Marcus had read the sleep research. He owned a good mattress. He’d tried every recommended sleep hygiene technique at some point over the preceding five years: consistent wake times, no screens before bed, a cooler room, magnesium, melatonin. None of it had produced lasting change.
He consistently went to bed between midnight and 1am. He consistently woke at 6am for an early call. He was running on five to six hours of sleep, felt it every day, and had come to treat it as a fixed feature of his life as a senior operations executive at a mid-size logistics company.
What he hadn’t tried was understanding why.
The Initial Session: Establishing the Coaching Context
Marcus started using Beyond Time (beyondtime.ai) after a colleague mentioned using it for workout consistency. He wasn’t expecting much from a “habit app,” but the coaching flow felt different from what he’d encountered before.
The opening conversation didn’t ask for his goals. It asked for his history.
The first prompt the system presented was something like: Tell me about your track record with sleep over the past two years. Not what you’ve tried, but what actually happened — your typical bedtime, wake time, and your best theory about what’s driven the inconsistency.
Marcus spent about 15 minutes on this response. It was the most honest account he’d given of his sleep situation — partly because no one had asked for it in that particular way.
What emerged: he consistently went to bed late not because he was doing productive work but because the late evening was the only time in his day that felt entirely his own. His schedule was meeting-heavy from 7am to 8pm. By 9pm, the house was quiet, no one needed anything from him, and he could read, watch things he actually wanted to watch, or simply exist without being on call. The late-night window wasn’t poor discipline. It was the only autonomy in his day.
He had never articulated this before. It had been invisible to him.
The Diagnosis: The Real Barrier Was Never Sleep
The coaching response to his context didn’t suggest sleep hygiene techniques. It asked a diagnostic question: When you imagine going to bed at 10:30pm, what’s the first thing you feel? Describe it specifically.
Marcus’s answer surprised him. What he felt was something like loss. The sense of giving up the only hours that belonged to him.
The follow-up: If you had two guaranteed hours in your day that were entirely yours — not late at night, but at a time that didn’t cost you sleep — would the late bedtime still be as appealing?
He thought about it and said no.
This was the diagnosis. The sleep problem wasn’t a sleep hygiene problem. It was a self-determination problem — a structure in which the only autonomous time in his day was at the expense of his sleep. Any intervention aimed at getting him to bed earlier without addressing that structure would fail, because it would require him to give up the one thing in his day that felt self-directed.
Previous approaches had all tried to solve a behavioral problem. The actual problem was motivational and structural.
Weeks 1–3: The Wrong Prescription, Corrected
The first prescription the system suggested — with Marcus’s input — was to protect a 90-minute window in the mid-evening for personal time, explicitly marked on his calendar as non-negotiable. The theory: if his autonomous time was earlier in the evening, the psychological driver of late nights would be reduced.
He tried it for two weeks and reported back honestly: it was partially working but not completely. He was going to bed slightly earlier — closer to 11:30 instead of 12:30 — but the protected window was frequently invaded by work, and even when it wasn’t, he found himself reluctant to leave it.
The coaching session for week 3 dug into why.
What emerged from that session: the mid-evening protected window felt fragile. It existed only because he’d claimed it; anyone could remove it by scheduling over it or sending an urgent message. The late-night hours felt different — they were genuinely safe from intrusion by circumstance. The structural protection wasn’t just about time allocation; it was about imperviousness to interruption.
New diagnosis layer: the issue wasn’t only about having autonomous time. It was about having time that no one could take from him.
The Pivot: Addressing the Identity Dimension
The coaching conversation at week 4 introduced a question Marcus hadn’t expected: What does working late signal about you to yourself?
This was where the coaching moved into territory that simple behavioral approaches can’t reach.
His answer, after some resistance and honest reflection: working late signaled seriousness. Commitment. That he was the kind of person who did what the job required, regardless of the hour. His father had been the same way. His early career mentors had modeled this. Leaving work “early” — even at 9pm — felt like a signal of insufficient dedication.
The coaching response didn’t challenge this directly. Instead it asked: When you imagine yourself at 60, looking back — does the version of you that succeeded do so by working until midnight, or by working sustainably?
And: Is there any version of ‘serious’ and ‘committed’ that includes sleeping eight hours?
This was reinforcement coaching in the technical sense — not cheerleading, but values clarification. The goal was to surface the conflict between his stated values (health, longevity, being present for his family) and his operational values (demonstrated by his behavior), and to help him resolve it rather than simply override it with willpower.
