Four AI agents for a car hub reply to every request at once, guide the client to handover and catch good lots
The hub imports cars from auctions abroad and sources them for each client. Four agents take over the routine, from the first reply to a new request to an alert about a good lot. Money and decisions stay with people.
- Client
- Car dealer hubimports cars from US and Korean auctions, sources them on the domestic market, sells
- Service
- requests, calls and client follow-through
- Status
- In production at the clientfour agents work as one team
- Before → After
- Requests that end in a sale: 3% → 5%first reply: 1–1.5 days → up to 3 min, and 50–60% of requests are paid for with ads
per client data the client gave us this numberestimate we calculated it; the assumptions are next to it
The client's situation
The hub brings cars from auctions in the US and Korea, sources cars on the domestic market and sells cars from stock, some of which it buys with its own money. Requests come from the website, Telegram, Instagram and Facebook, and a new request waited 1–1.5 days for a reply, even though 50–60% of requests come from paid ads. Instagram messages were answered from a phone, duplicates in the CRM were hunted down by hand, there was no dedicated account manager, and good “Buy Now” lots were tracked manually for 1–2 hours a day, with some of them missed.
After the purchase, clients kept asking in the chat where their car was, and someone had to collect the status from the logistics team.
of requests ended in a sale, 2% in weak months
per client dataa new request waited for the first reply
per client dataof requests come from paid ads
per client datais all the time a good “Buy Now” lot stays available; before, the auctions were watched by hand 1–2 hours a day
per client dataThe rule for all four agentsNo amounts or dates outside the system: the manager edits every lot offer, and a lot is bought only after the manager taps the button.
What the system does
- Messages a new client in Telegram within 3 minutes, using the name and number from the form, day and night, while the interest is still fresh.
- Calls a new client by voice right after the request: clarifies what they are looking for in Ukrainian and logs a call summary in the CRM, so the manager knows what the person needs before the first conversation.
- Answers inquiries from the website, Telegram, Instagram and Facebook by the company's rules and collects the details so the client is contacted no more than twice.
- Finds out what the client needs and creates the CRM record right away, with no duplicates, adding a summary and tags, so the manager sees at once who this is and what they need.
- Sells like the best manager on the team: knows the services, the “we import at our own cost” terms, credit, leasing and current promotions from a knowledge base that is updated from call recordings.
- Finds cars in stock that match the request and sends the client options while they are still in the chat.
- Picks auction lots to the manager's spec and prepares an offer with a turnkey cost breakdown, which the manager edits before it goes out.
- Follows up with those who said “I'll think about it” when there is a reason, such as a new promotion or a similar car, and stops once they reply. It hands a hot client to the manager with the full history, and at night it collects everything for the morning or messages the person on duty.
- After the deposit, tells the client in Telegram where the car is: auction, container, port, customs, service shop, so the manager no longer chases the logistics team for updates.
- Calls the client when the car is ready at the service shop: reports that it is ready and agrees on a pickup time, so the car does not sit around and the manager does not have to call clients one by one.
- Watches “Buy Now” lots at US auctions and buys a lot with one tap from the manager: compares the price with the market by year, mileage and condition, sends an alert with a button in Telegram, and while the lot is still live, the system buys it. The decision is always a person's.
How it is built
The four agents share one foundation: a knowledge base, the company's CRM as the source of truth, and an action log. The first line and the account manager also talk to clients by voice. A person stands wherever money and commitments are involved: the lot offer, the purchase, amounts and dates for the client.
- Channels
- websiteTelegramInstagramFacebookvoice calls
- Integrations
- the company's CRMtelephony and call transcriptsauctions: the “Buy Now” categoryTelegram alert channel
How it works
- clientRequestFrom the website or a messenger
- first lineFirst replyTelegram within 3 minutes and a voice call right after the request
- agentClient detailsNo more than two contacts with the client
- CRMCRM recordNo duplicates, with the call summary
- agentsA car for the requestFrom stock or an auction lot
- managerLot purchaseAlert with a button: the system buys the lot after the tap
- manager and clientDealA hot client with the full history, money stays with people
- account managerFollow-throughCar status from the auction to the service shop
- voiceCar is readyA call to the client and a pickup time
“Buy Now” runs in parallel: it watches the auctions by filters and sends an alert with a button in Telegram. A lot is bought only after the manager taps it.
Result and payback
from the request to the first reply in Telegram
per client dataof requests end in a deal
per client datais all the time a good “Buy Now” lot stays available: the system spots it at once and sends the manager an alert with a button
per client dataProject payback: ≈2 months estimateEvery 100 requests now bring 5 deals instead of 3. At ≈300 requests a month (an assumption), that is +6 deals at $1,000 profit each, so a system of four agents at $12,000 (within the “System” range on the pricing page) pays for itself in ≈2 months.
What is running
- running
First reply in Telegram within 3 minutes, a voice call, a CRM record without duplicates
- running
Client follow-through in Telegram and auction lot search with a turnkey cost breakdown
- running
“Buy Now”: an alert with a button in Telegram, and the lot is bought after the manager taps
A similar task in your business?Solomiia, Pavlo's AI assistant, will ask about the details and prepare your conversation with him.
emilus. Agent
An AI agent that answers your customers in chats. Diagnostic $300 → pilot $2,500–3,500 in 3 weeks → system $6,000–15,000; we name the exact system price after the diagnostic. Support $300 or $500 a month; you pay for AI usage directly to its provider.