Retail and distribution, six simulated engagements.
Six companies with fictional names, across Europe and the United States, from a former bookseller's venture to an eight-store DIY chain. In each simulation, AI forecasts, reads, drafts, and alerts. Decisions remain with the teams, from placing supplier orders to issuing customer refunds.
These simulations demonstrate our method on representative cases. The companies and their exchanges are fictional.
How AI is already reshaping retail, from the shop floor to customer service.
of French micro-enterprises and SMEs use at least one AI solution in 2026, up from 26% a year earlier.
of French online shoppers already use generative AI to plan their purchases, rising to 49% among 15 to 24 year olds.
increase in visits driven by AI assistants to US retail websites during the 2025 festive season compared to 2024.
of US online sales were expected to be returned in 2025. 85% of surveyed major retailers use AI to combat returns fraud.
Since 28 June 2025, e-commerce websites and mobile applications have been required to be accessible to people with disabilities under the European Accessibility Act. Micro-enterprises providing services are exempt.
Six companies, six starting points.
An online running and hiking apparel brand whose customer service team is swamped with tracking queries and size exchange requests.
Every customer receives an answer, in French as well as in Spanish.
- Aunage designs running and hiking apparel in Nantes, manufactured in Portugal and sold through its website in France and Spain. The brand was founded by a former trail runner. Its fourteen employees are split between Nantes, for product and brand design, and Madrid, where a small team supports Spanish customers.
- After each collection launch, messages pile up, and the same questions keep recurring, from parcel tracking to size exchanges. Two advisers reply manually, in French and Spanish, running several days behind during sales periods.
- End-of-season unsold stock is another heavy burden. In France, legislation has prohibited destroying unsold non-food items since 2022, requiring them to be given a second life.
- An assistant replies on the website and by email, in French and Spanish, introducing itself as an AI from its very first message. It tracks parcels with carriers and checks the brand's exchange policy, then logs size exchanges directly whenever requests meet the criteria.
- For lost parcels or out-of-policy requests, it hands over to an adviser along with a summary of the exchange and the order details. The adviser replies without asking the customer to repeat themselves.
What remains with the advisers
Advisers alone decide on exceptional refunds and goodwill gestures. The assistant never promises delivery dates that are not confirmed by the carrier.
- First response time
- Requests resolved without an adviser
- Size exchanges logged on the same day
- Messages in Spanish handled in Spanish
A family-owned chain of DIY and garden centres on both sides of the Rhine, where shelves run out of de-icing salt at the first frost and overflow with bulbs at the end of the season.
Every store has the right stock before the first frost.
- Boisseau is a family business operating eight DIY and garden centres in Strasbourg, Mulhouse, Freiburg im Breisgau, and Basel. It employs two hundred and fifty people, and each store has its own department managers, from gardening to tools.
- Orders are placed weekly, based on the previous year's sales and department managers' intuition. At the first frost, de-icing salt runs out in Freiburg im Breisgau while sitting idle in the Mulhouse stockroom.
- At the end of the season, unsold gardening stock piles up. In France, the law has banned its destruction since 2022, requiring every French store to find a second life for it.
- An agent forecasts weekly sales for each store, department by department, using sales history and weather forecasts. For French stores, public Météo-France data has been free of charge since 1 January 2024. The agent also factors in promotions and public holidays in each country.
- It then suggests restocking, as well as transfers between stores when one lacks what another holds in excess. Department managers receive these suggestions on Monday mornings, complete with the rationale for each, and approve or adjust them.
What remains with the department managers
Department managers decide on every order and transfer. Buyers negotiate with suppliers directly, and the agent alters no pricing.
- Shelf stockouts during peak periods
- Backroom stock at the end of the season
- Suggestions accepted without adjustments
- Transfers between stores
An electrical and plumbing supplies wholesaler whose sales reps spend their mornings retyping unstructured orders, often sent as photos.
Every order ready before the tradesperson arrives at the counter.
- Toise Distribution supplies electricians, plumbers, heating engineers, and HVAC specialists across south-eastern France and north-western Italy. Its four branches are located in Marseille, Nice, Genoa, and Turin, employing one hundred and ten people, including around forty counter and telephone sales reps.
- Tradespeople order however they can, straight from the building site. A scribbled list captured in a photo, a voice note, an email with half-accurate part numbers, or a phone call between jobs. Every morning, sales reps retype everything into the management system, while tradespeople arriving at the counter wait for their orders to be prepared.
- In Italy, all business-to-business invoices have been routed through the tax authority exchange system since 2019. In France, receiving electronic invoices has been mandatory since 1 September 2026.
