How a Solo Founder Uses an AI Weekly Planning System to Run Two Products

Marcus runs two SaaS products as a solo founder — one in its third year with paying customers, one in early validation. He has no employees, a small group of contractors he coordinates asynchronously, and roughly 45 hours per week of working time.

He has been running a weekly AI planning session every Sunday for eight months. This is an account of how that system works, what broke before it worked, and what the data shows about the difference it has made.

Note: Marcus is a composite representative of the planning patterns described in this article. The workflow, failure modes, and observations reflect real practices documented across our user community.

The Problem Before the System

Before building a structured weekly planning process, Marcus ran on what he calls “inbox gravity” — his day’s work was determined primarily by what arrived overnight and whatever felt most urgent. He was productive in the sense that he was always working. He was not productive in the sense that the two things most likely to determine whether either business succeeded rarely got his focused attention before 4pm, if at all.

“I had a list of important things I never started and a calendar full of things I never should have agreed to,” he says. “I thought the problem was focus. It was actually planning. I was not deciding what mattered — I was letting the loudest inputs decide for me.”

The specific failure mode was what researcher Sophie Leroy has described in attention residue research: the cognitive cost of switching between the reactive surface of the business (email, support, contractor coordination) and the generative core (product development, customer insight, strategy) is not just a time cost. It is a quality cost. Marcus was losing not just hours but the cognitive state required for the work that mattered most.

The First Attempt

His first attempt at structured weekly planning lasted three weeks. He used a popular planning template — a two-page spread with goals, priorities, and reflections — and spent 30 minutes Sunday evening filling it out.

The problem: the template was static. It asked the same questions regardless of what had actually happened last week or what was genuinely pressing next week. By the third Sunday, filling it out felt like paperwork rather than thinking.

“It was a beautiful template. I just wasn’t actually making decisions with it. I was filling it in to feel organized, not to get oriented.”

Building the System That Worked

The turning point was moving from a template to a prompt-driven conversation. Instead of filling in predetermined fields, Marcus started each Sunday session by giving an AI assistant a raw brain dump of the week — what happened, what he finished, what surprised him, what felt unresolved — and asking it to surface three things: the highest-leverage unfinished item, the decision he was avoiding, and the pattern he should pay attention to.

This approach worked for two reasons. First, the brain dump required genuine recall rather than performance — there were no fields to fill, so there was nothing to make look good. Second, the AI’s response consistently surfaced something he had not explicitly articulated, which made the planning session feel generative rather than administrative.

From there, he built the full five-step process.

His Current Sunday Session: The Walkthrough

Marcus runs his session on Sunday between 6:30 and 7:15pm — after the day has wound down, before the week feels imminent. He uses a combination of AI chat and Beyond Time for tracking.

6:30 — Brain dump (10 minutes)

He opens a blank note and writes everything that happened in the past week without editing. He includes what he worked on across both products, what he avoided, what felt like progress, and what feels stuck. This takes exactly as long as it takes — some weeks it is three paragraphs, some weeks it is a page.

He does not read it back. He pastes it directly into his AI assistant.

6:40 — Review analysis (8 minutes)

He uses this prompt:

“Here is a brain dump from my week as a solo founder running two products: [paste]. What are the three most actionable observations about my week — one about where I made real progress, one about what I avoided and why that might matter, and one about a pattern that has appeared multiple times in recent weeks?”

The “multiple times in recent weeks” instruction reflects one of the most useful adjustments he made over time: asking the AI to look for recurring patterns rather than just analyzing the current week. When he provides context from previous sessions, this yields substantially better insights.

6:48 — Three weekly outcomes (10 minutes)

He defines outcomes for both products separately — typically two outcomes for the primary business and one for the validation project. The prompt:

“My current highest-priority items across both products are: [list]. Given the following constraints this week — [meetings, scheduled calls, any travel or personal commitments] — what three specific outcomes would make this week genuinely successful? One should be for Product A, one for Product B, and one should be a ‘force multiplier’ — something that makes subsequent weeks easier regardless of which product benefits.”

The “force multiplier” framing came from a realization that his most valuable weekly outcomes were often not the most urgent ones — they were the ones that removed ongoing friction (automating a recurring task, clearing a decision that was blocking multiple things, having a conversation he had been postponing).

6:58 — Calendar block (8 minutes)

He opens his calendar for the coming week and assigns one 90-minute deep-work block to each outcome. He has learned through painful experience that he cannot reliably protect more than three blocks per week — his contractor coordination and customer responsibilities generate enough reactive demand to fill everything else.

He labels each block with the outcome, not the project name. “Draft onboarding flow v2” not “Product A.”

7:06 — Monday opening move (4 minutes)

A single sentence in his planning note: the first action he will take Monday morning before opening email. He treats this as a commitment, not a suggestion. The move is specific: not “work on the onboarding flow” but “open the Figma file and complete the decision point I left unresolved on Friday.”

7:10 — Constraint scan (5 minutes)

He asks the AI one more question:

“Given these three outcomes and this week’s calendar, what is the single most likely way this plan falls apart? What should I do on Monday to prevent it?”

The answer is usually obvious in retrospect (“You have a four-hour product call on Wednesday that will wipe out Thursday morning — book the Product B block on Tuesday instead”) but Marcus finds the act of making the risk explicit prevents the surprised feeling that used to accompany every week’s disruption.

What the Data Shows After Eight Months

Marcus tracks his planning data through Beyond Time, which connects his planning sessions to his time logs. After eight months, several patterns are clear enough to act on:

Outcome completion rate: His three weekly outcomes are achieved approximately 70% of the time when defined as genuine outcomes, versus roughly 35% when he retrospectively identifies them from task-list thinking. The definition quality matters as much as the intent.

Most valuable session step: The review analysis step has the highest correlation with good weeks. Weeks where he ran a thorough review (brain dump of 200+ words, AI response engaged with rather than skimmed) show meaningfully higher outcome completion than weeks with a compressed review.

Biggest persistent failure: Wednesday afternoon. Almost every week, something claims his Wednesday afternoon that was not on the plan Monday morning. He has stopped fighting it and now treats Wednesday as structurally reactive, scheduling deep-work blocks only on Monday, Tuesday, and Thursday.

The compounding effect: The most significant benefit has appeared over time rather than immediately. He now has eight months of Sunday session notes and can ask the AI to analyze the full dataset: which categories of outcome he consistently achieves, which he consistently defers, and what the weeks with the highest subjective satisfaction scores had in common. This longitudinal picture has informed several structural changes to how he runs both products — changes he would not have made without the data.

What He Would Tell Someone Starting

“Do not try to run a perfect session. Run an adequate session every week. The practice is worth far more than any individual output.

The part that was hardest to accept was that planning is not time you steal from working — it is what makes your work time worth more. Forty-five minutes of real orientation produces better execution than five hours of well-intentioned busyness. I know that intellectually. It still took about six weeks of data before I felt it.”

Your action: Take Marcus’s review prompt and use it for your next weekly session. You do not need the full system yet — just the brain dump and the AI analysis. Run it once, note what the AI surfaces that you had not explicitly thought, and decide whether that kind of input is worth building a session around.

Tags: founder weekly planning, solo founder productivity, AI planning case study, weekly planning real example