AI for founders: 7 jobs to hand over, a lean stack and a 30-day plan

A founder does six jobs before lunch. AI will not make the decisions, but it can do the research, first drafts and build work that used to need a hire, so the hours go into customers and choices.

The short answer

AI for founders in November 2026 works best as a lean three-tool stack: Perplexity for market questions with sources, Claude from Anthropic for strategy memos and writing (Claude Opus 5.5 on paid plans since 22 September 2026), and Notion as the place plans and notes live. Add Lovable to build a landing page or first app by chatting. Hand over research, drafts and prototypes. Keep customer conversations, hiring, fundraising relationships and the final call on strategy yourself.

Research with sourcesPerplexity (Deep Research on paid plans)
Thinking partner and writingClaude: Opus 5.5 on Pro and above, Sonnet and Haiku on Free
Home for docs and plansNotion (full Notion Agent on Business and Enterprise)
Build a page or appLovable: Free to test, Pro for real projects
Watch outFree and Pro plans of Claude, Perplexity and Lovable can train on your content unless you opt out

How founders use AI in 2026

Founders use AI differently from teams in large companies. There is no IT department choosing tools and no specialist for each job, so the question is which few tools cover the most ground. The answer we see most in founders' shared setups, and the one on our founders stack, is Perplexity for research, Notion for notes and plans, and Claude as a thinking partner for strategy and writing.

The tools are much more capable than a year ago. Claude now merges chat with Claude Cowork, so one conversation can take on a whole task and hand back an editable document, spreadsheet or presentation (rolling out gradually, starting with Pro and Max). Perplexity Computer runs multi-step research and builds documents in a cloud sandbox on paid plans, paid with credits. Lovable builds and publishes full-stack web apps from a chat. Our best AI for small business page covers more options.

The 7 founder jobs worth handing to AI

1. Market and competitor research (Perplexity)

Market size claims, competitor pricing, recent funding and regulation questions all need sources. Perplexity lists numbered sources in every answer, and Deep Research on paid plans goes further. The "A competitor report for the board" stack on /stacks adds Gemini Notebook to hold the sources and Claude to write the one-pager.

2. Strategy memos and decision write-ups (Claude)

Write down the decision, the options and what you believe, then ask Claude to argue against your preferred option as hard as it can. It is a cheap way to find the hole before an investor or co-founder does.

3. Investor updates and pitch narrative (Claude)

Give Claude your metrics, wins, misses and asks, and one update you were proud of. Ask for a draft in that format. Claude Slides, available on every plan including Free, exports to PowerPoint, PDF and Google Slides when you need a deck.

4. A landing page to test an idea (Lovable)

Claude writes the copy, Lovable builds and publishes the page. Lovable's Plan mode writes a reviewable plan and never edits code, so you can agree the structure first. On Free, apps publish publicly with an Edit with Lovable badge and you cannot use a custom domain, so test on Free and move to Pro for anything real. This is the "A landing page in an afternoon" stack.

5. Customer interview synthesis (Gemini Notebook)

Upload interview transcripts to one notebook and ask for the problems people describe, in their words, with the interview each quote came from. Answers in Gemini Notebook are grounded only in your sources, with citations. Then go back and read the transcripts behind the strongest themes.

6. Your operating system in Notion

Keep plans, decisions and meeting notes in Notion. Notion Agent can create and edit pages and databases, and Custom Agents run on triggers and schedules, but both need the Business or Enterprise plan; Free and Plus have only a limited AI trial. Notion says customer data is not used to train models by default.

7. First drafts of everything else (ChatGPT or Claude)

Job descriptions, contracts to review before your lawyer sees them, customer emails, support macros. Use whichever assistant you already pay for, and give it real context: who it is for and one example of what good looks like.

What founders should keep human

  • Talking to customers. AI can summarise interviews. It cannot have them, and the insight is often in what people did not say.
  • The strategy call. AI can list options and argue each side. You choose and own the consequences.
  • Hiring and co-founder decisions. Use AI for the job description and interview questions, never for judging the person.
  • Investor and partner relationships. Drafts are fine; the relationship is yours.
  • Legal and financial sign-off. AI is useful for a first read of a term sheet, but a qualified lawyer and accountant make the call.

