Product
Turning scattered needs into coherent modules, priorities and journeys.
Dossier · Project 01 / 13 · Personal product · 2026
Martelazo, my local command centre.
A personal application that connects the scattered information of my working day to concrete actions, without exposing my data to a public service.
Working prototype · local use
| Category | Marketing tools · applied AI |
|---|---|
| Period | 2026 |
| Role | Design, prioritisation, assisted build and testing |
| Stack | Next.js, React, TypeScript, SQLite, Google APIs, AI |
| Principle | Local-first, data stays on my machine |
| Demonstration | Entirely fictitious data |
This project did not start from a wish to build an application. It started from a concrete problem: tasks, documents, clients, content and decisions lived in too many separate places.
01 · Problem
Seeing the whole without losing the detail
I wanted a single interface to start my day, follow my priorities and find a piece of information without rebuilding the context in my head. Privacy had to stay central: the application runs on my Mac and keeps its SQLite database locally.
02 · Scope
A marketing tool suite, not a to-do list
Around forty modules bound together by the same context. The ones that matter day to day cover the work itself, not personal organisation.
- Acquisition: an SEO audit of a page with keyword leads, a GEO / AEO module that measures whether a brand is cited by generative engines and produces the Schema markup, and Meta Ads tracking.
- Sales: prospecting and pipeline, proposals, invoices, and a profitability module that shows what a client really brings in once the hours are counted.
- Production: editorial calendar, assisted content studio, video, LinkedIn, Instagram and WhatsApp.
- Management: objectives, priority reviews, reports, and a module that cross-references tasks, objectives and finances to say what should move.
- AI under control: search across my own files, assistant, e-mail and calendar, always with human approval before anything leaves the machine.
03 · Honest status
A test bench, not a product for sale
I designed and built these tools, and I use them every day on my own business. They do the job they were made for. They have not been deployed at a client: no client result is attributed to this tool, and it is not for sale.
What it proves lies elsewhere: to automate an SEO audit, a prospecting pipeline or a profitability calculation, you first have to understand the work in fine detail. The GEO / AEO skill is in fact applied to this very dossier, whose llms.txt and structured data can be checked online.
04 · Demonstration
What I learned building my own business
This demonstration recreates four real modules of the tool, on entirely fictitious data. It shows no client, no figure and no real document: it shows the way of working I built for my own business, and that I apply to the projects I run.
Martelazo
Friday · LyonInterface recreated here · fictitious data
Today’s view
Priorities, appointments and signals gathered before any work starts. The beginning of the day.
- 09:30Follow up the proposal sent to Client A
- 11:00Publish the article scheduled in the editorial calendar
- 14:00Check this week’s Meta Ads budget
What it changes: I start the day with the priorities, not with the inbox.
04priority actions
GEO / AEO
Measuring whether a brand is cited by generative engines, then producing the markup that helps it be cited.
Brand analysedExample LtdWith real web search
Signals present
- Complete business listing
- Valid structured data
- Mentions in three directories
What is missing
- A frequently asked questions page
- Markup for the main service
Schema to generateProfessionalServiceLocalBusinessOrganization
Example result: it does not depend on the brand you type. The real audit queries the engines.
What it changes: you stop guessing whether a brand exists for AI engines, and start measuring it.
2/5engines cite the brand
SEO
Auditing a page, checking the fundamentals and producing keyword and content leads.
Page auditedexample.com/servicesReviewed
- Title and meta description within the right lengths
- A single H1, but two near-identical H2s
- No internal link to the contact page
Keyword density62% on the main intent
What it changes: the audit becomes a list of fixes, not one more report.
3fixes proposed
Profitability
What a client really brings in once the hours are counted. Ranked by hourly margin.
| Client | Billed | Hours | Margin / h |
|---|---|---|---|
| Client A · tradesperson | 4 200 € | 38 | 110 € |
| Client B · restaurant | 2 800 € | 34 | 82 € |
| Client C · practice | 3 100 € | 52 | 60 € |
What it changes: the biggest-billing client is not the most profitable, and I know it before renewing.
84 €average hourly margin
Assistant
The assistant prepares and organises. Nothing goes outward without human confirmation.
Proposed action
Prepare the follow-up on proposal no. 2026-014, sent to Client A twelve days ago.
En attente de confirmation humaine
What it changes: AI saves time without ever deciding in my place.
1 clickbefore any outward action
This recreation shows the logic. I can show you the tool live during the interview.
05 · Skills demonstrated
Designing a system before choosing its screens
Technical
Working with a local database, APIs, tests and a modern, AI-assisted architecture.
Accountability
Protecting privacy and keeping a human approval step for outward actions.