Quick Answer
A solo founder now covers the work of a 4-person team by making AI the default first pass on every task, then editing instead of creating from scratch.
What you get
- Clear a 40-email inbox in 20 minutes with AI-drafted replies you edit, not write
- Ship 1-2 small product changes a day using an AI coding agent
- Turn every sales and user call into notes and follow-ups automatically, zero manual typing
- Run the entire founder stack for $54/mo required, $84/mo with research and docs added
Step-by-Step Workflow
7 steps · 3 hours to set up · runs inside your normal day ongoing
Workflow at a glance
7 steps · 3 hours setup
Inbox triage
ClaudeShip in Cursor
CursorLive call notes
Research pass
PerplexitySupport + admin
ClaudeFounder writing
Claude
NotionShut down
ClaudeNotionInbox triage
Ship in Cursor
Live call notes
Research pass
Support + admin
Founder writing
Shut down
- 1
9:00 - Inbox and the day's one thing
Open your inbox with Claude in a second window. Paste the subject lines and first lines of unread mail and ask: 'Sort these into reply-now, reply-later, and ignore. Draft one-paragraph replies for reply-now in my voice: direct, warm, no filler.' You edit the drafts, you do not write them. Then ask Claude: 'Given my inbox and my goal of [current sprint goal], what is the single most important thing I should do today?' Commit to that one thing before anything else steals the morning.
Claude - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 20 minOutput: Empty inbox, 5-8 sent replies, one committed priorityTools: ClaudeTip: Save a reusable Claude Project or system prompt with your voice rules and current goals. Reusing it beats re-explaining yourself every morning.
- 2
9:30 - Deep work: ship the product change
This is the protected block. Open Cursor and describe the change in plain English to the agent: 'Add a usage meter to the settings page that reads from the billing table and warns at 80%.' Let the agent draft the diff, then you review every line before accepting. You are the senior engineer reviewing a junior's PR, not a typist. Two small, reviewed changes a day compounds faster than one big unreviewed one.
Cursor - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 2-3 hrsOutput: 1-2 reviewed, shipped product changesTools: CursorTip: Never accept an agent diff you have not read. The failure mode of solo AI engineering is shipping code you do not understand, then being unable to debug it at 2am.
- 3
12:30 - Sales and user calls, hands off the keyboard
Granola runs quietly in the background of every Zoom, Meet, or in-person call and takes structured notes without a bot joining the room. You stay present and actually listen instead of typing. After the call, Granola gives you a clean summary, decisions, and action items. Ask it to draft the follow-up email and paste it to the prospect within the hour while the conversation is warm.
runs during callsOutput: Notes, decisions, and a drafted follow-up per callTools: GranolaTip: Add your own two-line gut read to each Granola note (are they a real buyer, what is the blocker). The AI captures what was said; only you capture what you felt.
- 4
14:00 - Research an unknown before you decide
Founders lose hours to shallow research. When you hit a decision that needs facts (pricing a new tier, a competitor's positioning, a compliance question), run one focused Perplexity Pro query with the constraint: 'Cite sources, prefer data from the last 12 months, and flag anything you are unsure about.' Read the sources, not just the summary. Fifteen minutes here saves a bad call later.
Perplexity - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 15-30 minOutput: A sourced answer you can act onTools: PerplexityTip: Keep research time-boxed. The goal is a good-enough decision, not a perfect literature review. Set a 30-minute timer.
- 5
15:00 - Support and the busywork pile
Batch the small stuff. Paste support tickets into Claude with your docs and ask for draft replies you approve. Ask it to turn today's Granola notes into a CRM update, draft the weekly investor note from your commits and metrics, and write the changelog entry for what you shipped this morning. This is the work that quietly eats founder time; AI drafts it in minutes and you approve in seconds.
Claude - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 30-45 minOutput: Support cleared, CRM updated, changelog and investor note draftedTools: Claude, Granola - 6
16:30 - Write the thing only you can write
Positioning, a hard hiring or pricing decision, the next feature's spec, a personal note to a churned customer. Use Claude as a thinking partner here, not a ghostwriter: 'Argue the opposite of my plan to price at $49. What am I missing?' The output you keep is yours; AI is the sparring partner that pressure-tests it.
45-60 minOutput: One high-leverage decision or document, finishedTools: Claude, NotionTip: Draft the first 100 words of anything customer-facing yourself. AI can extend your voice but it cannot start it convincingly.