- An agent reads every order as soon as it arrives, whether sent as a photo or a message, recognising part numbers even when poorly written. It checks branch stock and prepares the order in the management system.
- When an item is out of stock, it suggests an equivalent from another brand, or the same product from the nearest branch. The sales rep validates the order, and the tradesperson receives a message once it is ready.
What remains with the sales reps
Sales reps validate every order and every suggested equivalent. The credit manager alone determines the credit limits granted to tradespeople, and the agent never pursues overdue payments without their approval.
- Time between order receipt and system entry
- Orders ready when the tradesperson arrives
- Equivalents accepted by tradespeople
- Part number errors detected at delivery
A former bookseller who wants to showcase the stock of independent retailers in central Rennes online, with no shopkeepers subscribed yet.
The offering refocused before approaching the first shopkeeper.
- Chaland is the venture of a former bookshop manager in Rennes. She wants to display the stock of city-centre independent retailers on a map, so customers can check whether a book or a toy is waiting for them before heading out. Retailers would pay a subscription.
- The project consists of a mockup and a list of retailers she knows. She has neither proof that a shopkeeper would pay, nor any idea of what proportion of shops keep their stock truly up to date.
- A number of retailers are already wary of major platforms, and she does not want to pitch the wrong offering when approaching them.
- Within a few days, the mockup becomes an online demo connected to the till software of a friendly bookshop. Before setting up any meetings, the offer is tested against a panel of simulated buyers, AI agents modelled on independent retailer profiles.
- The panel highlights an objection she had not anticipated. Many retailers do not manage inventory through their till system, meaning the showcase would display out-of-stock items. The offering is refocused on shops whose till already tracks stock, such as bookshops and toy stores, combining online reservation with in-store payment.
What remains with the founder
A simulated buyer never pays; real validation requires a shopkeeper to subscribe and pay. The founder conducts every meeting herself and sets the subscription price.
- Offer versions tested against the panel
- Meetings secured with retailers
- Retailers with subscriptions
- Reservations generated through the platform
A Portland-based online shop for second-hand outdoor gear, where every item waits for days before being catalogued and listed for sale.
Every item gets its listing upon arrival.
- Steelyard buys second-hand outdoor gear from individuals and takes items on consignment, from rain jackets to tents, reselling them online after inspection. Its Portland workshop employs eighteen people, and orders ship across the United States.
- Every item requires a complete listing, detailing the brand, size, measurements, and any visible defects. Two employees write them manually, and gear piles up in the workshop waiting to be listed online.
- Buyers then ask the same questions about the true condition of items, and items are returned when their condition falls short of the listing's description.
- The employee photographs the item from multiple angles, including the label. An agent reads the photos and drafts the listing, including the brand, size, measurements recorded by the employee, and visible flaws. It suggests a condition grade according to Steelyard's grading scale.
- The employee reviews the listing and assigns the final grade. They also note anything the photos cannot capture, such as a stiff zip. The agent then answers buyers' questions based on the listing and photos, introducing itself as an automated assistant.
What remains with the team
Employees alone assign the final condition grade to each item, and the team sets resale prices. Disputes and refunds are always handled by a team member.
- Time from item arrival to listing online
- Listings corrected by the employee
- Buyer questions prior to purchase
- Returns due to condition not matching the listing
A neighbourhood pet shop chain whose loyal customers buy their kibble elsewhere simply because they did not think of it in time.
Every loyal customer alerted before the bag runs out.
- Pottle operates six neighbourhood pet shops in Minneapolis and Saint Paul, featuring a food aisle and a grooming salon in every store. The family-owned business employs eighty people, with tens of thousands of customers enrolled in its loyalty programme.
- Customers restock kibble every four to six weeks, rarely on the exact day the bag runs empty. They end up buying from supermarkets or major online retailers, eroding loyalty without anyone noticing.
- Yet the loyalty programme retains the purchase history of every enrolled customer. Nobody has the time to put it to use.
- For each enrolled customer who has opted in to SMS, an agent identifies when their bag is likely to be running low. It drafts a personalised reminder, including the pet's name and usual store, offering to set a bag aside.
- When the customer replies, the agent answers while introducing itself as an automated assistant, forwarding each reservation to the store. Customers can unsubscribe from all reminders by replying STOP.
What remains with the store team
The team sets campaign rules, such as frequency and audience selection, and approves every new message template. The agent provides no animal health advice and redirects medical questions to a vet.
- Loyal customers returning before running out
- Reservations made via SMS
- Opt-outs following a reminder
- Questions forwarded to stores
The trade professional uploads plans to the website, and the CRM generates the quote
No case study has been published for this sector yet. The MAZLOKS case study covers the website and CRM of a supplier selling to trade professionals. It is filed under Real estate and construction.