Weeks 5–8: A Different Prescription
The revised prescription, built on the deeper diagnosis, had two components:
First: reframe the 10:30pm bedtime not as “giving up time” but as “protecting tomorrow.” Marcus was a logistics operations thinker. The framing of sleep as input to the next day’s execution — rather than as a sacrifice of current autonomy — was more cognitively compatible with his professional identity.
Second: create a genuine wind-down ritual that felt like autonomous time, not like obedience. A 30-minute reading window before sleep, with the explicit rule that it couldn’t be interrupted by anything except a genuine emergency. Not reading for self-improvement. Reading for pleasure — whatever he actually wanted to read.
This second element was subtle but important. The late-night autonomy he’d been protecting could be relocated to the 30 minutes before sleep, provided those 30 minutes were genuinely his. The quality of autonomy mattered, not just the quantity of time.
Weeks 5 through 8 showed his most consistent progress in five years. He was averaging 10:45pm bedtimes — still not perfect, but 90 minutes earlier than his baseline, sustained across four weeks.
The Relapse and What It Revealed
Week 9 brought a major work crisis — a supplier failure that required near-constant availability for several days. His bedtime reverted entirely. Two weeks after the crisis resolved, it hadn’t recovered.
The coaching session at week 11 examined what had happened.
The relapse pattern was instructive. When external pressure became very high, the psychological safety net of the late-night autonomous window reasserted itself completely. The new system had been built on a foundation of “when things are normal” — but for an operations executive, things are frequently not normal.
This led to the final prescription: stress-proofing the system. When his working hours exceeded a threshold (he defined it as more than two back-to-back 11-hour days), he would trigger a simplified version of the wind-down ritual that could happen in 15 minutes rather than 30, and the bedtime target would relax from 10:45 to 11:30. This wasn’t giving up — it was building a contingency that acknowledged reality.
By week 12, his average bedtime was 10:55pm. Not perfect by sleep research standards, but a genuine 90-minute improvement from baseline, sustainable across both normal and high-stress conditions.
More significantly: he understood exactly why his sleep had been poor for five years. That understanding is what makes the change durable.
What This Case Study Illustrates About AI Habit Coaching
Several things worth noting:
The diagnosis took three sessions to develop fully. The first session surfaced the autonomy problem. The second session revealed the imperviousness dimension. The third — the identity work — revealed the underlying belief structure. Real diagnosis isn’t immediate.
The relapses were data, not failures. Week 9 produced the most important diagnostic insight: that the system needed stress-proofing. Without the relapse, that wouldn’t have been discovered.
The coaching never told Marcus what to do. Every prescription emerged from his own answers to diagnostic questions. This is what motivational interviewing literature predicts will happen when coaching is done well — the person being coached generates the insight and owns the solution, rather than being handed it.
And the AI’s advantage in this case was consistent, patient, non-judgmental questioning across 12 weeks. A human coach could have done the same work. What AI provided was accessibility (five-minute check-ins at 11pm when that was when reflection happened), absence of social judgment (Marcus reported honestly about the identity dimensions because he wasn’t worried about what his AI thought of him), and structural memory that maintained the diagnostic thread across sessions.
The question this case raises: What’s the thing about your habit that you’ve never articulated — the actual barrier that all your previous attempts haven’t addressed? That’s the starting point for effective coaching.
Explore the Coach Stack framework that structured this case: The Coach Stack. For a walkthrough of Beyond Time’s coaching flows: Beyond Time Habit Coaching Walkthrough.
Frequently Asked Questions
-
Is this a real case study?
This case study is a composite, constructed from real patterns that emerge consistently in AI habit coaching work. The details — the specific habit, the diagnostic sequence, the relapses, the resolution — reflect documented failure and success patterns in habit formation research and coaching practice. No individual's private information is used.
-
Why is sleep a good test case for AI habit coaching?
Sleep hygiene habits are among the most commonly failed habit interventions precisely because the barriers are complex and individual. Simple accountability tools consistently fail because the problem usually isn't remembering to go to bed — it's a tangle of work pressure, identity (executives often pride themselves on low sleep needs), and environmental design failures. That complexity is exactly where coaching approaches outperform compliance approaches.
-
What made the AI coaching different from what this person had tried before?
The diagnostic depth. Previous attempts had addressed symptoms — buying blackout curtains, setting phone cutoffs, downloading sleep apps — without identifying the actual root cause. The coaching process surfaced that the core issue was a work identity problem, not a behavioral design problem. That's a diagnosis that simple tracking tools can't make.