Work

Real situations. Solutions put to work.

Five different contexts, with my specific contribution to each.

01 / E-commerce / AI agent

Drapeur · E-commerce

Product listings ready for review, without the copy and paste.

Prepare, then review

Copy + visuals

ShopifyDraft

You stay in control of publishing.

The work to be done

Drapeur prepared 20 to 30 product listings a day. The merchant already used ChatGPT, but still had to copy the content and create each listing in Shopify.

Before

Content could be prepared with ChatGPT, but every piece of copy and every visual still had to be transferred into the store by hand.

What was built

We built an agent that prepares copy and visuals from the merchant’s guidelines, then fills out the product listings in Shopify. Everything remains in draft for review before publishing.

Today

Copy and visuals arrive in draft listings. The merchant reviews them and decides what to publish. Over time, the client expanded the instructions independently.

What Antoine handled

I built and installed the agent for this workflow. The client uses it independently and continues to refine the instructions.

Results and boundaries

Listings remain in draft until the merchant approves them. The client expands the instructions over time.

One to two weeks of work saved per month, by the client’s estimate.

“It creates all my product listings for me. […] So a huge thank you, it saves me an incredible amount of time.”

Translated from FrenchAlexis · CEO, Drapeur

02 / Prospecting / AI agent

A wine merchant · Wine trade

Qualified prospects that match their criteria, without searching one by one.

Search, qualify, record

Your criteria

CRMTo review

You decide who to contact.

The work to be done

This client was finding wine shops and wholesalers by hand: searching, checking each criterion, entering the result in their CRM. The same work came back every week.

Before

Each prospect was found, checked against the merchant’s criteria, and entered into the CRM by hand.

What was built

We first wrote down their method: where they look, what disqualifies a contact, what they record in the CRM. I turned it into an agent that runs those steps, applies their judgment, and saves the results in their CRM. It took several passes before the qualification was truly theirs.

Today

They describe what they are looking for in one instruction and the agent returns a qualified list in their CRM. They decide who to contact.

What Antoine handled

I documented their method, built the agent, and refined the qualification with them until it reflected their criteria.

Results and boundaries

The agent prepares and records the list. The client reviews the results and decides whom to contact.

03 / Automation / Back office

Brest Bretagne Handball · Professional handball club

Connecting tools to remove duplicate entry.

From manual transfers to an automated connection

Information

Back office

An automated connection

The work to be done

Business information was moving manually through a back-office process.

Before

The team had to move the information in scope from one tool to another.

What was built

I automated that process end to end by connecting existing tools to avoid re-entering the same information.

Today

The back-office processing and transfers in scope are automated instead of being moved manually between tools.

What Antoine handled

I built the agreed flow between the existing tools.

Results and boundaries

The automation covers the transfers and back-office processing agreed for this engagement.

04 / Data / Practical AI

Edifix

Connecting project information to sales workflows.

Organize information

Notion

A separate focus: AI visuals for sales

The work to be done

Project information was scattered across local files, drives, and emails. Sketches needed to become visuals for sales conversations.

Before

Useful project information was scattered. Producing sales visuals was a separate need.

What was built

I structured information in Notion and worked on AI visuals to support sales.

Today

Project information is structured in Notion, with AI visuals used to support sales conversations.

What Antoine handled

I organized project information in Notion and worked separately on the AI visuals component.

Results and boundaries

Data organization and visual creation were handled as two distinct parts of the project.

05 / Product / Technical execution

Fabrik Labs

Rebuilding without losing the business logic.

From scoped requirements to implementation

Scoped needs

Build

My proposals, within a confidential project

The work to be done

An existing low-code system needs to be rebuilt from requirements already defined.

Before

The functional requirements were already defined and needed to be translated into a new technical implementation.

What was built

I implement already-scoped requirements, contribute my proposals, and respect the project’s confidentiality.

Today

Already-scoped requirements are turned into technical implementation, with my proposals. Delivery details remain confidential.

What Antoine handled

I contribute technical execution and proposals within the scope agreed with the team.

Results and boundaries

The feature and delivery details remain confidential.

Diagrams are concept illustrations, not screenshots of the systems.

Your next step

Which task would you rather not do by hand again?

Describe what comes in, what you do today, and the result you need. We will work out what can be handled for you.

Your task does not look like any of these examples? Describe it anyway.

Describe my task Book a 30-minute call

Scope, timeline, price, and the expected outcome are defined in writing.