Your honest week with AI
Before the week starts, two guesses. There is broad research on how much time AI tools save knowledge workers — and on how much of that saving never actually lands as free time. Slide your honest estimates, then reveal the figures this lab runs on. Revealing seeds the meter’s opening balance for your week.
Saved time is not free time until you protect it. Three forces dissolve it: the AI productivity paradox (a hidden cost — supervising and correcting the output — eats the net gain), Parkinson’s Law (work expands to fill any unclaimed minutes), and the mere-urgency effect (we reach for what feels urgent over what’s genuinely valuable). Time management in the AI era is three decisions: what to hand to AI, paying the verification tax with open eyes, and defending the hour before something else spends it. The meter above tracks exactly that fight.
What to hand to AI — and what never to
The old urgent/important grid tells you what to do first. It does not tell you what to hand to a machine. For that you run four questions on every task before AI touches it — because “AI can” is never the same as “AI should.”
1 Value
Is this close to your real priorities, or is it overhead? High-value, judgment-heavy work is what you want the reclaimed time for — not what you hand away.
2 Reversibility / Stakes
If the AI is wrong, how bad and how undoable is it? A draft you re-edit is reversible. An auto-sent client reply or an irreversible decision is not — those demand a human checkpoint matched to the stakes.
3 Verification cost
How long does checking the output take? This is the botsitting tax. If verifying costs more than doing it yourself, delegation is a net loss no matter how fast the draft arrives.
4 Frequency
One-off or recurring? Recurring, low-stakes, easily-checked work is the sweet spot for automation — but automations rot, so they carry a maintenance cost you must budget for.
How to read the lens: high frequency + low stakes + cheap verification → automate fully. High value + high stakes + expensive verification → do it yourself. The interesting middle — useful but checkable — is AI-draft + verify, where you pay the tax on purpose. Low-value and low-frequency? Delete or decline it; not every task deserves to exist.
Sort Maya’s morning inbox
Run the lens on six items from Maya’s 9am. Click an item to cycle it through the four bins. The point is discrimination: a draft-and-verify item is not an automate-fully item, and a cheap-to-decline item should never reach AI at all.
The busy morning
9:10am. You have a 120-minute focus block before your first meeting. Five items land in the inbox, one at a time. Route each one with the lens. Every choice spends or reclaims minutes, and the meter above updates live. Watch the long tail: a choice that “saves” eight minutes now can leak twenty-five later.
The botsitting tax, then the urgency trap
1. Charge the verification tax
Pause the morning. That campaign proposal you had AI draft? Before it goes anywhere you have to verify it — read it, fact-check figures, fix the AI’s confident-but-wrong claims. This supervising time is the botsitting tax, and most people forget to count it. Estimate it, then watch it leak the meter.
2. Round 2 — the urgency trap
11:30am. The tax recalibrated your meter, and the same balance carries in. Two things compete for the same 90 minutes: a request that feels urgent, and a block of genuinely high-value deep work. The mere-urgency effect says you’ll reach for the urgent one. Watch what that costs — the meter can move backward.
Audit the delegation, then spend the hour
1. Spot the bad delegation
A colleague proudly shows you their fully-automated week and the “15 hours” it saved. Read it like an auditor. Click the four delegation flaws — the moves that violate one of the four lens-questions. Don’t flag the parts that are actually fine; good automation isn’t a flaw.
Click each flawed move. Then check to see which lens-question each one breaks.
Sam’s “AI runs my week” setup
“Every morning my agent triages the inbox and . It also files the routine receipts and posts my standup — . On Tuesday it so I didn’t have to think about it. I had it draft the quarterly report and — it sounded confident. I set up a dozen automations back in January and . I do still . Honestly, it’s freed up my whole week!”
2. You reclaimed an hour. Now what?
The triage paid off: Maya has a clear 60-minute block this afternoon that AI gave back. Here are two honest plans for it. Both look productive. Only one actually reclaims the hour — pick it, then check your reasoning.
3. Build your Reclaimed-Hour Budget
You have 73 reclaimed minutes to allocate. Toggle where they go and which protections you turn on. The net reclaimed meter responds live: allocate well and it holds near 73; leave minutes unallocated and Parkinson’s Law drains them; skip delegation maintenance and your automations rot into a leak.
The Friday that goes sideways
It’s Friday, 3pm, deep inside your protected hour. A client escalation lands — loud, but not actually a fire. You have built the block, the budget, and the protections. This is the test. Each path plays out across the rest of the week, and the carried meter moves with it. It can move backward.
Your reclaimed-hour commitment
Make it concrete. Three short answers turn the week you just ran into a plan for your own. Write yours first — it autosaves — then reveal a model answer to compare against. No grading; the value is in writing the specific commitment down.
A model commitment
Stop: “I’ll decline the two recurring agenda-less syncs and stop hand-building the weekly status roll-up — that’s ~40 minutes a week leaking into busywork.”
Hand to AI + verify: “AI drafts first passes of proposals and routine summaries. Anything internal: a 2-minute glance. Anything client-facing or with numbers: a real ~10-minute verification pass before it leaves my hands. If checking would cost more than doing it, I keep it.”
Protect for: “One calendar-blocked, do-not-disturb deep-work hour each morning for the single highest-value task on my plate — with an auto-decline rule and a weekly 10-minute maintenance review so the block and the automations both survive.”
The Reclaimed-Hour cheat sheet
One screen to keep. The lens that decides what to delegate, the rule that keeps the verification tax honest, and the checklist that makes a reclaimed hour survive the week.
1 The four-question lens
- Value — priority or overhead?
- Reversibility / Stakes — how bad if it’s wrong?
- Verification cost — the botsitting tax
- Frequency — one-off or recurring?
- High freq + low stakes + cheap check → automate
- High value + high stakes → do it yourself
- Useful but checkable → AI-draft + verify
- Low value + low freq → delete / decline
2 The verification-tax rule
- Never count an AI “saving” before subtracting the time to verify it
- If verifying costs more than doing it, don’t delegate
- Match the checkpoint to the stakes: glance for internal, real review for client-facing or numeric
- Auto-send only what is reversible and low-stakes — never the angry client
3 The protection checklist
- Block the hour on the calendar (recurring, busy)
- Auto-decline / batch rule for interruptions
- Allocate every reclaimed minute — unallocated time leaks (Parkinson)
- Weekly maintenance review — automations rot
- Triage urgency before surrendering the block (mere-urgency effect)
- Reclaim for value, don’t reinvest back into busywork
You ran the week.
You triaged an inbox to AI, paid the botsitting tax with open eyes, watched the urgency trap and over-automation move the meter backward, built a budget that holds, and protected the hour against the Friday that went sideways. Saved time only becomes reclaimed time when you decide what to delegate, verify it honestly, and defend the hour before something else spends it for you.