Examples

One tool built for someone else's need, four tasks small businesses asked us about, and the tools I build for my own work.

01The example

SIMAC Prep: the full story

Her words

“Thirty years in sales, and for the past few years a sales coach for future entrepreneurs. What I want to pass on is the practical side of the trade: listening, reacting to what the client says, building the soft skills that will make them comfortable in the field. But my sessions slide into theory, and I lose the role plays, where the real transmission happens. With AI, I want to hand off the theory and focus on my know-how: the human side.”

Camille C., sales coach. Name changed.

Who it is for

Trainers who want their sessions to go to practice. And early entrepreneurs who do not know whether their idea is mature, what “mature” means, or who their competitors are.

What it solves

Paste your offer: the tool rates its maturity from 1 to 5, names what is missing and builds a competitor grid, before it is too late. The figure is the tool's assessment, not an audit. Then the blank page: what do I say, in which order? Paste who you are meeting: it profiles each person, asks the one question that matters, what is the objective of this meeting, writes the script and the follow-up email, and keeps a clean list of your client actions.

The screenshots show an invented yoga business meeting an invented client. Demo data throughout.

SIMAC Prep rating an offer 4 out of 5 and listing what the offer sheet does not say yet — Exhibit 01The offer rated by the tool, with what is missing. Demo data.

The offer sheet filled by the tool from the description: fourteen fields filled from the description — Exhibit 02The offer sheet, filled from your description: 14 of 14 fields. Demo data.

The client sheet separating sourced facts from hypotheses to check — Exhibit 03The client sheet: sourced facts apart from hypotheses. Demo data.

The meeting script with one advantage written for each person in the room — Exhibit 04The script, one argument per person. Demo data.

The follow-up step with the result, the decision maker and three dated actions — Exhibit 05Three dated actions, the follow-up email is built around them. Demo data.

How it was built

She described the need and handed over her training deck. The tool was built with AI and the method we teach, and its output compared with a meeting she had prepared by hand. Six hours of my work, then two more of tests and adjustments with her. It is published with GitHub Pages, the address is a DNS redirect at OVH, there is no server to pay for. The app is free and open source. It runs in the browser with the user's own AI key, free or paid, and any charges from that provider are the user's. What the user types is sent to that provider; company and web searches go to the official French registries and to Jina; keys and meeting data stay in the browser; an email address is asked once at the start and confirmed by email. Free keys are for public or sample data. Business data belongs on a paid plan at the provider. Provided as is, without warranty.

Open SIMAC Prep (in French)

02Where to start

Pick one small, everyday task. Two of these were built by their owners after a one-to-one session. The other two are typical first projects, not requests we heard.

A shop page for baby clothes and gifts

Built by the owner, September 2026

The need: They sold by phone and through friends. They wanted a page where customers see the collection and gift ideas, order and pay later, and write to the shop.
Built: The page, by the owner, in one week, after a one-to-one session, with ChatGPT.
Tested: By the owner, before real use.
Result: In use since September. Orders and requests arrive on the site; questions go straight to the shop by email.
Stays out: Online payment, stock and delivery; an existing shop platform fits those better. No AI assistant on the page.
Data: Orders and messages land on the site and in the shop's email.

Payment follow-up for a physiotherapist

Built by the owner, September 2026

The need: The bank shows amounts, not who paid for which sessions. They wanted a phone app with a record of who paid, when, how much and for how many sessions.
Built: An iPhone app, by the owner, in three days, after a one-to-one session, with Claude.
Tested: By the owner, before real use.
Result: In use since September. Payments are matched directly in the app.
Stays out: Invoicing, accounting and the health-insurance side.
Data: Records stay on the phone and in a spreadsheet on their cloud drive. The app calls no AI while running, so health data never reaches a public AI chat.

A recurring report

A typical first project, not a request we heard

Today: Every week someone exports the figures and rebuilds the same report by hand.
The tool would: Read this week's export, produce the usual report in the usual layout, flag what moved the most.
Stays out: A person still checks the figures. No automatic sending.
Data: The export stays on the company computer. Once built, the tool may not need AI to run.
First step: One past export and one past report, side by side, on the free call.

A request tracker

A typical first project, not a request we heard

Today: Requests arrive by email, phone and messages. The lost one shows up when the client calls back.
The tool would: One shared list, each request with an owner and a status, readable from the office and from a phone.
Stays out: Beyond a handful of people or a large volume, an existing tool fits better. We say which.
Data: Client names stay on your own account, never in a public AI chat.
First step: The last twenty requests and where each one got stuck.

03Built this way

Real tools, built with AI and the same method we teach.

Book production, organized

I publish activity books: texts, puzzles, illustrations, layout. In about a week I built a local dashboard with AI. It shows what comes next, keeps every chapter version, and asks before spending AI credits.

My own tool, not a client project. It runs on my computer; its costs are AI usage and my maintenance time.

Side-by-side comparison of the current chapter and an earlier saved version, with a review checkbox before approval — Exhibit 06Comparing a saved version before restoring it. Nothing changes until I approve.

AVApmo, a complete product

A project-management assistant for small businesses. Send a client email, get a clear plan; nothing changes without your confirmation. Built and maintained with the Agentic Development Protocol, the method we teach.

Visit avapmo.com How it was built

AVApmo turning a client email into a project canvas with purpose, scope and deliverables — Exhibit 07Email to plan (demo data)

AVApmo showing a change request with its schedule impact, actions and decisions, and a delay explained — Exhibit 08What needs you (demo data)

AVApmo portfolio view where the assistant proposes and the user decides — Exhibit 09AI proposes, you decide (demo data)

More things you can build with AI

Learning games

Maths and reading games in the browser, including letter games for dyslexic learners.

An automated publication

Articles prepared and published by AI assistants, following a process set in advance.

dltrevolution.io

Watering that follows the weather

A small device waters according to how damp the soil is; an AI installed on site takes the forecast into account.

Ready to talk about your first agent?

Describe your need

Already know what you need? Book a call (opens a new tab). Or write to contact@avapmo.com.