Prototyping in a day: the go/no-go decision
Prototyping in a day means building, with the company's team and its real data, a first version of an AI agent and measuring it against how that work is done today. The day ends with a written decision from the process owner: go on, adjust or stop. The name describes the build, never a timescale for results.
Updated: 5 November 2026 · Deibson Silva · Forward Deployed AI Engineer and Founder
What is prototyping in a day?
It is the fourth of the seven stages of the Legathum AI-First Hybridisation Method v3. It comes once the case has been qualified, someone has sat with the person who does the work and the in-house expert's rules are written and approved.
On that day a first agent is built on a single process, with the company's data and with its team in the room. The goal is a prototype the client can touch with their own hands.
One day forces you to choose a single case and to work with the data there is. In return, you find out whether the idea survives contact with real documents.
It also teaches the team something. People who watch the agent get things right and wrong on their own paperwork understand what to ask of it better than people who sat through a presentation.
How is it different from a demo and from a pilot lasting months?
Three ways of trying out an idea. What changes is what you test with and how it ends.
| Demo | Prototype in a day | Pilot lasting months | |
|---|---|---|---|
| With which data | Examples chosen by whoever prepares it | The company's real data, which stays in its systems | Real data, once it is finally ready |
| Who builds it | The supplier, alone | The builder, with the client's team at the keyboard | A technical team, far from the people who do the work |
| How it is measured | It is shown, without measuring | Against the process baseline, with real cases | At the end, if someone fixed beforehand what to compare with |
| How it ends | With applause | With a written decision: go on, adjust or stop | Often, without a clear decision |
| What is at risk | Little, and little is learnt | One day's work | Months of budget |
What has to be ready the week before?
The day goes well or badly depending on what has been prepared. If anything on this list is missing, it is postponed.
The case
- 01A single process, with an owner, a measured baseline and the expert's rules already approved.
- 02The success criteria in writing: what the agent has to do well and how that will be checked.
The data
- 01Read access sorted out. The day is not spent asking for permissions.
- 02Real data that stays in the company's systems.
- 03The expert's real cases, split into two groups: those used to build and those kept sealed for the blind test later on.
The people
- 01The process owner, who decides at the end of the day.
- 02The person who does the work today, because they know the exceptions.
- 03A champion from the team, who will later help the others use it.
How does the day run?
Four blocks. The order matters.
Understand
The team sees what an agent does with an example from their own process. No theory: one of their documents, one of their rules and the result.
Build
The first version is built with the client's team at the keyboard. The builder guides, and every decision of the day is written down.
Measure
The prototype is measured with the open cases, against the process baseline. First with the most capable model available, to find the ceiling of the case before thinking about cost.
Decide
A plan is presented with what worked, what did not and what is missing. The process owner decides.
How is it decided whether the case goes on?
With a one-page memo. It sets out what worked, what did not, the measurements against the baseline and what it would take to continue: data, rules or access.
There are three ways out. Go on, if the prototype meets the success criteria or comes close. Adjust, if the case is worth it but something specific is missing. Stop, if it does not hold up.
The process owner decides. The builder presents the facts and gives a view, but the signature belongs to the company.
Stopping is a valid outcome. One day has been spent and the company knows why the case does not work, which is more than is usually known when a months-long pilot is left without an owner.
If the answer is to go on, the real work starts then: building with tests on real cases, passing security and deploying in phases.
Where you learn it
- Forward Deployed Engineer LicenceWhat it includes, price and terms.
- Forward Deployed Engineer: what it is and what it doesThe craft and the seven stages of the method.
- The certification at the AcademyThe Prototype in a day module and its lab.
- Forward Deployed Engineer certificationProgramme, labs and assessment, week by week.
Honest limits
- Prototype in a day is about the build. The first measured result in production comes later and is measured; nobody promises it.
- A prototype that works in a day has not yet passed the security tests or the Mirror. It does not go into production.
- If the agent is going to talk to people, bear in mind Article 50 of the EU AI Act, which has applied since 2 August 2026: people talking to an AI system must be told it is an AI, unless that is obvious from the context.
- In the programme's lab, the day is rehearsed on a synthetic company. In a real project, the company's data never leaves its systems.
- Today the programme is taught in Spanish: 27 Spanish videos are published at the Academy, and the English videos arrive in November 2026.
- The certification attests a competence, never a job or an income.
What people ask most
01Can you really build an AI agent in a day?
A first version, yes, when the case, the data and the rules have already been prepared. It is a prototype for measuring and deciding. It is not in production and has not passed the security tests.
02Who decides whether the project goes on?
The process owner at the company. The builder presents a one-page memo with what worked, what did not and the measurements against the baseline. The decision is recorded: go on, adjust or stop.
03What happens if the decision is to stop?
The method has done its job. The company has spent one day and knows why that case is not worth pursuing. It can choose another process from its list and repeat the preparation.
04Which data is used on the day?
The company's real data, with read access and without it leaving the company's systems. Part of the real cases is kept sealed and is not touched: it is for the blind test that comes later.
05How is it different from a proof of concept?
It is built with the client's team, measured against a baseline fixed beforehand and ends with a signed decision. A proof of concept usually shows that something is possible; a prototype in a day helps decide whether to do it.
Sources
- Regulation (EU) 2024/1689, the AI Act (EUR-Lex): Articles 50 and 113 · accessed 23 October 2026

Founder of Legathum and iPersonality and Forward Deployed AI Engineer. He leads the SciTechBiz Community from Madrid and examines the FDE Certification defence.
Start from your dashboard
On the licence page you request access; requesting access costs nothing. You then sign the contract and pay the annual licence from your dashboard.
iPersonality Global Institute, S.L. · NIF B27580042 · Madrid · Privacy policy
Prices in euros, exclusive of VAT. In Spain, Spanish VAT is added. In the rest of the EU, VAT is calculated at payment according to your country and VAT number. No Spanish VAT is charged to UK customers.