Three things a farmer needs from a dealer, done by AI
Look at the plant. Say what to do. Show how to do it. Agro Replay does all three through the channel farmers already use, under the dealer’s name.
Scan — photo diagnosis
The farmer photographs a leaf, stem, fruit or root. A vision model classifies disease, pest damage and nutrient deficiency, and returns a confidence score. Low-confidence cases are routed to the dealer with a suggested question rather than a guess.
Launch crops: rice, wheat, maize, soybean, tomato and potato — the crops most smallholders grow. Regional crops are added as labelled field images are collected from pilot dealers.
Ask — advisory in the farmer’s language
The advisor answers questions about what to apply, when, how much and what to avoid, by text or voice. It reasons over crop and season guides, the local weather and the dealer’s own product catalogue, and it answers in the farmer’s own language — 80+ supported, from Spanish, French and Portuguese to Hindi, Bangla, Swahili, Indonesian and Vietnamese.
Every recommendation stays inside the rules the dealer sets in the console: approved products, dose ranges, safety intervals and the cases that must be escalated to a human.
Show — a short video for every answer
Text is hard to act on in a field. For any recommendation the advisor can generate a 30–60 second video in the farmer’s language — how to mix, when to spray, what safety gear to wear — using the same AI video pipeline our team built for FacelessClip AI. Videos are rendered in batch and delivered over WhatsApp, which works in 180+ countries, so they play on a basic smartphone and can be shared with neighbours.
Dealer console
A web dashboard for the dealer or cooperative: upload the product catalogue, set rules, review escalated cases, see which crops and problems are trending in their area, and keep a record of every farmer conversation.
Data is isolated per dealer. Farmers’ photos and messages are used to answer them and to improve the dealer’s own service, and are never sold.
How it is built
Agro Replay is a pipeline of specialised models rather than one chatbot. Photo diagnosis runs on a vision model; the advisor is a large language model with retrieval over agronomy guides and the dealer catalogue, routed through Oxyy AI so each request goes to the best-fit model for that language and task; speech-to-text and text-to-speech handle voice in 80+ languages; video generation runs as batch jobs. A Postgres database keeps each dealer’s data separate.
Where we are
| Milestone | Status | Timing |
|---|---|---|
| Advisor chat + photo diagnosis (launch crops) | In development | 2026 Q3 |
| Dealer console | In development | 2026 Q3 |
| First dealer pilots | Planned | 2026 Q4 |
| Video advisory at scale, regional crop packs | Roadmap | 2027 |
Updated as milestones are reached.
What dealers ask first
Does this replace an agronomist?
No. It handles the routine questions and photo checks that never reach an agronomist today, and it escalates anything uncertain to the dealer or their expert.
Who decides which products get recommended?
The dealer. The advisor recommends only from the catalogue and rules set in the dealer console, and it explains why.
Which languages and crops?
80+ languages, wherever WhatsApp works (180+ countries). Launch crops are rice, wheat, maize, soybean, tomato and potato; regional crops are added as we collect labelled field images with pilot dealers.
Do farmers need to install anything?
No. It works over WhatsApp. A lightweight app is available for dealers who prefer it.
What does it cost?
A per-dealer subscription, priced by number of farmers served. Pilot partners get the first season on special terms.
Run the first pilot with us
Input dealers and cooperatives anywhere: tell us your country, region and main crops.