A WhatsApp chatbot for an agrovet or farm input supplier should answer the questions that have one correct answer: prices, stock, opening hours, branch locations, order status and payment. It should not invent spraying advice. For anything about which chemical, what rate or how long before harvest, the bot should quote the PCPB-approved label word for word or hand the farmer to a trained person in the shop.
That split is the whole design. Get it right and the bot clears most of the planting-season queue on your phone. Get it wrong and a bot confidently tells a farmer in Kitale the wrong dose for a product it half-remembers.
Key Takeaways
- Automate the questions with one right answer (price, stock, location, order status) and route every agronomy question to a person or to the approved label text.
- Never let a general AI model generate pesticide doses or pre-harvest intervals; pull them from the PCPB register and its linked labels instead.
- Reply inside the 24-hour window where you can, because Meta does not charge for non-template replies sent in that window.
- Collect clear opt-in before sending planting reminders or offers, and stop messaging anyone who asks you to.
- If you run one shop from one phone, start with the free WhatsApp Business app; move to the paid platform only when the queue or your branch count demands it.
What can a WhatsApp chatbot actually do for an agrovet?
Most messages to an agrovet in March or September are repeats. Do you have DAP? How much is a 50kg bag of CAN? Are you open on Sunday? Has my order left Nakuru? These have one correct answer that lives in your stock list or price sheet, so a bot can answer them reliably if it reads from that list.
The harder messages are the valuable ones: a photo of yellowing maize, a question about fall armyworm, a farmer asking whether a fungicide is safe on tomatoes he plans to sell on Friday. Those need judgement, and in the case of agrochemicals, they carry legal and safety weight. That is where the bot should step back.
| Task | Can a bot handle it well? | What this means for you |
|---|---|---|
| Price and stock checks | Yes, if connected to your live stock list | Worth automating first, but a stale price list makes the bot wrong at scale |
| Opening hours, branch location, delivery areas | Yes | Easy win; set it up in a day |
| Taking an order and confirming it | Yes, with a confirmation step | Ask the customer to confirm items, quantity and pickup branch before payment |
| M-Pesa payment request | Yes, through Safaricom’s Daraja APIs | Needs a developer and a till or paybill; match payments to orders automatically |
| Order status and delivery updates | Yes | Utility messages inside the 24-hour window are free on Meta’s platform |
| ”Is product X registered for my crop?” | Yes, only if it reads from the PCPB register | Quote the registered use and link the label; never paraphrase from memory |
| Dose, mixing and pre-harvest interval | Only by quoting the approved label | Hand anything beyond the label to a trained staff member |
| Diagnosing a pest or disease from a photo | Not on its own | Treat any AI suggestion as a first guess for your agronomist, not an answer for the farmer |
My view: the bot’s job is to clear the queue so your trained staff spend their time on the hard questions, not to replace the person behind the counter.
Why must the bot never make up pesticide advice?
Because the label, not the chatbot, is the authority. Kenya’s Pest Control Products Board publishes an online register of products registered for use on crops, and each entry lists the registration number, active ingredient, local agent, registered uses (crops and target pests, with re-entry and pre-harvest intervals where they apply) and a link to the approved label.
In August 2026 the Kenya News Agency reported that PCPB had published this database with approved labels and urged users to check the label before buying or applying any pesticide. Labels show the approved use, dosage, first-aid measures, environmental precautions and safe handling, and the Board stressed following pre-harvest intervals after spraying.
The rules on who sells what are also strict. A PCPB compliance officer told KNA in November 2025 that crop protection products must only be sold through registered agrovet outlets, that labels must be in English and Kiswahili, and that a registered product carries a number starting with “PCPB (CR)”. In August 2026, PCPB’s head of compliance and enforcement warned agrovet dealers to check every product against the official register, and the report noted fines of Ksh100,000 to Ksh750,000 in cases that ended in convictions.
Now put a general-purpose AI model in front of that. These models are fluent and often right, but they also fill gaps with plausible numbers. We saw this ourselves running our own publishing desk: an AI writer will invent statistics and client stories unless checks stop it. A bot that invents a spray rate is the same failure, with a crop or a farmer’s health on the line.
So the safe pattern is retrieval, not generation:
- The farmer asks about a product or a pest.
- The bot looks up the product in your own copy of the PCPB register entries for the products you stock.
- It replies with the registered use and the label link, in the farmer’s language, and adds “Read the full label before use.”
- Anything not covered by the label goes to a named staff member in the chat.
