How to implement AI in a small business without a big project

How many small businesses actually use AI?

According to INSEE, France's national statistics office, 20% of EU businesses with 10 or more employees used at least one AI technology in 2025.

Among small firms with 10 to 49 employees, the EU figure drops to 17%. In France it was 15%, against 58% for companies with 250 employees or more.

Of the French businesses not using AI, 54% cited a lack of expertise and 38% cited data protection. That is what most SMBs look like: no AI team, and documents organized for people to read.

Is a ChatGPT or Copilot license for every employee enough?

ChatGPT Business is priced per user. Microsoft 365 Copilot is an add-on billed per user per month, with an annual commitment and a qualifying Microsoft 365 license on top.

Every employee gets an assistant that writes, summarizes and translates well. With a connector, it can find the files that person is already allowed to open.

Nobody has defined the job with you, or how your company does it. People re-explain their context with every question, and the bill grows with headcount. The page "Is ChatGPT or Copilot enough for my company?" goes through this comparison.

Should you hire a consultancy to build a custom AI project?

A consultancy or systems integrator can build exactly what you ask for. That path takes a written specification, a project lead on your side and a maintenance budget after delivery. Check timelines and prices with each provider.

In July 2024, Gartner predicted that at least 30% of generative AI projects would be dropped after proof of concept by end of 2025.

Gartner pointed to poor data quality, weak risk controls, rising costs and unclear business value. So ask any provider three questions up front: which data, what it costs after delivery, and which specific task it serves.

What about an AI agent platform you configure yourself?

Microsoft describes Copilot Studio as a platform for creating and managing custom agents with no code. Dust, a French company, is priced per seat per month.

These tools give you the building blocks. Someone in-house still has to design the agents, test them and keep them up to date. Dust itself talks about "AI Operators" who build agents and run them for their team.

If you have that person, it's a good option. If you don't, you're back to the lack of expertise that 54% of French businesses without AI cited in the INSEE survey.

How does Saména approach AI adoption?

We start from the documents you already have, where they already live: SharePoint, OneDrive or Dropbox. Saména reads them with read-only access, only in the folders you pick, and you don't have to sort anything first.

Our engineers, trained at Télécom Paris and specialized in AI, work with you to understand how your teams operate. Then they build one agent per task: a sales proposal, meeting notes, a reply to a client.

You can start with one department and a first agent. Your teams use it in the Saména app, and also from Claude, ChatGPT or Copilot. The agent cites its sources and drafts Word documents and PowerPoint decks.

• One department, one task that comes up often
• Your existing documents, with no preparation
• A first agent tested on your real cases, then adjusted
• Other departments next, once the first agent is in use

How do you measure whether it works?

Before you start, write down how the task gets done today. Who does it, how often, which folders they dig through, how many rounds before the final version.

During the trial, give the agent recent, real cases. Check each answer against the document it cites. Note what gets used as is, what gets edited and what gets thrown out.

After a few weeks, look at one simple thing: does the team use it without being reminded? That tells you more than a productivity gain estimated up front.

What is leadership's role in an AI strategy for SMBs?

Leadership picks the first department and the task to serve. It names a point person who tells our engineers what is missing.

IT, or whoever manages it, opens access to the chosen folders and signs off on security.

The CEO then decides whether to extend, based on what the team has seen. If the CEO uses it too, the team sees that the trial matters.

What happens to your data?

Your files stay in SharePoint, OneDrive or Dropbox. Saména keeps no copy, and what it retains from your documents is encrypted with a key unique to your company. The master key is stored outside the database.

Servers are in France and the database is in the European Union. The models come from Mistral AI, under a contract that prohibits training on your data.

Saména can also run on your own servers or on a dedicated instance, and everything is deleted on request. The page "Where does our data go with enterprise AI?" has the details.

Frequently asked

Do we need an in-house AI team?

No. Our engineers build and adjust the agents. Your teams ask them questions the way they would ask a colleague, and a point person flags what is missing.

Do we have to clean up our data first?

No. Saména reads your documents as they are, in the folders you choose, and links them by client, project, method and team.

Which department should we start with?

The one where a task comes up often and means digging through folders: sales proposals, meeting notes, replies to clients. The page "AI agents for business: definition, examples and how to set one up" gives examples by role.

Will employees have to switch tools?

No. The agents live in the Saména app, and also inside Claude, ChatGPT or Copilot. People who already use one of these assistants keep it.

What does a first step cost?

A setup fee based on your document volume, then a monthly subscription for the whole company with unlimited users. We share pricing in a demo. The page "How much does a custom AI agent cost for a small business?" explains the model.