AI automation agency: the agent prepares, you approve

It designs and installs systems that handle the repetitive part of a process inside the software you already have, instead of selling you one more tool to open.

For us, this breaks down into three areas of work, each tied to a result.

• Find customers. A prospecting agent applies your criteria, identifies the right contacts, prepares a message based on a verifiable observation, follows up, and keeps the CRM up to date.
• Respond to customers. An inquiry and quote agent reads incoming messages, recognizes the customer, looks up references, and prepares the proposal with its sources.
• Reduce administrative work. Message sorting, quote follow-ups, CRM updates, accounting documents: this is the field of custom automation, where a fixed rule is often enough and AI does not always belong.

The word “agency” covers very different lines of work in this market: some design AI products, some train teams, and some connect tools. We do one thing: remove a repetitive task from a workstation and make it verifiable.

One task, chosen because it comes up every day, is done in writing, and produces a result that can be checked. Never “automate everything.”

A task is suitable for an agent when it meets three conditions: it arrives in writing (a message, form, or file), answering it requires finding information elsewhere, and a person can review the result in a few seconds. A price request with twelve references, a follow-up on a quote sent three weeks ago, and sorting messages in a shared inbox all fit this category.

By contrast, we refuse to start with a project that replaces an entire position, a task performed by voice or in the field, or a process that no one in the company can describe precisely. An automated unclear process remains an unclear process, executed faster.

There is also organizing work that no one sells and that determines everything that follows: writing down what the agent needs to know about your company before choosing any tool. We explain it in AI knowledge base: organize what your agent needs to know before changing models.

The agent prepares and proposes; a person approves anything that commits the company toward a customer, supplier, or employee. This boundary is written before construction, not guessed afterward.

The third column is the most important. An agency that cannot tell you what it refuses to automate has not yet told you where its system stops. The same reasoning applies to access: an agent connected to your email and CRM inherits your permissions. In four questions to ask before connecting an AI agent to your tools, we listed what you should require from any provider, including us.

Two to three weeks for a first system in one role, including two weeks of trial operation. During the trial, the agent works without sending anything; you decide whether to put it into service at the end.

• The discovery audit (free, 20 to 30 minutes, by video call or on site): one task, your tools, and the decisions that must remain human. You leave with a written opinion.
• The written scope: what the agent may do alone, what goes through you, the authorized sources, and a fixed-price quote. When the need extends beyond the task reviewed, we propose a paid scoping engagement whose scope and price are written before work begins.
• Construction: connection to your tools in read-only mode first, rules, output template, and tests on your past cases.
• Trial operation, two weeks included in the delivery timeline: the agent works without sending anything, we compare its proposals with what your teams actually answered, and we make corrections.
• Ongoing support: supervision and adjustments each month, with no commitment and one month's notice.

The scope and price are written before work begins, and no paid work starts until you have approved the quote. Hosting and third-party tools are in your name, paid directly by you, with no markup from us. See the framework and offers.

What using AI will cost you does not depend that much on the model you choose: it depends on how many rounds the agent takes to finish a task. That is the subject of the cost of an AI agent: pay for the task, not the token.

Then we say so during the discovery audit, and you have spent nothing. Some repetitive tasks call for a fixed rule, a setting in existing software, or an organizational decision that cannot be coded.

Three cases where the conclusion is not to use AI: a quote follow-up that always goes out after the same delay does not need a language model, a simple automation is enough and will cost less to maintain; an already structured order flow can be handled by a connector between two software tools; and a shared inbox that no one wants to open is often first a problem of assigning responsibilities.

The audit separates these cases. An AI agent makes sense when the task requires interpreting free-form text, searching several sources, and producing something a person will review. Not before.

One human and agents. Équipage IA is led by Hakim Ouahabi, the company's only human, with a team covering nine AI agent roles: operations management, prospecting, customer relations, customer service, monitoring, writing and search visibility, social media, studio, and development.

Each role has a job description and follows the same charter: the agent prepares, a human approves anything that commits the company, and nothing is sent outside the company without Hakim's approval. Their paid third-party tools (research, images, voice) have daily caps and are logged. The roles are not all at the same stage: monitoring runs every day, while prospecting is still in trial operation. This is the system we install for our customers: it runs in our company first, and that is where we find its defects before you do, including ordinary ones such as slowness, the subject of why your AI agent is slow.

Small and medium-sized businesses that receive inquiries, prepare quotes, or prospect in writing, anywhere in France, with an on-site meeting available in Île-de-France.