? That trigger is your whole business. My guess is it's not a feature, it's a visible plateau: the app showing them "you've done everything right for six weeks and the scale hasn't moved," which is exactly the diagnostic problem only the human tier solves. Do you surface that gap on purpose, or wait for them to feel it?
Good question, and worth being precise about, because the answer is more nuanced than "AI can't do this."
AI is already good at the math: apps like MacroFactor or Welling calculate personalized calorie and macro targets from your stats, and adjust them as you progress. That part isn't the wall.
The wall is about accountability, not calculation. None of those apps take clinical responsibility for a plan, they all carry a "consult a professional before making significant dietary changes" disclaimer, for a reason. The moment there's a health condition, a medication interaction, or a pattern that needs real follow-up, that's not a math problem anymore, it's a judgment-and-liability problem. That's the part we keep on the human side, with certified professionals actually accountable for the call.
To your actual question, at the moment, right now it's closer to "wait for them to feel it" (if they ever feel it. As with the AI alone you can get amazing results) than a designed trigger, doesn't mean this cannot change, but we are still learning. What exists today is a bridge between the SaaS and the human side, so when someone hits that plateau and wants to jump to 1:1, the path is there. But a proactive "you've plateaued, here's why this needs a human" nudge isn't built yet, but is definitely something worth taking a look at. Thanks for sharing.
the part i'd push further: the AI half and the human half aren't equal halves. the AI half is the commodity — everyone's model generates a decent workout now — and the human/judgment half is where both the margin and the retention actually live. which flips what the AI is FOR in the funnel: it's not the product, it's the cheapest possible way to prove to someone that you get their problem, so the human offer converts. build the AI to earn the right to sell the human part, not to replace it.
Exactly! That's the vision behind both ProductFitCoach (1:1 coaching) and FitDots (the AI SaaS). I built the AI to earn the right to sell the human part. And the human part isn't just "a person", it's a system of certified professionals that scales behind it, which I lead on the business side, while staying the one directly coaching and supervising clients myself. That's a second ingredient AI can't replace, and a big part of why I don't feel competitive pressure from apps that just ship a better AI workout generator.
The distinction that stood out to me is that you stopped dividing the product into AI and human, and started dividing it into logistics and judgment.
That's a much stronger way to think about it. AI can remove operational friction at scale, but judgment often becomes more valuable once the logistics are no longer the bottleneck.
Yeah, that's a great way to frame it: "logistics vs. judgment".
I think that in a year or two the AI half will just get better at logistics, but judgment will not get commoditized the same way. I might actually go back and edit the post language to reflect that. Thanks for sharing!
Glad it resonated.
Your reply made me think there's one strategic decision sitting underneath that logistics vs. judgment distinction which becomes much more significant as the business grows, but I don't think I can explain the reasoning properly in a thread without oversimplifying it.
If you're interested, what's the best email to reach you on?
Hey aryan_sinh. Feel free to drop me a DM through https://www.linkedin.com/in/productfitcoach/
Looking forward!
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.
This comment was deleted 2 hours ago.