The auto repair shop asks every customer to leave a review, and AI helps draft the replies
Let's take an independent auto repair shop with 5 lifts as an example. Its rating on the maps is 4.1, and new customers go to nearby chain shops rated 4.7 and above. Here is the scheme. Eighteen hours after collecting the car, every customer gets the same WhatsApp message with a thank-you and links where they can leave a review. If a customer is unhappy, the owner finds out right away. GigaChat, a Russian AI model, drafts replies to reviews. By the estimate, if customers are happy with the repairs, the rating reaches 4.7 in 4 months and enquiries from the maps grow by 47%.
With a 4.1 rating, new customers drive past
The shop in the example is in a residential area of a big city. On the maps, it sits next to chain shops and large repair centers. The rating affects how high a listing appears in Yandex Maps and 2GIS search results. At 4.1, the shop ends up below its neighbors. Even after opening its listing, a new customer usually picks the neighbor rated 4.7.
The shop itself does decent work: quality is average for the market, prices are moderate. But happy customers hardly ever leave reviews: the car is fixed, and they get on with their day. An unhappy customer writes straight away, and that one-star review sits unanswered for weeks until the manager gets round to it.
So the maps show a lopsided picture. A good experience barely leaves a trace there. A bad one stays in view for everyone who opens the listing. In this example, the shop has collected only 23 reviews in a year and a half.
Eighteen hours after the repair, the customer is asked to rate the job
When a work order is closed in Alpha-Auto, a 1C-based program for running auto repair shops, n8n starts a workflow. n8n is a tool that carries out a chain of actions on its own, when an event happens or at a set time. Eighteen hours after the car is collected, the customer gets a WhatsApp message. It contains a thank-you, links to Yandex Maps and 2GIS and a question about how everything went. Every customer gets the same message. Low ratings go straight to the shop owner. Separately, GigaChat drafts replies to new reviews on the maps.
Alpha-Auto sends the customer's details and the list of work done to n8n
The customer has driven the car, the impression is fresh, and the request doesn't feel pushy
A thank-you, links to Yandex Maps and 2GIS and a question about how it went, with rating buttons from 1 to 5 stars
The workflow passes any rating of 3★ or below to the owner, who calls the customer to sort things out. The customer already has the map links
GigaChat drafts replies; the manager checks and publishes them
Why 18 hours
If you ask when the car is collected, the customer is in a hurry: they say thanks and leave. A week later, they have forgotten the details. After 18 hours, they have driven the car and can tell how it runs after the repair, and the impression is still fresh. The request also comes some time after payment, so it is less likely to annoy.
The same request for every customer
Every customer gets the links to Yandex Maps and 2GIS, whatever their rating and before they answer the question. The shop asks for an honest review and doesn't suggest what to write or what rating to give. Customers get nothing for a review: no discount, no bonus, no gift. They write their reviews on the maps themselves, in their own words.
What happens after a low rating
The question about how it went is for the shop itself: the answers show the owner what to improve. If a customer gives 3 or below, the owner is notified right away with the work order details and calls the customer to find out what went wrong and fix it. The customer already has the map links and decides for themselves whether to leave a review. Nobody asks the customer to delete or change a review, and nothing is offered in return.
As of September 2026, Yandex Maps and 2GIS allow businesses to ask customers for reviews and send them a link to the listing. Both platforms forbid paying for reviews and rewarding only positive ones. Yandex also asks businesses not to suggest what to write. That's why the scheme sends the same request to everyone, with no hints and no rewards. Customers write reviews on their own phones. The platforms update their rules from time to time, so before launch I check the message texts against them.
GigaChat drafts replies to reviews
Yandex Business and 2GIS notify the owner of new reviews by email. Every 4 hours, n8n reads these emails, and for each new review GigaChat writes 2–3 draft replies that match the review's tone and details. The manager picks one, edits it if needed and publishes it in the Yandex Business or 2GIS dashboard.
Yandex offers no official way for other programs to reply to reviews, so the manager does this last step by hand. Negative reviews get a separate template: the reply acknowledges the problem and offers a way to fix it. The goal is to reply within an hour of the workflow finding the review.
