# The first-build rule: ship something working in week one Source: https://getstartedwithai.net/blog/first-build-rule Site: Get Started With AI Tagline: Build AI automations that actually run your business. Audience: solo professionals, freelancers, consultants and AI implementers in training Contact: hello@getstartedwithai.co.uk This Markdown companion is generated from the prerendered page for AI crawlers. # The first-build rule: ship something working in week one How to start an AI project: pick the smallest repeatable process, build the thinnest working version of it, and ship that version inside week one. After running 120+ AI implementations across 15+ industries, this is the only opening move I trust. Most engagements that miss week one drift to month four with nothing deployed. Most engagements that hit week one keep shipping. This post is the rule, the checklist that defines a real first build, the kinds of work that fit, and what to do with week two. ## Why week one is the credibility window A working build in week one is the credibility moment for the engagement. The client team can see the thing, describe it to their colleagues, and show it to the CEO. Without that early artefact the engagement lives as a deck and a Notion page, both of which look identical to every failed AI project the team has already seen. Cash impact in a UK SMB is the second reason. Every week the build slips is a week of staff time still being spent on the manual version of the work, while a paid engagement runs at full rate. A six-week project that ships in week one starts paying back from week two. A six-week project that ships in week six pays back nothing inside the engagement window. ## What 'working' actually means - Connected to real data, not test data. - Producing real output that the named owner is willing to send, file or act on. - Reviewed by a named human inside the client team, not the consultant. - Scheduled to run on the actual cadence the workflow needs (daily, every Monday, on form submission). - Documented in a one-page handover that the owner could read in a meeting without rehearsing. - Logged in a place the team will look without being prompted, so silent failures get caught. ## First build vs MVP vs prototype These three words get used interchangeably and they mean different things. A prototype is a sketch that proves a shape. An MVP is the smallest product that someone would pay for. A first build is neither. It is the smallest production-shaped version of one repeatable workflow, running on real inputs, producing real outputs, owned by a named person on the client side. The reason to keep the words separate is that prospects who say they want an MVP often want a prototype, and prospects who say they want a prototype often need a first build. The Audit step in the WIRE Method gets the language right before any code is written. ## What can ship in a week, and what cannot About four shapes of work fit cleanly into a week. A Custom GPT that drafts a single recurring email or document on the firm voice. A Zap or Make scenario that catches a form submission, runs it through an LLM, and posts the result to Slack or a CRM. A Notion AI question and answer layer over a tightly scoped knowledge base. A scheduled summarisation job that reads inbound email or meeting transcripts and writes a daily digest. None of these need a model to be trained, infrastructure to be provisioned or a vector store to be tuned. Anything that needs custom model training, multi-step agent orchestration, or a UI for non-technical users is a multi-week build, and the right move is to scope a smaller first build under it. ## The Fastest Win pattern, applied Inside the WIRE Method, picking the first build is the Fastest Win step. The candidate scores well on three things at once: the time it currently takes is high (Time Tax), the cost of a wrong output is low (the work can be reviewed before it lands), and the inputs are already documented somewhere a builder can read. If a candidate scores well on two of the three but poorly on the third, it is not the first build. The full saving is real, but the build will not finish in week one and the engagement will lose its credibility moment. Pick the candidate that wins on all three; the larger savings get built second, on the trust earned from the first build running cleanly. ## What to do in week two Iterate the week-one build before adding a second. The first version of any AI workflow is wrong in small but specific ways: tone drift on certain inputs, edge cases the documentation did not cover, a trigger that fires too often. Week two is for catching those, logging the corrections, and updating the prompt or the trigger where the correction is structural. The fortnight of output review is the next two weeks of the WIRE Method, and it is the part that decides whether the build survives past month three. Skipping it to start a second build is the most common reason a fast first build fails in month four. ## How to recognise a project that will fail week one - The team cannot name the workflow without using marketing words. - The owner will not be in the kickoff meeting. - The data needed for the first build is in three different systems and no one has a sample to hand. - The success metric has not been agreed in writing. - The first proposed build is a chatbot for the public website, which is rarely the right first build for an SMB. ## A worked example from a UK lettings firm A 12-site lettings agent in March 2025. The candidate workflow was the inbound maintenance request triage. Volume was 60 to 90 requests a week across the sites. Each one took an admin 3 to 6 minutes to read, classify by urgency, and route to the right contractor. Week one shipped a Make scenario plus a Claude prompt. The flow read the incoming request and classified it on a four-band urgency scale. The contractor brief was drafted automatically, then both items posted into the agent inbox for human review. Time per request dropped to 30 seconds of review. Week two was a fortnight of output review, with prompt updates after the first 200 cases. The build is still running 14 months later under the property manager who owns it. ## Build your first AI system live with Richard A 60-minute live build with Richard Batt. Bring one repeatable task. Leave with a mapped automation, success metrics and a deploy plan you can test the next morning. Free seat, no upsell. ### Live build session The next live session covers the WIRE Method end to end on a real audience build. [Join the next session](/webinar) ## Questions people ask before they start ### undefined undefined ### undefined undefined ### undefined undefined ### undefined undefined ### undefined undefined ### undefined undefined