Timing
A human session is scheduled, so there's a gap of days between “it arrived” and “I can start”. The AI answers the moment you open the app.
Consumer AI · 0→1
Replacing a calendar slot with an AI coach
New Sportstech owners had to book a human coach before their first workout, and many never got past the booking. We replaced it with an AI coach that sets people up instantly. Then people started coming back to it every day.
At a glance
01 Problem
To get started, new owners had to book a session with a human coach. We lost a lot of them right there.
02 Insight
Excitement peaks at unboxing and decays fast. A slot days away cost us the best moment we had.
03 Solution
An AI coach that turns onboarding goals, body metrics and the device into setup guidance and a first workout, instantly.
04 Result
3× daily active sessions per user. The reported outcome is daily use, not the share of new users reaching three sessions in two weeks.
01 Context
The machine arrived. The coach didn't.
Sportstech sells smart treadmills, walking pads and strength equipment, mostly in Germany. The app is where a machine turns into a routine.
When I joined, a new owner's first step in the app was to book a session with a human coach, who set up the device and gave them a plan. Our data showed a big drop-off right at that step.
I owned AI Coach adoption and daily use: making the product useful beyond workouts.
02 Problem
We were putting a calendar between users and their motivation.
A booking before the first workout
To get set up, you picked a slot with a coach. Until then, an expensive machine sat in the living room.
The wait hit at the worst moment
Days between “my machine arrived” and “I can start”, exactly when people were most excited.
It got worse as we grew
Coaching capacity grew with headcount. More sales meant longer waits, right when things were going well.
Time to first workoutDays
Excitement after unboxing
03 Insight
People didn't need a coach first. They needed to start.
“Excitement peaks at unboxing and decays fast.”
The interviews kept telling the same story. People had just spent a lot of money on a machine that was now sitting in their living room, and we were asking them to wait for a calendar slot.
The scheduled session wasn't the value. It was the thing standing in front of it, delaying time-to-value at the exact moment motivation was highest.
04 Why AI?
Why AI, and not just more coaches.
Availability
People work out at 6 a.m. and at 11 p.m. A coach isn't there then. The AI is.
Cost & scale
Coaching cost grew in a straight line with device sales. A funnel bottleneck and a cost line at once, both worse exactly when sales went well.
Repeat use
A human session answers your setup questions once, then it's over. The AI stayed. People kept coming back with questions about training, health and the device.
05 Solution
No booking, no waiting: a plan from what we already knew.
The AI Coach reads what onboarding already collects, fitness goals and body metrics, plus the connected device, and turns them into setup guidance and a first workout in one conversation. No booking, no waiting.
I worked with design, engineering and data science on what the coach needed to know at minute one, so its first answer is a plan, not another questionnaire. Then something we hadn't planned for happened: people kept talking to it.
What we designed for
Setup, the evening it arrives
The coach already knows the goal, the body and the machine, so the first answer is a plan.
Context the coach used
- GoalBuild stamina, lose 5 kg
- Body34 y · 168 cm · 74 kg · beginner
- DeviceConnected treadmill, incline
- HistoryWorkouts & streak so far
06 Measurement
Daily use is the reported outcome.
Metric definitionDaily active sessions per user
- Definition
- Daily active sessions per user. The resume reports a 3× lift, not a 3× increase in an activation rate.
- Cohort
- The reported sessions-per-user metric; no new-user-only cohort is implied.
- Window
- Daily activity. The resume does not specify the aggregation window.
- Comparison
- 3× the baseline daily active sessions per user. The chart is indexed to 1× and 3×, not absolute session counts.
Separate exampleThe 14-day window demo below illustrates cohort alignment. It is not the definition or evidence for the 3× daily-sessions result.
2 of 4 counted
A4 ✓
B2
C3 ✓
D1
Aligned to onboarding, everyone gets the same 14 days. A user who signs up on a Friday isn't penalised for it.
07 Results
3× more new users built a routine, mostly because they came back.
3×Daily active sessions per user
3×Reported daily-sessions lift, not a new-user activation percentage
0bookings standing between unboxing and a first workout
The setup fix removed the wait. But a setup flow, however fast, gets used once. The bigger effect came from repeat use: people asking the coach everyday questions about training, health and their machine. That's where most of the 3× came from, not the setup fix.
08 Guardrails
What keeps the 3× honest.
Built into the metric
Each user's clock, not the calendar
Cohorts by onboarding week and a window from each user's own day 1, so a big sales week or a Friday sign-up can't skew the comparison.
Built into the metric
A routine, not curiosity
A new AI feature gets opened once out of curiosity. Repeat daily use is the outcome; the 14-day demo is a separate teaching example.
Attribution
Know which part did the work
The setup fix wasn't the main driver. Most of the 3× came from repeat use, which changed how I thought about the feature: a companion, not an onboarding step.
09 What I learned
Ship the obvious fix, then watch what people do with it. We built a faster setup; users turned it into a daily companion, and that's where the value was.