Mira is a fourth-year PhD candidate in cognitive neuroscience. She’s studying, among other things, the mechanisms of habit formation in rodent models — which made it particularly frustrating that her own habits were in disarray.
By her own assessment, she had three habits she’d been “trying to build” for over two years without success: a daily writing practice, a consistent exercise routine, and a proper end-of-day shutdown ritual. She’d tried apps, accountability partners, 30-day challenges, and multiple variations of “start Monday.” None had stuck past six weeks.
The problem, she recognized when she actually read the relevant literature carefully, was not motivation or character. It was design.
The Diagnosis: What Was Actually Going Wrong
Before designing new habits, Mira ran an audit of her previous attempts using AI to help structure the analysis.
Her prompt:
I've tried to build three habits repeatedly over two years and keep failing.
Here's what I know about each attempt: [she described the habits, her previous
approaches, and where each had broken down].
Help me diagnose what went wrong in each case. For each habit, I want to know:
(1) Was there a specified cue, or was it general intention?
(2) Was there a contingency plan for disruptions?
(3) What was the typical failure mode — initiation failure, maintenance failure,
or recovery failure after a slip?
Don't suggest solutions yet. Just help me diagnose.The AI’s analysis identified three consistent patterns:
No stable cues. Mira’s habit intentions were time-based (“in the morning”) without behavioral anchors. Her mornings varied significantly — seminars, teaching duties, and fieldwork meant no two mornings were the same. The cue wasn’t stable, so neither was the habit.
No contingency plans. When academic obligations disrupted her routine, she had no pre-specified alternative. She’d improvise, usually unsuccessfully, then experience the day as a failure — even if she managed the core behavior in modified form.
Starting over after slips. Each time she missed three or four days, she treated it as a reset rather than a disruption. She’d restart the 30-day challenge from day one, repeatedly, never making meaningful progress on the automaticity curve.
Recognizing these as design problems rather than discipline failures was, in her words, “surprisingly difficult but important.”
Habit 1: Daily Writing Practice
The design
The target behavior: 45 minutes of focused manuscript writing.
The MVB (minimum viable behavior): open the document and write one sentence. That’s all. The habit was “done” after one sentence, even if she continued for much longer.
The cue: after sitting down at her desk with coffee — a behavior that occurred every working day without exception, regardless of whether she had teaching or seminars.
The implementation intention: “If I’m sitting at my desk with my coffee, then I will open the manuscript document before opening email or any other tab.”
The contingency plan: “If I’m not at my desk (conference, travel, off-site), then I will write one paragraph in my phone’s notes app during the first break of the day.”
The AI-assisted setup
My writing habit target: 45 minutes of focused manuscript work.
My MVB: open the document and write one sentence.
My anchor: sitting at desk with coffee — happens every working day.
Please write:
1. A complete implementation intention for this habit
2. Three contingency plans for the most common disruptions I've described:
[she listed conferences, teaching days, travel]
3. One question I should ask myself in my weekly review specifically for this habitThe AI produced three contingency variants and surfaced a question she hadn’t considered: “Is the document opening fast enough, or is the friction of finding and opening the file contributing to avoidance?”
She moved the manuscript to her desktop and renamed it “OPEN THIS FIRST.” The friction reduction was immediate.
The outcome
By week 6, Mira reported that opening the document had become genuinely automatic — it happened before she consciously decided to do it, triggered by sitting down with coffee. By week 10, she was averaging 38 minutes of writing on days when she opened the document.
The automaticity score (1–10) in her weekly reviews: week 2 = 2, week 4 = 3, week 6 = 5, week 8 = 6, week 10 = 7. A normal development curve.
Habit 2: Exercise
The previous failure analysis
Mira’s previous exercise attempts had always targeted mornings. This made sense in theory — research on decision fatigue suggests that habits that require less willpower perform better when placed earlier in the day before cognitive load accumulates. But her academic schedule meant that morning availability was inconsistent, and the cue (“in the morning”) was not tied to any specific behavioral anchor.
The failure mode: on days when seminars or teaching started before 9am, the exercise intention was bumped with a mental note to “do it later.” Later never happened reliably.
The redesign
Target behavior: 30 minutes of moderate exercise.
MVB: put on running shoes and walk outside for 5 minutes. The habit was complete at the door.
New cue: immediately after her 12pm check-in message to her research group — a daily ritual she never missed, regardless of what else was happening.
The implementation intention: “If I’ve just sent the noon message to my group, then I’ll put on my shoes and go outside immediately.”
Contingency: “If I’m in a building I can’t leave during the noon break (seminar, meeting), then I will take the stairs between floors for 10 minutes before my next commitment.”
