// Case study
AutoHunter
An AI hunter for good car deals on AUTO.RIA. It checks new listings every 2 hours, compares the price with the market and sends a Telegram card with advice on what to inspect.
// Task
Good deals disappear within hours
Good offers on AUTO.RIA go fast. Catching them means scanning hundreds of listings by hand several times a day, comparing prices and filtering out doubtful ones.
The goal: a system that does it on its own, runs around the clock and costs pennies to maintain.
// Solution
Three scenarios, a database, and AI only where needed
The user sets filters in plain words in Telegram, and AI turns them into search parameters. Every 2 hours the system pulls new listings, calculates the average market price and keeps the ones below market.
Only the shortlisted cars go to AI for review: pros, risks and an inspection checklist. All heavy logic lives in Supabase SQL functions, so Make spends a minimum of credits.
- Natural-language filters with a "does this car exist" check
- Rated against the market, not just "cheap"
- Invite-code access and a user limit
- A "nothing found" message so the bot never goes silent
// Architecture
How it works
// Result
The outcome
AutoHunter demo video (in Ukrainian) · 2:26
By moving the logic into the database, I cut Make usage from 158 to 39 credits per run. AI is called only for cars that have already passed selection.
// Want the same?
Need similar monitoring?
AutoHunter adapts easily to other marketplaces: real estate, electronics, tenders. Tell me what you need to track.