If you want the bot to answer in Kiswahili or Sheng, test that carefully. Our guide on serving customers in Kiswahili with AI covers what to check before you trust a translation with anything technical.
How much does Meta charge for WhatsApp chatbot messages in Kenya?
If you use the WhatsApp Business Platform (the API that chatbots run on), Meta’s pricing documentation says it charges per message, a model in place since 1 July 2025. Kenya falls under the “Rest of Africa” market on Meta’s rate card, and charges depend on the country code of the recipient’s number. Meta publishes the exact rates as downloadable rate cards, and it may only change prices on the first day of a quarter, so check the current card before you budget.
| Message type | When Meta charges | What this means for you |
|---|---|---|
| Free-form reply (non-template) | Never; but only allowed inside the 24-hour customer service window | Most chatbot answers to farmer questions cost you nothing from Meta |
| Utility template (order confirmation, delivery update) | Free inside the window; charged outside it | Send order updates while the conversation is still open |
| Marketing template (planting reminder, offer) | Charged every time | Budget for every broadcast, and only send to farmers who opted in |
| Authentication template (one-time codes) | Charged | Rarely needed by an agrovet |
| Replies after a Click to WhatsApp ad | Free for 72 hours if you reply within 24 hours | Useful if you advertise in planting season |
The window opens when the customer messages you and closes 24 hours after their last message.
The separate WhatsApp Business app is free, but it is built for small businesses messaging from a single device, with limited integration with other tools. Your chatbot developer or platform provider will also charge for their service on top of Meta’s message fees.
Is a chatbot even allowed on WhatsApp in 2026?
Yes, for customer service. Meta’s WhatsApp Business Platform terms (last modified 23 September 2026) bar AI providers from using the platform when the AI itself is the main product on offer. An agrovet using AI to answer customer questions about its own products is a different case: the AI is incidental to your business, and the terms say you may retain an AI provider as your solution provider.
Two other rules in Meta’s WhatsApp Business Messaging Policy matter directly:
- Opt-in and opt-out. You may only message people who gave you their number and agreed to receive messages from you, and you must respect any request to stop, whether it comes on WhatsApp or off it.
- A way to reach a human. If you use automation, you must offer prompt, clear and direct escalation paths, such as in-chat transfer to a person, a phone number or an in-store visit.
One more to check before you build a catalogue. The same policy lists “hazardous goods and materials” among restricted goods you may not buy, sell, promote or facilitate on WhatsApp. We have not found Meta guidance on how this applies to PCPB-registered crop protection products, so ask your platform provider before you put pesticides into a WhatsApp catalogue or marketing broadcast. Fertiliser, seed and animal feed price checks are a safer place to start.
Should an agrovet use the WhatsApp Business app or the platform?
Honest answer: most single-branch agrovets should start with the free app. It already gives you a product catalogue, automatic greeting and away messages, quick replies and chat labels. A well-built set of quick replies for your top 20 questions, plus an away message with your hours and location, takes a lot of pressure off the counter phone.
Move to the platform when one or more of these is true: you have several branches sharing one number, more than one person needs to answer at the same time, you want orders and M-Pesa payments matched automatically, or you want the bot to read live stock levels. Meta also supports “coexistence”, where a number stays on the Business app while connected to the Cloud API, with messaging history kept in sync between both. It requires app version 2.24.17 or higher, so confirm with your provider that it works for your number.
Our WhatsApp Business API setup guide for Kenyan SMEs walks through registration and choosing a provider.
How do orders and M-Pesa payments work inside the chat?
The usual flow looks like this: the farmer picks items, the bot reads back the order with prices, the farmer confirms, and the system sends an M-Pesa payment request through Safaricom’s Daraja platform, which gives developers access to M-Pesa APIs. When payment comes through, the bot confirms the order and tells the farmer when and where to collect.
The parts people underestimate are not the payment itself. They are matching each payment to the right order, handling the farmer who pays the wrong amount or pays from a different number, and keeping the stock list accurate so the bot never sells a bag you do not have. Our breakdown of M-Pesa Daraja integration covers what the setup involves, and AI inventory management for Kenyan shops covers keeping the stock side honest.
If payments are new to you, start with order capture and confirmation, let staff send the M-Pesa till number by hand for the first month, and automate payment only once orders flow cleanly.
Worked example: an agrovet in Kitale in the long rains
This is a hypothetical example with stated assumptions, not a client result.
