How to Do Evening Planning with AI: A Step-by-Step Walkthrough

Most people approach evening planning the wrong way. They either do an exhaustive review that takes 45 minutes and burns out after a week, or they do a vague “check-in” that produces no useful output. Neither works as a habit.

What follows is the approach that actually sticks: a structured, AI-assisted session with three distinct jobs and a hard 20-minute ceiling.

What You Need Before You Start

Before opening the AI tool, two conditions make the session work.

A clear endpoint for your workday. Evening planning only works if there is actually an “evening” — a moment where work stops. If you are someone who drifts between work and non-work all evening, the first thing to fix is not the planning tool, it is the boundary. Set a shutdown time. Hold it. The planning session lives at that boundary.

A text interface with memory. Use a tool that lets you have a back-and-forth conversation and, ideally, review past sessions. This is not about picking the “best” AI. It is about having a persistent thread so you can look back. A named conversation called “Evening Planning — May 2025” beats scattering your sessions across random chat windows.

Step 1: The Brain Dump (5 Minutes)

The first job is emptying your working memory of every open loop from the day. Tasks half-done, things you said you would follow up on, ideas that surfaced in meetings, low-grade anxieties about tomorrow — all of it out.

Do not narrate. Do not organize. Just dump.

Use this prompt to start:

“I’m about to close out my workday. Help me clear my head. Ask me the following questions one at a time, and after each answer I’ll say ‘next’ to move on: 1) What did I leave unfinished? 2) What did I promise or imply to someone I would do? 3) What ideas came up today that aren’t captured anywhere yet? 4) What am I still low-key worried about from today?”

The one-at-a-time format matters. When everything is asked at once, the brain processes it as a single large task and resists. When questions arrive sequentially, each answer frees up space for the next one.

After you have answered all four, ask the AI to compile your answers into a clean numbered list. That list is your open-loop inventory.

Step 2: Triage the Inventory (3 Minutes)

Now look at your compiled list. Most items fall into one of four buckets:

• Do tomorrow: Goes onto tomorrow’s task list.
• Do this week, not tomorrow: Goes into your weekly list or project file.
• Delegate or defer: Mark it and send the relevant message or set the reminder now, while you still care.
• Drop: Not actually worth doing. Let it go.

Tell the AI your list and ask it to help you sort:

“Here’s my open-loop inventory from today: [paste list]. Help me triage each item into: do tomorrow, do this week, delegate/defer, or drop. Ask me clarifying questions if something is ambiguous.”

This takes about three minutes. The AI will flag things you are half-heartedly including out of guilt rather than genuine priority. Let it.

Step 3: The One-Paragraph Reflection (5 Minutes)

Not a diary entry. Not a performance review. One paragraph that answers three questions:

1. What was the highest-leverage thing I did today?
2. What got in the way of what I intended to do?
3. What would I do differently tomorrow?

Use this prompt:

“Based on our conversation so far, help me write a one-paragraph daily reflection. Push me to be specific — not ‘I was distracted’ but ‘I checked Slack every fifteen minutes during a block I had set aside for deep work.’ Keep it honest rather than generous.”

The AI has the context of everything you shared in the brain dump. It can reflect patterns back at you. It will ask “You mentioned the client proposal three times across your answers — what’s the actual status there?” That kind of observation is harder to get from journaling alone.

Write the paragraph. Read it. Move on.

Step 4: Set Tomorrow’s First Move (3 Minutes)

This is the single most important output of the session.

The first move is not a to-do list for tomorrow. It is a specific description of the first 60-90 minute block of focused work: what you will work on and what you expect to produce.

“Write the first draft of Section 3 of the product spec, aiming for 500 words with a clear recommendation” is a first move.

“Work on the product spec” is not.

Use this prompt:

“Given the open loops we triaged and the reflection we wrote, what should be my first task block tomorrow morning? It should: start within the first 90 minutes of my workday, be something I can make meaningful progress on in a single focused block, and be the highest-leverage use of my morning energy. Give me one specific option with a concrete output description.”

Take the suggestion, edit it to fit your actual context, and write it somewhere you will see it in the morning. Not buried in an app. Visible: a sticky note on your monitor, a pinned message, the title of your first calendar block.

Step 5: The Shutdown Statement (30 Seconds)

This sounds small, but Cal Newport makes a compelling case for it in Deep Work: ending the session with an explicit verbal or written cue trains the brain that work is actually done.

Type or say: “Shutdown complete.”

Some people find this embarrassing. Do it anyway. The ritual signal — however small — is part of what tells your nervous system that it is safe to stop processing work problems.

What to Do When the Session Goes Off the Rails

You get pulled into solving problems during the brain dump. Stop. Write the problem on the list and keep moving. The evening session is not for solving; it is for capturing. Solving happens tomorrow during the first-move block.

The AI gives you generic advice. Give it more specifics. “Help me triage my tasks” produces worse output than “I’m a solo founder with one major client deliverable due Friday and two smaller internal projects. Here’s today’s open-loop list — help me triage.”

You run out of steam after step two. That is fine. Steps 1 and 4 are the non-negotiables. If you only complete the brain dump and set the first move, you have captured the most important value of the session.

You have a terrible day and do not want to reflect on it. Do the minimum: one sentence about the day (just accurate, not eloquent), and set the first move. The reflection phase is the most skippable one. The capture and the set are structural.

The Minimum Viable Session

For days where you have three minutes, not twenty, this is the stripped-down version:

“Quick evening check-in. Three things: 1) What’s one thing from today I need to capture before I forget it? 2) What’s the single most important open loop I need to address tomorrow? 3) What’s tomorrow’s first task? Let’s keep it fast — one question at a time.”

Three minutes. One capture, one priority identified, one first move set. That is enough to prevent the morning reactive scramble and close the most critical loop.

The full session is better. The minimum session, done consistently, beats the full session done twice a week.

Building the Habit: The Anchor Approach

The most reliable way to make evening planning stick is to anchor it to an existing behavior rather than scheduling it as a standalone task.

The natural anchor: the moment you would otherwise close your last work application without any ritual. That exact moment is when you open the AI conversation instead.

Do not set a separate calendar event for “evening planning” in the first month. Just catch yourself at the natural endpoint of work and redirect into the session. After four weeks of consistent anchoring, it will require less conscious effort.

For the research behind why evening planning works, see The Science of Evening Planning. For five copy-paste prompts to get started immediately, see 5 AI Prompts for Evening Planning.

Your action tonight: Open your AI tool, paste the brain dump prompt from Step 1, and run through just that one phase. Five minutes. See what surfaces.