AI Weekly Planning System — My Exact Process

By Rasumon Manuel, PMP · Brainchild360 · Published July 2026

I used to start every Monday the same way. Open Notion, open email, open Slack, open my calendar, open the PM tool the client was using, and then spend 90 minutes trying to assemble a coherent picture of what mattered most that week. I would get to 10am with a rough plan and the distinct feeling that I had already wasted the morning doing it.

That wasn’t planning. That was archaeology. I was digging through the week’s rubble before I had built anything.

Research from Asana’s Anatomy of Work Global Index (2023) found that knowledge workers spend 58% of their workday on “work about work” — coordination, updates, status checks, and meetings about meetings. Weekly planning that isn’t systematised becomes part of that 58%. It looks productive. It isn’t.

I rebuilt my weekly planning system around two Claude prompts and a 10-minute Friday review. After 16 weeks of running it across live client engagements, the Monday setup time dropped from 90 minutes to 18 minutes. More importantly, the output quality — meaning: whether I actually executed the plan I made — improved measurably. I stopped over-committing by Wednesday.

This is the exact system. The prompts are verbatim. The constraints are real.

Why Weekly Planning Breaks Without a System

Weekly planning without a review step is wishful scheduling. You sit down on Monday, look at an undifferentiated list of tasks, and try to construct a rational week from a starting position of incomplete information. You don’t know what actually shipped last week versus what got rolled over. You don’t know what your real capacity is because you haven’t reconciled last week’s interruptions with this week’s commitments. So you guess. Then Wednesday arrives and the plan is fiction.

The McKinsey Global Institute’s “Superagency in the Workplace” report (January 2025) found that AI tools free up an average of 5.7 hours per week for professionals — but those professionals only redirect 1.7 of those hours to higher-value work. The other 4 hours disappear. They don’t disappear because people are lazy. They disappear because people don’t have a system for deciding where the recovered time goes.

A weekly planning system fixes both problems: it tells you what to work on, and it explicitly allocates the time you’ve freed up. Without it, the savings from your AI workflow library compound nowhere.

The System: 4 Steps, 18 Minutes

This is a thought-leadership system, not a productivity philosophy. I’m not selling you on deep work or time blocking as a worldview. I’m telling you what I run on Fridays and Mondays and why it works when most weekly reviews don’t.

1 The Friday Review — 10 Minutes

The Friday review is 80% of why this system works. Most people skip it. They end Friday by starting the weekend, which means Monday planning begins from zero — no clean record of what shipped, what slipped, or why. My review captures three things in Notion:

• Done: Everything that was completed or delivered this week, with a one-line note on what it enabled or unblocked
• Slipped: Tasks that didn’t happen, with a one-line reason (time, dependency, de-prioritised, scope changed)
• Carry-forward: What rolls into next week and whether the priority has changed

This takes 10 minutes when you do it weekly. It takes 45 minutes if you skip two weeks in a row. The review is the source of truth that makes Monday’s planning honest rather than optimistic.

Notion AI is where I store the three-section review. Each week’s review lives in a database row. The AI summarises patterns across multiple weeks when I ask — “what categories of work keep slipping?” — which is insight the raw log doesn’t surface without querying it.

2 Monday Brain Dump — 5 Minutes

Before opening Claude, I write down everything competing for the week in one document. Not organised, not prioritised — just a raw list. Client deliverables, internal projects, admin tasks, follow-ups, anything I said I would do. I add four pieces of metadata to each item:

• Deadline: Hard (external commitment), soft (internal), or none
• Estimated time: My honest read, not optimistic — I’ve learned to add 30% to any estimate involving writing or client feedback
• Who it involves: Just me, client-facing, or requires someone else’s input first
• Context from last week: Anything from the Friday review that affects this item’s priority

This takes 5 minutes because I’m not deciding anything yet. Decision fatigue starts when you try to capture and prioritise simultaneously. Separate the capture from the ranking.