Assumptions: a two-counter agrovet in Kitale gets 120 WhatsApp messages a day in the March planting rush. Each message takes staff about 3 minutes to read, check and answer. The mix is 40% price and stock, 20% hours and location, 15% order status and 25% product or pest advice.
Total staff time today: 120 messages x 3 minutes = 360 minutes, or 6 hours a day.
What the bot can take: price and stock (40%) + hours and location (20%) + order status (15%) = 75%. That is 120 x 0.75 = 90 messages a day.
Time freed: 90 x 3 minutes = 270 minutes, or 4.5 hours a day. The remaining 30 advice messages (120 x 0.25) still go to staff, who now have time to answer them properly.
Meta’s cost for those replies: close to zero, because the farmer messaged first and the bot replied inside the 24-hour window with free-form messages.
The marketing side: if 600 farmers have opted in and you send one fertiliser top-dressing reminder, that is 600 marketing templates. Your cost is 600 x the current Rest of Africa marketing rate on Meta’s card.
The caveat: the 75% only holds if your price and stock list is updated daily. A bot quoting last week’s price for 90 farmers a day creates 90 arguments at the counter.
How to set up a WhatsApp chatbot for your agrovet this week
- Count your messages. For three days, tally every WhatsApp message into five buckets: price, stock, location and hours, order status, advice. This tells you what the bot should handle.
- Write your top 20 answers. Turn the most common questions into clear quick replies in English and Kiswahili. Put them into the free Business app today.
- Build your product sheet. One spreadsheet with product name, pack size, price, stock status and, for every crop protection product, its PCPB registration number and label link from the register. Remove anything you cannot find on the register.
- Decide your hand-off rule. Write it down: any question about dose, mixing, a sick crop or a product not on your sheet goes to a named person. Add that person’s name and the shop number to the bot’s reply.
- Set up opt-in. Add a line to receipts and your shop poster: “Reply YES to get planting-season reminders on WhatsApp.” Keep a record of who said yes.
- Choose app or platform. If one phone and one branch is coping with quick replies, stay there. If not, speak to a provider about the platform and coexistence, using the decision points above.
- Test before launch. Have three staff send the bot the 30 hardest real questions from your tally. Fix every wrong or vague answer before a farmer sees it.
For a wider view of what chatbots cost and how they are built, see our guide to AI chatbot development in Kenya.
Mistakes to avoid
- Letting a general AI model answer agronomy freely. It will sometimes produce a confident, wrong dose. Quote the label or hand over.
- A price list nobody updates. The bot repeats the error to every customer, all day.
- Broadcasting to everyone in your phone book. Meta requires opt-in, and every marketing message costs money.
- No way to reach a person. Meta’s policy requires an escalation path, and farmers with a sick crop will not wait for a menu.
- Listing products you have not checked on the PCPB register. The bot then advertises a product you should not be selling.
Frequently asked questions
Can a WhatsApp chatbot recommend pesticides to farmers?
It can point a farmer to products registered for their crop and pest, if it reads from the PCPB register and links the approved label. It should not generate doses, mixes or pre-harvest intervals on its own. Anything beyond what the label says belongs with a trained staff member.
Is a WhatsApp chatbot free for a small agrovet?
The WhatsApp Business app is free and covers quick replies, greeting and away messages and a catalogue. The WhatsApp Business Platform, which full chatbots use, has no access fee from Meta, but Meta charges per template message and your provider charges for their service. Replies inside the 24-hour window are free from Meta.
Can the chatbot take M-Pesa payments?
Yes, through Safaricom’s Daraja APIs, which let the system request payment and confirm the order once paid. You need a till or paybill and a developer to connect it. Plan for wrong amounts and payments from other numbers, because those are where most problems come from.
Can the bot reply in Kiswahili?
Yes, and for many farmers it should. Fixed answers such as prices and hours are easy to translate once and reuse. Technical answers about products need a human-checked translation, and the approved label itself is already in English and Kiswahili, so link to it rather than rewording it.
What happens if a farmer asks something the bot cannot answer?
It should say so plainly and pass the chat to a named person, or give the shop number. Meta’s policy requires a clear escalation path for automated chats. A bot that guesses instead of handing over is worse than no bot.
The bottom line
A good agrovet chatbot is a fast, accurate counter clerk for the easy questions and a polite messenger for the hard ones. Start with your real message tally, the free app and a product sheet tied to the PCPB register. Add the platform, ordering and M-Pesa once the basics work.
If you would like a second opinion on whether your shop needs a chatbot at all, or help scoping one, WhatsApp us on 0711 344 702.