Skip to content
Marsanne Studio

Retail

AI for retail businesses with 20 to 500 employees

Replenishment, product sheets, quotes for non-catalogue requests. Marsanne puts AI systems into service on the processes that make up a retailer's core business, and every result they produce is checked by a second system.

Updated on

AI in retail: where businesses stand

In 2025, 13% of trade companies with 10 or more employees use at least one AI technology, compared with 18% across all industries. Across all industries, 71% of companies that do not use it say they see no use for it (Insee, ICT survey 2025).

Business leaders mainly expect AI to improve what already exists: 94% of the SME and mid-sized company leaders surveyed cite an optimisation objective, and 54% a development objective (Bpifrance Le Lab, 2025).

In a business with 20 to 500 employees, losses are concentrated in four places:

  • The catalogue. Suppliers send PDFs, spreadsheets and photos, each in its own format. The online catalogue stops where the person doing the copying has had time to get to.
  • Replenishment. A single rule is applied to thousands of items. Stockouts on fast-moving items and overstock on the rest coexist.
  • Advice. Knowledge of technical products sits in the heads of two or three salespeople.
  • Non-catalogue requests. Made-to-measure quotes, a part to track down: they take too much time and are turned down from the outset.

Demonstration scenario

Answering the made-to-measure quote requests you used to turn down

The company described is fictional. This scenario shows how we approach a problem; it does not report a result achieved for a client.

A shelf of identical bottles, and among them a blue vase of a different shape.

The situation

A bathroom and tile merchant: three showrooms, forty-five employees. Some requests fall outside the catalogue: a vanity unit with a non-standard dimension, a worktop cut to measure.

Pricing one of these requests takes about three days: prices from several suppliers, feasibility with the manufacturer, a search for a comparable previous quote, formatting. Salespeople have got into the habit of turning them down, and the customer goes to a competitor.

What the company already has

Three years of made-to-measure quotes, won and lost, with their actual cost price. The supplier price lists currently in force. The feasibility rules, known to the technical manager: minimum thickness of a cut worktop, actual lead times for each manufacturer.

What the system produces

The salesperson enters the request as the customer states it, with the dimensions and a photo. The system finds the closest past quotes, applies the price lists and the feasibility rules, and returns a complete quote: components, price, lead time, and the points that call for a human decision. The quote goes out within twenty-four hours.

What is checked

A second system, separate from the one that priced the quote, checks each quote before it reaches the salesperson:

  • the margin stays within the limits set by management;
  • the lead time quoted is consistent with the manufacturer’s actual lead time;
  • each component is available, or its replacement is flagged;
  • each price points back to the line of the supplier price list it comes from.

When a quote falls outside the limits, the check states which criteria it does not meet. The record stays attached to the quote. The principle is described in detail on the home page.

What the system does not do

It does not send the quote to the customer and does not set the commercial discount. It works on an export of quotes and price lists; connecting it to the ERP is discussed once the value has been demonstrated.

How the result is measured

The additional revenue generated on the requests the company used to turn down.

Product sheets, replenishment and customer service: three other uses of AI in retail

Stockouts and overstock cost retailers worldwide 1,700 billion dollars in 2024, of which 1,200 billion for stockouts alone (IHL Group, 2024).

Product sheetsReplenishmentCustomer service
TodayOne person copies out the supplier documents. Older sheets are never updated.A multiple of last month’s sales, applied to the whole catalogue.Questions about an order, a delivery, a return or a warranty arrive by email and by phone. Each reply means cross-checking the order history, the carrier tracking and the terms of sale.
The system producesThe attributes extracted from the original documents, filed in the company’s product repository, and the sheet written up.A forecast by item and by store, and an order proposal ranked by importance.A reply prepared from the customer’s file: their order, its tracking, the terms that apply. Requests that involve a goodwill gesture are flagged.
What is checkedEach published attribute points back to its supplier document. Mandatory mentions are checked separately.Minimum order quantity, storage capacity, expiry dates. A rejected line is rejected with its reason.Each statement points back to a document in the file. No refund or goodwill gesture is proposed without approval.
Data neededThe raw supplier documents.Two to three years of sales history by item and by store.The history of orders and exchanges, the general terms, an export of delivery tracking.

What Marsanne does not do in retail

  • The general-purpose chatbot on the website. Tools on the market do this, and it does not touch the core business.
  • Writing marketing content. Consumer tools are enough.
  • CV screening. The market is saturated and the regulatory ground is risky.
  • Software that already exists. When a vendor covers the need, we say so and tell you which one.

To gauge the expected return of a project before launching it, see How to measure the return on investment of AI.

Frequently asked questions

Do we need to connect our ERP or our tills?

Not to get started. The system works on exports: sales history, price lists, supplier documents. A few people use it for real, and the question of connecting it comes up once the value has been demonstrated.

What data do we need?

It depends on the use case. For replenishment, two to three years of sales history. For product sheets, the raw supplier documents. For made-to-measure quotes, past quotes with their cost price. If the company does not have this data, we say so before we start.

How long before we see a first result?

Development generally takes a few weeks. Users benefit as soon as they start using it.

What happens to our data?

Once the solution is in service, Marsanne no longer has access to your data. You remain its sole owner.

Will our teams need training?

No. The system is designed around the actual work of the team that uses it, and can be used without training.

Articles — Retail

In the press

The latest articles from the studio's AI watch on AI in this industry, summarised in English with their source.

All 80 articles on this industry in the AI watch →

Find out where AI can help in your business

The executive diagnostic takes ten minutes. It compares your answers with the situations we encounter in retail and gives you concrete leads.

Start the diagnostic