The AI-supported weekly review
Each Sunday, she ran a brief habit review with an AI:
Week [number] exercise review:
Noon habit occurred: [number] days
I completed the MVB (shoes on, outside): [number] days
Skips: [describe circumstances]
Automaticity score: [number]
Note on what happened physically during the sessions: [brief]
What patterns do you see? Is there anything I should adjust?By week 4, the AI identified a pattern: she was consistently completing the habit on days when she sent the noon message from her own desk, but skipping on days when she sent it from a seminar room or shared space. The cue was working — but only in one physical context.
She added a secondary implementation intention for seminar room days, and completion rates improved significantly in week 5.
The outcome
At 12 weeks: automaticity score of 6–7. Exercise was happening 5 of 5 available days. She had extended the MVB to a 25-minute run on most days without any formal decision to do so — the behavior expanded naturally once the initiation chain was reliable.
Habit 3: End-of-Day Shutdown
Why shutdown is harder than most habits
An end-of-day shutdown ritual — closing open tabs, writing tomorrow’s three priority tasks, and logging off — is behaviorally more complex than most habits because the cue is time-based (“end of day”), which varies, and the routine involves multiple steps.
Previous attempts had failed at the cue stage: “end of day” was not a reliable trigger because her actual day endings varied by 2–3 hours depending on what was happening.
The redesign
MVB: write one sentence summarizing what was accomplished. That’s the ritual floor.
Cue: the moment she closes her last Zoom call or meeting of the day — not a time, but a behavioral event that occurred daily, usually between 4 and 6pm.
Implementation intention: “If I close my last meeting or Zoom call, then I will immediately write one sentence in my shutdown log before doing anything else.”
Secondary cue for days with no meetings: “If it’s 5:30pm and I haven’t had any afternoon meetings, then I will start the shutdown ritual now regardless of what I’m in the middle of.”
Tracking automaticity with AI
At week 8, Mira ran a more thorough automaticity assessment:
I've been doing a shutdown ritual for 8 weeks anchored to closing my last
meeting. Here are my automaticity ratings for the past 4 weeks: [4, 4, 5, 5].
I want to answer these questions honestly:
1. Does the behavior initiate before I consciously decide to? [partial — yes
after Zoom, less reliable after in-person meetings]
2. Does skipping feel noticeably wrong? [yes]
3. Do I sometimes complete it without fully remembering starting? [no]
4. Has the cognitive effort dropped significantly from week 2? [yes]
What does this suggest? Am I on track? Anything to address?The AI flagged what the data showed: the habit was forming well for Zoom calls (reliable, digital cue) but lagging for in-person meeting endings (less reliable cue because the transition out of a physical room involves variable friction). It suggested a specific implementation intention for in-person endings: “If I walk out of the seminar room or colleague’s office, then I will take out my phone and write one sentence in my shutdown log before walking anywhere.”
By week 12, automaticity scores reached 7 for both cue contexts.
What the Case Study Illustrates
Several themes run through Mira’s experience that are broadly applicable:
Diagnosis before design. Two years of failed attempts became useful information when analyzed for specific design failures rather than treated as evidence of low discipline.
MVB as initiation engine. The minimum viable behavior was not a concession — it was the mechanism. Initiation is the hardest part of most habits. An MVB that succeeds 90% of the time builds the cue-routine-reward association far more effectively than a full-length behavior that succeeds 40% of the time.
Cue specificity matters more than behavior specificity. A vague cue (“in the morning”) attached to a very specific behavior still fails. A specific behavioral anchor attached to a rough behavior routinely succeeds. The cue is the load-bearing element.
AI as a pattern detector. The weekly review with AI surfaced the seminar-room context effect in week 4 — a pattern that would have taken much longer to notice through self-reflection alone. AI doesn’t do the behavior. But it does identify patterns in the data that help you refine the design.
Beyond Time is built around this architecture — structured habit installation with weekly review, automaticity tracking, and AI-assisted pattern identification — making it practical to run this kind of science-based process without building your own system from scratch.
For the methodology behind this case study, see How to Apply Habit Science with AI. For the underlying research, see the Complete Guide to the Science of Habit Formation.
Your action: Run your own version of Mira’s initial diagnosis. Pick one habit you’ve tried to build and failed. With an AI, analyze: did you have a stable cue? A contingency plan? Did you treat slips as resets? The diagnosis usually takes 15 minutes and tells you more than any new habit app.
Frequently Asked Questions
-
Is this a real case study?
The case study is a composite portrait drawn from the kinds of challenges and outcomes commonly reported by knowledge workers applying research-based habit methods. The specific details — behaviors, prompts, and timelines — are realistic and consistent with what the literature predicts. Names and identifying details are illustrative rather than referring to a specific individual.
-
Does this approach work outside academia?
Yes. The behaviors in this case study — focused writing, exercise, and end-of-day shutdown — are relevant to any knowledge worker: software engineers, consultants, analysts, writers. The methodology is domain-independent: the same cue design, implementation intention, and weekly review process applies regardless of what the habit is or who is doing it.