# The 3-minute investor demo: a script for AI products

*A second-by-second structure: hook, why now, one live path to the magic moment, proof, business, ask. With prep checklist and Q&A.*

An investor demo is not a product tour. In three minutes you have to make one person feel one problem, see the product solve it once, and believe it can become a business. Everything else — settings, the admin panel, your architecture — is noise. This script gives you the structure with timings, what to say and what to show, a worked example (CallNote, a fictional product — numbers are illustrative), and the preparation that keeps a live AI demo from going sideways.

**Time:** 1 hour to write, 5 rehearsals · **Format:** 6 segments + prep checklist · **Online version:** https://stepikin.com/templates/investor-demo-script/

---

## Three rules

1. **One persona, one path.** Pick the single journey that ends in your "magic moment" and cut everything else.
2. **Show, then tell.** Each claim should land on screen within seconds of you saying it.
3. **Rehearse with a timer.** Three minutes means three minutes. Most first drafts run six.

## The script, second by second

### 0:00–0:20 · Hook: the problem as a moment

**Goal:** Make them picture one person at one painful moment, with one number.

**Say:** It's 6 pm. [Persona] has just finished [situation] and now has to [painful task] — [number] minutes, every day.

**Show:** Nothing yet, or the "before" screen: the messy doc, the empty CRM, the spreadsheet.

**Example:** "It's 6 pm. Maya, an account executive, just finished her sixth sales call today. Now she has to type all six into HubSpot — 40 minutes she doesn't have. So she skips it, and her manager forecasts on stale data."

### 0:20–0:40 · Who and why now

**Goal:** Size the pain and explain why this is possible or urgent now, not five years ago.

**Say:** [Number] people do this. It costs them [money/time]. Until [shift], it couldn't be automated reliably — now it can.

**Show:** Still the "before" screen, or one simple slide with two numbers.

**Example:** "Every B2B sales team has this problem — our design partners measured about 6 hours a week of CRM admin per rep. Transcription is solved — but turning a conversation into trustworthy, structured CRM fields only became reliable and cheap this year."

### 0:40–2:10 · Live demo: one path to the magic moment

**Goal:** Before → the AI doing visible work → the user in control → the result landing where they already work.

**Say:** Narrate what the user sees and decides, not what the system does internally. Name the magic moment out loud when it happens.

**Show:** Realistic input (not "test 123"). The AI working with visible progress. The user reviewing and approving. The result in the tool they already use.

**Example:** "Maya ends the call. Thirty seconds later, CallNote has a proposed update: next step, new close date, two objections — each with the exact quote from the call. She fixes one field, approves, and… here it is in HubSpot. That's the moment: the CRM is updated before she's even opened it."

### 2:10–2:35 · Proof

**Goal:** Show it works outside your laptop: usage, retention, a quote, or an eval number.

**Say:** One or two hard numbers and who they come from.

**Show:** A single proof slide or an analytics screen.

**Example:** "Three design-partner teams, 31 reps, 4 weeks: 68% of calls updated within an hour, 82% of suggested fields approved without edits. Two teams asked to pay before we'd sent a price."

### 2:35–2:50 · Business

**Goal:** Who pays, how much, and why the margin holds as usage grows.

**Say:** Buyer, price, cost to serve, expansion path.

**Show:** Pricing line or a one-line unit economics slide.

**Example:** "Sales managers buy it: $39 per rep per month. AI costs us about $4 per rep per month at current usage. We land with one team and expand across the sales org."

### 2:50–3:00 · Ask and close

**Goal:** State what you want from this meeting and repeat the one-liner.

**Say:** We're raising [amount] to [milestone]. [One-sentence product line again.]

**Show:** Final slide: logo, one-liner, contact.

**Example:** "We're raising $1.5M to get to 50 paying teams in 12 months. CallNote: your CRM, updated before you open it."

## Your script

| Time | Segment | What you say | What they see |
| --- | --- | --- | --- |
| 0:00–0:20 | Hook |   |   |
| 0:20–0:40 | Who and why now |   |   |
| 0:40–2:10 | Live demo |   |   |
| 2:10–2:35 | Proof |   |   |
| 2:35–2:50 | Business |   |   |
| 2:50–3:00 | Ask and close |   |   |

## Prep checklist for a live AI demo

- [ ] **A recorded backup video of the exact path** `BLOCKER` — Same narration. If anything breaks, switch without apologizing.
- [ ] **A seeded demo account with realistic data** `BLOCKER` — Real-looking names and content, no "Lorem ipsum", no other customers' data.
- [ ] **A deterministic path for the AI step** `BLOCKER` — Pre-tested input, low temperature, and a cached response ready if the provider is slow.
- [ ] **No login, onboarding or settings on screen** — Start already logged in, on the "before" screen.
- [ ] **Screen ready for sharing** — Notifications off, bookmarks bar hidden, 125% zoom, one browser tab, dark mode off unless it's your product's default.
- [ ] **No live typing of long prompts** — Paste or use a prepared input. Typing kills 20 seconds.
- [ ] **A sentence ready for the weird output** — "That's exactly why the user approves every change" — and move on.
- [ ] **Five timed rehearsals, one in front of a stranger** — Cut whatever makes you run over.

## Questions investors ask about AI products — prepare one-sentence answers

1. **"What happens when OpenAI (or Google, or Microsoft) ships this?"** — your answer is workflow, data, distribution or trust, not the model.
2. **"What does each task cost you, and what happens at 10× usage?"** — cost per task, margin, and your plan to keep it.
3. **"How do you know the output is good?"** — your eval set, acceptance rate and how a bad answer is caught.
4. **"Whose data is it, and can you use it?"** — permissions, privacy posture, what improves over time.
5. **"Why you?"** — the unfair insight or access your team has.

## Common mistakes

- Touring every feature instead of finishing one path.
- Explaining the architecture, the model or the prompt before showing the result.
- Starting on a login screen, an empty dashboard or settings.
- Apologizing for the UI or saying "this part is still rough".
- Ending on a feature instead of the ask.

---

Made by [Stepikin Studio](https://stepikin.com/) — design and engineering for AI products, MVPs shipped in weeks.
Want this done for you? [Get an estimate](https://stepikin.com/estimate/) · [Book a call](https://stepikin.com/book/) · More templates: https://stepikin.com/templates/