3 AI Priority Triage

I paste the brain dump directly into Claude with this prompt. I’ve refined it over 16 weeks. The version below is what I run now, unmodified:

The last line matters. I added it after noticing that early versions of the prompt would subtly weight things I had framed enthusiastically in the brain dump. The explicit instruction to rank by logic removes ego from the prioritisation. Claude reflects back the rational order. It’s rarely different from what I already knew — but seeing it written without my rationalisations makes it easier to commit to.

4 Time-Block Generation

Once the ranked list is confirmed, I run a second prompt to turn it into a day-by-day blocked schedule:

The final instruction — “do not compress the schedule to fit everything” — is the most important line. The planning fallacy is real: humans consistently underestimate task duration by 25–50% (Kahneman & Tversky, 1977, psychological research widely replicated in project management literature). Giving Claude permission to tell me when the week is overbooked prevents me from building a schedule I cannot honour.

The Numbers After 16 Weeks

The 80% time reduction is not the interesting number. The interesting number is how often I executed the plan I made. Before the system: I would estimate 3 out of 5 planned priorities actually happened as planned. The other 2 were replaced by reactive work I hadn’t anticipated. After 16 weeks on the system: I averaged 4.3 out of 5 priorities executed as planned. One slot per week is still reactive — that is intentional; the 20% buffer absorbs it.

This connects directly to the McKinsey finding. The 4 hours per week of AI-recovered time that professionals typically lose to reactive drift — my system allocates those hours during Monday’s planning session. They don’t disappear because the schedule has already claimed them. You can read the full breakdown of how these workflows compound in the 11-hours-replaced case study.

What This System Doesn’t Do

It doesn’t manage your tasks throughout the week. The daily execution layer — checking off work, handling Slack, managing inbound — stays in whatever tool you already use. I use Notion. The weekly planning system is just the start-of-week and end-of-week bookends: Friday captures what happened, Monday sets the frame for what’s next.

It also doesn’t replace judgment. Claude will rank a DEFER task above a strategically important one if the metadata you feed it is wrong. Garbage in, garbage out — the system amplifies your clarity, it doesn’t create it. If your brain dump is vague, the priority triage will be vague. The discipline is in the brain dump, not the prompt.

If you’re managing projects alongside client work, the Professional Workflow Library has the complementary systems: meeting-to-action in 8 minutes, client proposals in 30 minutes. The weekly planning system sits above all of them — it’s the container that decides when each workflow runs.

Frequently Asked Questions

How long does the AI weekly planning system take?

18 minutes when run consistently: 10 minutes for the Friday review, 5 minutes for the Monday brain dump, and 3 minutes to run both Claude prompts and review the output. The first run takes 25–30 minutes while you calibrate the prompts to your specific constraints and task types.

Which AI tool do you use for weekly planning?

Claude (Anthropic) for the priority triage and time-block prompts. Notion AI for storing the output, tracking task completion across the week, and running the Friday review against what was planned. The two tools complement each other — Claude for structured thinking, Notion for memory and pattern recognition across weeks.

Do I need Notion for this system to work?

No. The core system is two Claude prompts run once per week. Notion is where I store the output and track completion. Any tool that lets you write and search — Google Docs, Obsidian, even a plain text file — works for the storage layer. Notion adds the database layer for querying patterns across multiple weeks.

What if I skip the Friday review?

The Monday brain dump becomes significantly less useful. Without a completed-work log from Friday, you’re planning from memory — which consistently underestimates what was done and overestimates available capacity. In 16 weeks of tracking, every week I skipped the Friday review, Monday’s plan was over-committed by 20–30%.

Can this system work for team leads, not just solopreneurs?

Yes, with one modification: add a “dependencies” field to the brain dump. List any task that requires someone else’s input or approval before it can move. Claude will flag these when ranking priorities and suggest when to follow up. The team-lead version of the priority triage prompt should also include a brief “team capacity” note at the top of the brain dump.