It's Friday at 4 p.m. and the ops manager is clicking through the thing on the projector. She enters a real order, the queue updates, the confirmation email lands in the shared inbox. The room does the arithmetic out loud: this started on Monday. Nobody says 'impossible,' but two people are visibly thinking it.
A one week software prototype sounds like a consulting fairy tale, right up there with 'the migration will be painless' and 'we'll circle back after the holidays.' But it happens, repeatably, when the scoping is brutal and the builder has shipped this movie before.
The difference between the teams that demo on Friday and the teams that demo a login screen is never typing speed. It's what got cut before Monday morning.
What can you actually ship in one week? One working workflow, for one user group, running on real data: a proof with a pulse, not a product. The trick is a scope negotiated down to a single happy path before the week begins and then defended like a border for five straight days. Do that, and Friday's demo is real. Skip it, and Friday's demo is a Figma file with commitment issues.
Friday at 4 p.m.
The scene above works because of three details that are easy to miss. The user is real (the actual ops manager, not a product owner reading a script). The data is real (yesterday's orders, not lorem ipsum). And the bar is real: not 'is this the final product' but 'does someone's Tuesday get shorter.' A prototype that clears that bar earns the right to a hardening phase. One that doesn't still earns its keep, because it answered a question for the price of a week instead of a quarter.
Skepticism from the room is healthy and you should want it. The fastest way to spot a doomed prototype is a demo where nobody asks 'can I try it with a weird one?'
What a one week software prototype actually means
Let's be precise, because 'prototype' covers everything from a napkin sketch to a beta launch. In the one-week frame it means: one role, one workflow, one happy path, on production-shaped data, deployed somewhere a real user can click it. It is not an MVP (that word has been stretched past meaning), not production-ready, and not an architecture decision. It's an answer to a single question: does this workflow get meaningfully better with software? Every feature that doesn't help answer that question is a tax on Friday.
The Monday-to-Friday plan
Monday: shadow and data grab. Morning with the actual users doing the actual work; afternoon exporting real data and pinning down the happy path in writing. If Monday ends without real data in hand, the week is already in trouble.
Tuesday: walking skeleton. Login to list to detail, the ugliest version possible, deployed by end of day. Deploying on Tuesday feels absurd until you realize it makes every later problem a small problem.
Wednesday: the core workflow. The thing the user actually does all day, working end to end on real records. This is the day's only job.
Thursday: hardening and seed data. Error states for the three ways it will break in the demo, sane seed data for the edge cases, and a rehearsal of the exact clicks Friday will follow.
Friday: demo and handoff. Thirty minutes of real usage by a real user, then a one-page handoff doc: what it does, what it doesn't, what hardening would cost.
The Wednesday rule
If it isn't demoable by Wednesday lunch, cut scope Thursday morning. No exceptions, no 'one more day,' no heroics. The rule exists because week-long builds die of optimism, not of difficulty: every feature added after Wednesday borrows time from Thursday's hardening and Friday's credibility. A smaller demo of something that works beats a bigger demo of something that almost works, every single time.
Scoping like a surgeon
The keep list is short enough to memorize: one role, one happy path, real data, and a keyboard-first UI that respects the user's speed. The cut list is where engagements are saved: auth beyond a simple login, settings pages, exports, mobile layouts, dark mode, and every integration past the first one. A typical negotiation sounds like loss but is actually triage. Say the wishlist starts at three user roles, five integrations, twelve screens, and six reports; the shipped version lands at one role, one integration, five screens, and two reports. Those aren't compromises. That's what focus looks like with numbers on, and it's the same discipline as cutting MVP scope anywhere else.
A framing that works with stakeholders: nothing is deleted, it's parked. The parked list is the roadmap if Friday goes well, which makes the cuts feel like sequencing instead of rejection. For the deeper playbook, the rapid prototyping playbook covers the full toolkit beyond the one-week frame.
What you cannot ship in a week
Multi-role workflows, because each role multiplies the states. Payment flows, because money demands a correctness you can't rush. Anything requiring a compliance or security review, because those calendars don't compress. Native mobile apps, because the stores add a review queue you don't control. And 'just add AI,' which arrives in two flavors: a demo-grade LLM call that will embarrass you with a confident wrong answer, or a real AI feature with evals, which is its own timeline entirely. Promising any of these by Friday is how prototypes get their fairy-tale reputation.
After Friday
If the demo lands, the standard path is two more weeks of hardening: real auth, error handling, the second integration, and the parked features that earned their way back. Throwing the prototype away is rarer than folklore suggests; maybe one in five needs a rebuild, usually because the demo revealed the workflow itself was misunderstood. The build-on-versus-rebuild call comes down to one test: did Friday prove the shape right? If yes, harden. If the shape was wrong, be glad you learned it in five days.
Either way, you end the week knowing something a six-month discovery phase would still be guessing at. That's the whole trick behind shipping in days instead of months: small bets, real users, fast Fridays. Pick the workflow your ops lead complains about most, and put it on the calendar for Monday.