Split the work: let a model count and label, and make the choices yourself. A white space audit prompt, tested on Claude Opus 5.5 on November 6, 2026, counts open hours, labels blocks heads-down or heads-up after Dorie Clark's The Long Game, and checks hours per long-term bet. In the test it found 17.5 open hours and two of three bets with zero time.
Planning a week is two jobs that feel like one: counting the time you really have, and choosing what deserves it. Software is good at the first. It has no standing on the second, and most of the AI planner marketing quietly blurs the two.
The tool market splits along that line. Motion's Pro AI plan, listed at $19 a seat a month on annual billing, includes an AI task planner that schedules work into your calendar for you. Sunsama, at $22 a month or $17 on a yearly plan according to Ellie's comparison (prices checked October 26, 2026), walks you through a manual planning ritual and shows when you are over capacity. Both assume you already know what matters this week. Neither checks whether this week's calendar serves the things that matter over the next three years.
That gap is the one Dorie Clark writes about in The Long Game. Her argument is that perpetual execution mode crowds out long-term bets: you cannot pour more into a full glass, so white space comes first. She separates heads-down time, executing, from heads-up time, exploring and building relationships, and she quotes the line that we overestimate what we can do in a day and underestimate what we can do in a decade. Clark applies heads-up and heads-down to blocks of three to six months; I borrow the labels at the scale of a week, because a week is where the drift starts. The drift is getting stronger: an eight-month UC Berkeley Haas study of a 200-person tech company, described by Gizmodo on November 3, 2026, found that AI users took on broader scope and extended work into more hours, often without being asked.
So the workflow is one prompt, run on Friday or Sunday, which I call the white space audit. It gives the model four inputs: your long-term bets, next week's calendar, your working hours, and what happened to last week's intentions. It asks for six outputs in order: open hours per day, a label for every block, hours per bet, a survival estimate that uses last week as the base rate, at most two protected blocks and two cuts with their trade-offs, and three questions it must not answer. The calendar in the test is a sample founder week written for the test, not a real one.
Tested November 6, 2026. I had my research agent run this exact prompt against Claude Opus 5.5 (default settings, one run, no earlier conversation). The prompt:
You are my weekly planning partner. Do NOT pick my priorities for me. Your job is arithmetic, pattern-spotting and hard questions.
MY LONG-TERM BETS (things that pay off in 1-5 years, not this week):
1. Ship the self-serve onboarding so sales calls stop being the bottleneck
2. Write one essay a month to build reputation in my niche
3. Hire a senior engineer before Q1NEXT WEEK'S CALENDAR (Mon-Fri, I work 9:00-18:00):
Mon: 9-10 team standup+planning, 10-12 investor update prep, 13-14 customer call, 14-15 1:1 CTO, 15-17 sales demo x2
Tue: 9-9:30 standup, 10-11 candidate interview, 11-12 vendor call (CRM migration), 13-16 nothing, 16-17 podcast guest recording
Wed: 9-9:30 standup, 9:30-12 board deck review, 13-14 customer call, 14-15 sales demo, 15-16 marketing sync, 16-17 coffee with founder friend
Thu: 9-9:30 standup, 10-12 nothing, 13-14 candidate interview, 14-15 sales demo, 15-17 […]
The start of what came back, unedited:
## 1. White space
Assumption: 12-13 is lunch and not counted.| Day | Open slots | Hours |
|---|---|---|
| Mon | 17-18 | 1.0 |
| Tue | 9:30-10, 13-16, 17-18 | 4.5 |
| Wed | 17-18 | 1.0 |
| Thu | 9:30-10, 10-12, 15-18 | 5.5 |
| Fri | 9:30-10, 13-18 | 5.5 |
| **Total** | | **17.5** |Scheduled: 22.5h.
## 2. Labels
- **HEADS-DOWN (2h):** both candidate interviews.
- **HEADS-UP (4h):** both customer calls, the podcast and the coffee.
- **MAINTENANCE (16.5h):**
- standups and planning
- investor update prep
- CTO 1:1
- all 4 sales demos
- the CRM vendor call
- board deck review
- marketing sync
- all-hands
- weekly reviewCustomer calls are only HEADS-UP if you use them to learn about onboarding. The CTO 1:1 is only HEADS-DOWN if hiring or onboarding is on the agenda.
## 3. Hours per bet
- Self-serve onboarding: **0h**
- Monthly essay: **0h** (the podcast is related but produces no essay)
- Senior hire: **2h** (interviews only, no sourcing time)**Two of three bets get zero.**
## 4. Survival estimate
Last week you planned 6h for bets and delivered 0h. That is a 0% […]
One run is one run: the same prompt gives a different answer next time, and a different model answers differently.
What it did well: the arithmetic was right. I checked it by hand, and 17.5 open hours against 22.5 scheduled is correct, as is each day's figure. It flagged that two of three bets got zero hours, and it used last week's zero-of-six as the base rate instead of trusting the new plan. Its most useful line was the contradiction in point six: a founder running four sales demos while betting on self-serve onboarding to remove sales as the bottleneck. A coach notices that; a calendar app does not. It also obeyed the instruction not to pick priorities and left each protected block for the user to assign.
What it did badly: it left Friday afternoon, the largest open stretch at five hours, out of its proposal, and protected only five hours in total, close to the six that failed the week before. It pointed that out itself without suggesting what would make this attempt different. It filed the weekly review and the CTO one-on-one as maintenance, which depends on what happens in them, and it assumed a lunch hour the prompt never stated. One bad week is also a thin base rate.
The limits are the usual ones. The model only sees what you paste, so a calendar that hides the Slack hours produces a plan that hides them too. Run every week, the audit can turn into one more ritual you skim. And the three questions only work if you answer them yourself, in writing, before Monday. My rule: let the model count and label, make the two or three choices yourself, and put the protected blocks on the calendar before anyone else can claim them. For the Friday review side of the same week, see how to build an AI-assisted weekly review system.
Sources: Pricing, Motion, accessed November 6, 2026; Sunsama vs Motion (2026): Which Premium Planner Is Worth the Price?, Ellie, prices checked October 26, 2026, accessed November 7, 2026; AI Users Are Working Up to 90 Minutes Longer Every Day, Study Finds, Gizmodo, November 3, 2026.
Related: How to Find Your Passion · Best Self-Improvement Books · How to Make Better Decisions · What University Will Not Teach You