A lean AI stack for founders

JobToolStart onWhy
ResearchPerplexityFree, Pro for Deep ResearchSources on every answer
Thinking and writingClaudeProOpus 5.5, Research, Claude Code and Cowork tasks from Pro upwards
Notes, plans, docsNotionFree, Business when you want Notion AgentOne home for decisions; no training by default
Pages and prototypesLovableFree to test, Pro for real projectsCode syncs to GitHub, GitLab or Bitbucket
Source-grounded readingGemini NotebookStandard (free)100 notebooks, 50 sources each

If you write code yourself, add Claude Code (included from Claude Pro) or Cursor; see AI for software engineers. Check prices on each maker's pricing page or our compare tool.

A 30-day AI plan for a founder

  1. Days 1 to 3: settings. Switch off training on personal plans: Claude, Settings, Privacy, Help Improve our AI models; Perplexity, Account settings, Preferences, AI data retention; Lovable, Account settings, Preferences, AI model training. Your roadmap and numbers are confidential.
  2. Days 4 to 10: research week. Rebuild your competitor and market picture in Perplexity, with every claim sourced. Save the sources in a Gemini Notebook.
  3. Days 11 to 17: decision week. Write your three biggest open decisions as memos and have Claude argue against each. Store them in Notion.
  4. Days 18 to 24: build week. Ship one landing page or prototype with Lovable and put it in front of ten real people.
  5. Days 25 to 30: make it routine. Save your best prompts, set one recurring job (a weekly investor update draft, a newsletter digest) and drop any tool you did not open twice.

The risks for founders, and how to manage them

  • Training on your secrets. Claude Free, Pro, Max and Ultra, Perplexity Free, Pro and Max, and Lovable Free and Pro can use your content for training unless you opt out. Fix it on day one, or use team plans where it is off by default.
  • Confident wrong numbers. Market sizes are the classic hallucination. Never put an AI figure in a deck without the original source.
  • Credits running out mid-task. Perplexity Computer, Lovable and Notion's agents run on credits. Watch usage before a launch week.
  • Everyone's pitch sounding the same. Investors read many AI-polished decks. Your own numbers, customer quotes and plain sentences stand out.
  • Public by default. Lovable's free apps publish publicly. Do not put customer data in a test app.

More prompts are in prompts for founders and prompts for decision-making. To choose between the two main assistants, read ChatGPT vs Claude.

Prompts for the 7 founder jobs

One prompt for each job above. Replace the square brackets.

Prompts to copy

1. A sourced market picture

I am building [product] for [customer]. Find: the main competitors and their pricing models from their own pages, funding news from the last 12 months, and any regulation that affects this market in [country]. One line each with a source. Mark estimates and anything you could not confirm.

You get: A market picture you can defend line by line

2. Argue against my decision

Here is a decision I am about to make: [decision], with the options I considered and why I prefer [option]: [notes]. Argue against my preferred option as strongly as you can. Then tell me what evidence would change your mind, and the cheapest way to get it this week.

You get: The strongest counter-argument and a test to settle it

3. The monthly investor update

Write this month's investor update in the format of the example below. Metrics: [paste]. Wins: [list]. Misses and what we learned: [list]. Asks: [list]. Plain and honest, under 400 words, misses not hidden. Example: [paste a past update]

You get: An update in your format that investors trust

4. Landing page copy for a test

Write landing page copy to test demand for [product] with [audience]. Sections: headline, subheadline, three benefits with proof we actually have ([list]), how it works in three steps, one call to action to [join the waitlist]. No claims we cannot back up.

You get: Copy ready to paste into Lovable

5. Themes from customer interviews (Gemini Notebook)

From these interview transcripts only, list the problems customers describe, ranked by how many interviews mention each. For every problem, give two exact quotes and the interview they came from. Then list anything that contradicts our assumption that [assumption].

You get: Customer problems in their words, with the source

6. A weekly review page in Notion

Create a weekly review page from this week's meeting notes and tasks: a list of decisions made, open decisions with owners, the top three priorities for next week and anything that slipped. Link each item to its source page.

You get: A review page that keeps the team aligned

7. A job description that attracts the right people

Write a job description for our first [role]. What they will own in the first six months: [list]. What we can offer: [list]. Be honest about the stage, the risk and the pace. Under 450 words, no buzzwords.

You get: An honest job description that filters for fit

Start the Perplexity lessons Start the Claude lessons