From strategy to first project in 30 days
A strategy that does not produce a project in a month becomes a document. The four-week plan: decide, baseline and scope, prove, launch a pilot.
The short answer
A strategy that has not produced a running project within a month is on its way to becoming a document. The distance from a one-page strategy to a pilot on real work is about thirty days, provided the first project is chosen for probability of success rather than size of prize. Week one decides: the project, the owner, the success criteria and the decision date. Week two prepares: baseline, scope, real cases collected and labelled. Week three proves: a proof of concept on those cases and the first gate. Week four launches: pilot setup with a person at the gate, live for a small group. The plan works because each week ends in a decision, not a deliverable.
The four weeks
| Week | Goal | Activities | Ends with |
|---|---|---|---|
| 1. Decide | One project, one owner, one date | Score candidates on the seven criteria; name the internal owner; agree success criteria, decision date and the production owner and budget | A written one-pager: project, owner, criteria, date |
| 2. Prepare | Baseline and cases | Measure the current process; scope the smallest useful version; collect fifty real cases; the doers label them; set up accounts in the company’s name | A baseline and a labelled evaluation set |
| 3. Prove | Can it work on our data? | Build a throwaway proof of concept; run the evaluation set; look at the failures | Gate one: real data good enough, or a data task first |
| 4. Launch | Pilot running | Build the minimal pilot: retrieval, guardrails, a review screen, logging; brief the small group; go live with a person at the gate | A pilot live, with a decision date six weeks out |
What each week must not do
- Week one must not choose the biggest problem. It must choose the most likely win.
- Week two must not skip the baseline to save time. Without it, the pilot cannot be judged.
- Week three must not polish the proof of concept. It is thrown away; its job is the gate.
- Week four must not launch to everyone. A small group, a person at the gate, a log.
After day thirty
The pilot runs for six weeks with the log filling, the gate widening as evidence allows, and the metrics tracked against the baseline. At the decision date, the numbers say go or stop. Go means the production project: integration, monitoring, procedures, training, the monthly budget. Stop means a written lesson and the next candidate. Either way, the business has done AI on real work in under three months, learned what its data and people can support, and built the habit that every later project follows.
What this means for you
Turn the strategy into a project in thirty days: decide, prepare, prove, launch. Choose small, name the owner, take the baseline, label real cases, gate on the data, pilot with a person at the gate. Each week ends in a decision. After a month you are learning from real work rather than discussing a document, and the second project is already visible from the first.
Frequently asked questions
Is 30 days realistic for a business with no AI experience?
For a first project chosen with the seven criteria, with a partner who has done it before and an internal owner who clears time, yes. The month is not about building something impressive; it is about getting a small, measured pilot running so the business learns on real work. Ambition is what makes it unrealistic, not inexperience.
What if the proof of concept in week three fails?
Then you learned in three weeks and a small budget that the data or the task is not ready, which is the cheapest possible way to learn it. Write down why, fix what can be fixed, and choose the next candidate. The plan restarts at week one with better information. That is a success of the method, not a failure of the project.
Who needs to be involved?
The accountable person for a decision in week one and at the gate; the internal owner throughout; two or three people who do the work, for the cases and the pilot; a partner or builder from week two. Five to six people, part-time, for a month. More than that and the calendar becomes the bottleneck.