tellaprompt

AI prompts for founders

Last updated

A founder's AI leverage is in the documents that run the company: the pitch that survives hostile questions, the investor update that takes 30 minutes instead of 3 hours, the decision memo that forces clarity, the scorecard that fixes hiring before the job ad exists. Five templates, all built to be argued with — not admired.

The recurring pattern: you supply the facts and the stakes; the AI supplies structure and adversarial pressure. Vision stays yours.

How to prompt as a founder

  • Numbers before narrative: MRR, churn, runway, pipeline — paste them. Story built on stated numbers converts; story built on adjectives gets follow-up questions.
  • Make the AI the skeptic, not the cheerleader — you have enough optimism; you're paying for the pushback.
  • One decision per memo: cramming three calls into one document is how none of them get made.
  • Confidentiality: cap tables, term sheets and unreleased metrics belong in enterprise instances or redacted form.

The five templates

Pitch narrative, pressure-tested

The deck exists; the story wobbles. You want the narrative tight and the hostile questions rehearsed before the partner meeting.

Act as a skeptical [STAGE, E.G. SEED] investor with expertise in [SECTOR]. Below is my pitch narrative and key numbers.

Round 1 — attack: the 5 hardest questions you'd ask, ranked by how likely they are to kill the deal. For each: why it matters to an investor, and what a convincing answer must contain (structure, not my content).
Round 2 — after I answer each in the chat: critique my answer bluntly, then tighten it to max 3 sentences keeping MY facts.
No compliments. If a claim lacks a number, flag it as [UNBACKED].

Narrative + numbers:
[PASTE: PROBLEM, SOLUTION, TRACTION METRICS, MARKET LOGIC, TEAM, ASK]
Investor update in 30 minutes

Monthly update due. The discipline matters more than the prose — and the asks are the point.

Draft my monthly investor update from the raw notes below.

Format, fixed: 1) TL;DR — 3 lines, leading with the number that moved most; 2) metrics table vs. last month and vs. plan [PASTE: MRR, GROWTH, BURN, RUNWAY, YOUR 1–2 CORE METRICS]; 3) wins — max 3, each with its number; 4) misses & lessons — at least 1, honest, with what changed because of it; 5) asks — specific and answerable by hitting reply ("intro to X", "feedback on Y"); 6) one-line personal note.
Max 400 words. No hype vocabulary ("thrilled", "crushing"), no metric without comparison.

Raw notes:
[PASTE BULLETS, MESSY IS FINE]

Why this works: The misses-with-lessons section is what builds investor trust over quarters, and reply-able asks are what makes updates generate value instead of just reporting it.

Output formatFact densityAnti-AI style

Decision memo (one decision, forced clarity)

A real fork: pricing model, market, key hire, build-vs-buy. Writing it down is the decision technology.

Help me write a one-page decision memo on: [THE DECISION, PHRASED AS A QUESTION].

First: ask me up to 5 questions to extract what you need (options considered, constraints, deadline, what "wrong" costs, gut lean).
Then produce: 1) the decision question in one sentence; 2) context in 3 bullets; 3) options table — option, strongest argument FOR (steelmanned), strongest argument AGAINST, reversibility (one-way/two-way door), cost of delay; 4) recommendation with reasoning; 5) "What would change my mind" — 2 observable triggers; 6) decision deadline.

Push back if my reasoning smuggles in an assumption; name it.

Why this works: Interview-first extracts what's actually in your head; steelmanned options and change-my-mind triggers convert a vibe into a decision that can be revisited on evidence.

Ask-first loopOutput formatBrutal critic

Job scorecard before job ad

Hire #7 matters too much for a copied job ad. Define what winning looks like first; the ad falls out of it.

Build a hiring scorecard for: [ROLE] at my [STAGE, TEAM SIZE] company.

Context: this role exists because [THE PAIN]. In 12 months, this hire has succeeded if: [YOUR ROUGH IDEAS — MESSY OK]. Constraints: [BUDGET BAND, LOCATION/REMOTE, WHO THEY REPORT TO].

Deliver: 1) mission of the role in one sentence; 2) 4–6 outcomes for year one, each measurable ("reduce X from A to B by Q2"); 3) competencies split MUST vs. NICE (max 4 must); 4) 6 interview questions mapped to the outcomes — behavioral, no hypotheticals; 5) the job ad, 200 words, written from the outcomes — no "rockstar", no laundry list.
Ask me 3 questions first if the outcomes are too vague to quantify.
Customer discovery synthesis (what did we actually hear?)

Fifteen discovery calls done. Now: what did they say — versus what you wanted to hear?

Below are my notes/transcripts from [N] customer discovery calls about [PROBLEM SPACE].

Synthesize ruthlessly: 1) problems mentioned unprompted, ranked by frequency, each with 2 verbatim quotes; 2) current workarounds people described (what they DO, not what they say they'd do); 3) willingness-to-pay signals — actual quotes only, labeled strong/weak; 4) evidence AGAINST my hypothesis [STATE YOUR HYPOTHESIS] — this section may not be empty unless you show why; 5) the 3 riskiest assumptions still untested, each with the cheapest next test.

Rules: their words over your paraphrase; "nice to have" tone flagged as such; no invented enthusiasm.

Notes:
[PASTE]

Why this works: The evidence-against section is mandatory precisely because founders (and models) drift toward confirmation; verbatim quotes and do-vs-say separation are the discovery discipline that survives investor diligence.

Voice of customerBrutal criticFact density

Frequently asked questions

Which founder tasks should I NOT delegate to AI?

Vision, taste and stakes: the actual decision (the memo forces it back to you), the honest relationship with investors, firing and other conversations that define culture. AI compresses the documents around judgment; the judgment is the job.

Can I use these with sensitive company data?

Metrics and plans are competitively sensitive — use enterprise instances with no-training guarantees, or redact (the templates run fine on rounded numbers and code names). Term sheets and cap tables: lawyer first, AI never on consumer tiers.

How is the pitch template different from just practicing?

It structures the practice: ranked kill-questions, an answer-content checklist per question, and a tighten-to-three-sentences loop. Founders who rehearse against written hostile questions consistently report calmer partner meetings — the surprise is gone.

What's the fastest habit to adopt from this page?

The investor update format, monthly, even before you have investors — it forces metric discipline and the misses-with-lessons reflex early, and gives you a dated log of the company's actual story.