The deck says the customer portal will have three roles and a status board. The prospect nods. Then their operations lead asks what happens when a customer uploads the wrong file, and the deck has no slide for that.
A working prototype does. Click the upload, pick the wrong file type, and the prospect watches the app reject it and say why. That one moment does more for the pitch than the ten slides before it, because the room stops evaluating your promise and starts evaluating your work. On Bolt.new, an agency can build that prototype from the brief in an afternoon, on the prospect's brand, at a URL you can open in the meeting. Here is the fastest route from brief to a prototype you'd show a client, and the discipline that keeps it from costing you the scope.
Why a running prototype wins the room
Agencies fight harder for new business than they did five years ago. Client retention after an agency review fell to 21% in 2025, the lowest in eight years, and independent agencies took 13% of the $37.4 billion in billings reviewed (COMvergence, March 2026). Accounts are moving, and the pitch is where they move. Those figures cover media billings, so treat them as the direction of the market rather than a creative-agency statistic.
The research on prototypes is older and more careful than the vendor claims around it. Low- and high-fidelity prototypes surface the same usability problems, but people behave more like real users in front of a high-fidelity one (Nielsen Norman Group; Walker, Takayama and Landay). The historical case against high fidelity was cost: it took too long to build for a conversation that might not convert. That case is the one AI builders removed, which is our reading of the evidence, not a finding in it. No published study shows that AI-built prototypes close more pitches, and you should distrust any number that claims one.
What you can say is narrower and true. A prospect who clicks through their own workflow, on their own brand, has already imagined using it. A prospect who watched slides has imagined you describing it.
The fastest way to turn a client brief into a working prototype to present
- Pull one workflow out of the brief, not the whole engagement. The thing the prospect's customers will do most, with the two or three roles involved and the moment that goes wrong. If the brief is a portal, that's the login, the status view, and the upload.
- Write that workflow as a paragraph. Roles, screens, statuses, and the failure case, in plain sentences. That paragraph is your prompt. The sharper you write it, the closer the first build lands, so spend twenty minutes on it rather than five.
- Generate, then walk it as the prospect's customer. Paste the paragraph into Bolt and let the agent build the app: login, roles, data model, screens. Then use it the way a customer would, including the wrong-file upload. Where it breaks, describe the fix and let Bolt make it. A pitch prototype should survive the one edge case you know the room will ask about.
- Put their brand on it. Their colors, their type, their logo in the corner. If the prospect has a design system, bring it in and let the screens inherit it. An unbranded prototype reads as your template; a branded one reads as their product.
- Publish to a URL and rehearse two minutes. Deploy it through Bolt Cloud so the prototype has a link, not a laptop. Rehearse a two-minute walkthrough that ends on the failure case. Two minutes is long enough to prove the work exists and short enough that nobody asks for the rest of it.
An afternoon covers all five, because the build in step three is measured in minutes. The rest is thinking, which is the part agencies were already good at.
What to show in the room, and what to hold back
Show the one workflow, end to end, live. Show the edge case, and if the prospect asks about a second one, fix it in the meeting. It tells them more about working with you than the rest of the pitch does.
Hold back everything else. The pitch prototype is where scope creep is born; every extra screen you demo is a screen the prospect will assume is included. Tell the room this is a scoping prototype, built to agree on what the real product does, and that the engagement is the build. VISION-LINK AI HUB, a Nigerian startup, did a version of this with funders: a working caregiver dashboard built on Bolt.new in a few hours, shown as a proof of concept in place of a deck (customer story). The prototype opened the conversation. The company still had the hardware to build.
Then convert it. The prototype becomes the appendix to the statement of work: here is what we agreed the product does, screen by screen. Price the full build against that, including the parts the prototype skipped, which brings you to the trap.
The trap: pitching the prototype as the product
A prototype built in an afternoon is real software, and that's the problem. It looks finished enough that a prospect will ask why the engagement costs what it costs. Answer before they ask. The prototype proves the workflow. In the engagement you harden the authentication, connect the real data, handle the customers who don't behave, and hand over something the client can run without you. That's the work, and it's the work that keeps running after launch (what agency work actually lasts).
How that build goes is the client portal walkthrough. How it ends is the handoff guide.
The prototype is the right tool for exactly one job. Lenny Rachitsky and Noam Segal's survey of 1,750 product people puts the tool fit in one line: Bolt "for quick flexible prototypes" (Lenny's Newsletter, 2025). That's the pitch use. Build the prototype fast, let it win the room, and sell the engagement that turns it into the product.
The next brief in your inbox has one workflow in it that the prospect cares about most. Build that one before the meeting. See how agencies use Bolt for client work.
