Product demo script prompt builder
The demos that close are not feature tours — they are short stories where the prospect is the hero, their problem is the conflict, and your product is the resolution. The fastest way to lose a demo is to show everything; the fastest way to win one is to show only what answers a pain the prospect already feels. This builder turns your features, the prospect’s pains, and your flow into an LLM prompt that writes that story-driven script.
How it works
You list the features you plan to show and the specific pain points the prospect has, then describe your demo order and the time you have. The builder assembles a prompt that pairs each feature to a pain, opens on the prospect’s problem rather than your product, writes a talk track and transition for each section, paces the whole thing to your time budget, marks the key “aha” moments, and builds toward a single clear close. It instructs the model to drop any feature that does not map to a stated pain.
The anatomy of a demo that converts
Understanding the structure helps you give the builder better inputs — and helps you adapt the script the model produces.
The first two minutes should be about the prospect, not your product. Start with a crisp statement of the problem they have: “Most teams in your position spend X hours a week doing Y manually, and the biggest risk is Z.” This tells the prospect you understand their situation, earns attention, and sets up everything else as the solution to a problem they already feel. Do not start with company history or a feature list.
The feature-pain pairing is the core structure the builder enforces. For each feature you plan to show, the script says: “The reason teams struggle with [pain] is [root cause] — let me show you how [feature] handles that.” Each feature is introduced as the answer to a stated problem, not as a capability to admire. Features without a matching pain get cut — they are selling to no one and dilute the ones that do matter.
The “aha” moment is the screen or interaction that earns the sale. In most demos there is one thing that makes the prospect lean forward, and everything else is context-setting for it. The prompt asks you to identify it so the script can build toward it rather than presenting it at random among ten other features. At the aha moment, slow down, let it land, and do not rush to the next screen.
The close is a single, concrete next step, not “any questions?” — which signals that the demo is over without asking for anything. “Shall we run a two-week pilot with your own data?” or “Want me to send you a proposal by Thursday?” are closes. “Let me know if you have any questions” is not. The prompt builds the script toward one specific ask and keeps it proportionate to the stage of the conversation.
Common demo mistakes this builder helps avoid
- Starting with the product. Opening on “so, let me show you our dashboard” loses the prospect before you have earned their attention.
- Showing too many features. Every feature you show that is not relevant to the prospect’s pain dilutes the ones that are. The prompt drops features with no matching pain.
- Pacing poorly. Spending eight of thirty minutes on a detail the prospect does not care about and rushing the aha moment is one of the most common demo failures. The script is paced to the time you have.
- Trailing off instead of closing. Demos that end with “so… yeah, that’s the product” leave the prospect with nothing to do. The generated script ends with one specific, proportionate ask.
Tips and examples
- Open on their problem. Spend the first minute on the pain, not the logo. The prompt structures the script this way.
- Map every feature to a pain. If a feature has no matching pain, cut it. The prompt enforces this so the demo stays tight and relevant.
- Mark the aha moment. There is usually one screen that earns the deal — the prompt flags it so you slow down and let it land.
- Close on one step. “Shall we set up a two-week pilot?” beats “any questions?” — the prompt builds toward a single, concrete ask.
- Paste into your own LLM. The builder produces the prompt; run it in ChatGPT, Claude, or your preferred model to get the script.