- 7
17:45 - Shut down: log the day, tee up tomorrow
Dump the day into Notion: what shipped, what stalled, tomorrow's one thing. Ask Claude to summarize your day from your notes and commits into a three-line log. This closes the loop so you do not carry the day home in your head, and it gives you a running record of where the time actually went.
10 minOutput: A logged day and a clear priority for tomorrowTools: Claude, Notion
This is not a list of AI tools. It is an actual working day, timestamped, showing where AI does the first draft and where a human still has to decide. The stack is deliberately small: three required tools at $54/mo. Everything else is optional. The point is leverage per dollar, not a wall of subscriptions.
Stack cost breakdown
Public list prices as of July 2026. Optional tools are marked in the notes.
| Tool | Plan | Monthly cost | Notes |
|---|---|---|---|
| Claude | Pro | $20/mo | Required. Inbox triage, drafting, decisions, code review. |
| Cursor | Pro | $20/mo | Required. AI coding agent for shipping product changes. |
| Granola | Business | $14/mo | Required. Live meeting notes + follow-ups on every call. |
| Perplexity | Pro | $20/mo | Optional. Market, competitor, and prospect research. |
| Notion | Plus | $10/mo | Optional. Docs, planning, and the single source of truth. |
| ChatGPT | Plus | $20/mo | Swap for Claude, not additive - run one, not both. |
| Total | $54 - $84/mo($54 required, $84 with optional tools) | ||
Email me this stack as a checklist
Every tool, the plan to pick, and the monthly cost - in your inbox.
Real usage
What people actually run
No usage reports yet - be the first to share what you run. Tell us your real stack, your actual monthly cost, and any tools you swapped.
The prompts that run the day
Copy these into a saved Claude Project so you are not re-typing them every morning.
Morning inbox triage prompt
Here are my unread emails (subject + first line). 1) Sort into reply-now, reply-later, ignore. 2) For reply-now, draft a one-paragraph reply in my voice: direct, warm, no corporate filler, no 'I hope this finds you well'. 3) Flag anything that looks like a real sales opportunity or an unhappy customer. Then tell me the single most important thing to do today given my current goal: [goal].
Note: The 'flag opportunities and unhappy customers' line is what turns triage into judgment, not just sorting.
End-of-call follow-up prompt (from Granola notes)
From these call notes, draft a follow-up email to the prospect. Reference the two specific things they cared about, restate the next step we agreed on with a date, and keep it under 120 words. Add a one-line internal note on whether this is a real buyer and what the blocker is.
Decision sparring prompt
I am about to [decision]. Steelman the opposite choice. List the three strongest reasons I am wrong, the evidence I would need to be right, and the cheapest experiment to test it this week. Be blunt; do not hedge.
Swap options
Drop-in substitutions if a tool does not fit your budget or stack. These trade cost or effort for the recommended setup.
| Swap out | Use instead | When |
|---|---|---|
| Claude | ChatGPT Plus ($20/mo) | You prefer GPT for drafting; both cost $20/mo, run one not both |
| Claude | Google AI Pro / Gemini ($19.99/mo) | You live in Gmail and Google Docs and want AI native to them |
| Cursor | Claude Code or Windsurf | You want a terminal-first agent or a cheaper IDE agent |
| Granola | Otter.ai Pro ($16.99/mo) or Fireflies | You need a bot that joins calls to record rather than a local notepad |
| Notion | Notion Business ($20/mo) | You want full AI agents and Ask Notion across your workspace |
Common Pitfalls
- Collecting tools instead of using three well. A solo founder with 12 subscriptions and no workflow is slower than one with three and a routine.
- Accepting AI code you have not read. When it breaks in production you own a system you cannot debug.
- Letting AI write customer-facing copy end to end. It reads generic and buyers can tell. Edit every word that ships under your name.
- Skipping the shutdown log. Without it you carry the day in your head and start tomorrow reactive instead of intentional.
- Treating AI research as truth. Always open the cited sources before you bet a pricing or legal decision on a summary.
Frequently Asked Questions
Is $54/mo really enough to run a company solo?
Do I need Cursor if I am not technical?
Why Granola instead of a bot that joins the call?
Won't relying on AI make my judgment worse?
Claude or ChatGPT for the thinking-partner role?
How long before this actually saves me time?
How we built this workflow
ToolJunction's editorial team tests each workflow with real accounts and real budgets before publishing. Cost figures reflect public pricing pages as of July 2026. Reply rates, time estimates, and outcome metrics come from our own runs or vetted operator interviews. We update this page when a tool's pricing changes or a step stops working.
Last updated July 7, 2026; prices verified at publication.