A report for the owner
The report is built in Metabase, a tool for charts and tables. One screen shows how the rating changes day by day, how customers rate different kinds of work (diagnostics, tire fitting, suspension, engine repairs) and where new customers come from. This tells the owner whether the scheme works and what else can be improved.
Runs every step: gets the signal from 1C, waits out the pause, sends the customer the message, passes low ratings to the owner, calls GigaChat and watches for new reviews. Runs on a separate server
Reports that a work order is closed and passes the data to the workflow
WhatsApp messaging service: sends customers messages with map links and rating buttons
Drafts polite replies to reviews and goes through the survey results
An email about a new review starts the workflow; the manager publishes the reply in the dashboard
The same workflow handles emails about new reviews on 2GIS
The owner's screen: rating by day and ratings by type of work
Database that stores reviews, replies and customer ratings for reports
What should change in 4 months
The figures below are an estimate for the example in the first section, based on typical numbers for small and mid-sized businesses in Russia. Each card shows the starting point, then the target.
estimate for Yandex Maps and 2GIS, if new reviews average about 4.8
target after 4 months
target: a reply to every review, including those rated 1–3★; new reviews are checked every 4 hours
target compared with the month before launch; visits from the maps to the website are tracked with UTM tags, which are labels added to the link
The map rating is made up of all reviews. With 184 reviews, it reaches 4.7 if the 161 new reviews average about 4.8. That happens when almost all customers are happy with the repairs and simply didn't write reviews before. If the survey average is below 4.8, the map rating won't rise as much, and the report will show which kinds of work to improve first.
The main target is 47% more enquiries from the maps. The estimate assumes that with a rating above 4.5, the listing moves up in search results and appears alongside the nearby chain shops. On the maps, a customer's path is short: see the listing, tap, call. None of this requires paid advertising.
The report by type of work is useful in its own right: it shows which service pulls the overall rating down. For example, if suspension diagnostics averages 4.3 and tire fitting 4.9, start with diagnostics: how the car is checked in and what the mechanic explains to the customer.
What it costs and how long it takes
connecting to 1C, setting up n8n, testing the whole chain and training the manager
server, WAPI, GigaChat requests and support
from the start of work to launch
The first shop gets 30% off and a 90-day warranty. You can start with a 5,000 ₽ audit: I'll look at your listing and reviews and recalculate the figures for your shop.
Where else this scheme helps
The scheme is the same everywhere. A little after an order is closed, every customer gets the same review request and a short survey. Low ratings go straight to the owner, and AI helps answer reviews. It suits any business that closes orders in a CRM or accounting system and has a listing on Yandex, 2GIS or Avito:
- → Local shops and showrooms — a WhatsApp request to rate the purchase after delivery or a visit
- → Restaurants and cafes — the request arrives some time after the bill is paid in iiko or r_keeper, which are restaurant till systems
- → Medical centers and clinics — after the appointment, with tactful wording
- → Local services — cleaning, repairs, home visits, tutors
- → Wildberries and Ozon sellers — for orders through their own website, where the seller has the buyer's contact details
- A ready-made workflow in n8n: the pause, the same message for every customer, an alert to the owner about low ratings and an AI-drafted reply
- Instructions the AI model follows to write polite replies. I tune them to the company's style and industry
- Reading the emails about new reviews from Yandex Business and 2GIS
- A Metabase report: rating by day, ratings by service, where customers come from
If your customers are happy and your rating on the maps is still below 4.5, this scheme can help raise it
The scheme relies on regular free listings on the maps, so you don't need to buy advertising. Launch takes 3–4 weeks.
Related cases and ready-made solutions
After a missed call, the customer gets an SMS with the real ship date
The phone system reports a missed call, and the service finds the customer and their orders in two Shopify…
A store's own AI sales assistant won a blind test 34 : 12
Built for the brand's Russian site on Russian language models, so visitor data stays in Russia. It answers…
A voice agent that answers salon calls day and night and books clients into YClients or another booking system
It answers calls at any hour, books clients into YClients or another booking system with an API, and sends an…
Start with the 5,000 ₽ audit: a written report and a payback estimate
I'll tell you what to automate first, when it will pay for itself and whether your task needs AI at all. Sometimes it doesn't.
Or just ask a question. During business hours I reply within 2 hours