Ask a customer how they would like to reach you and, in a great many markets, the answer is WhatsApp. Email feels slow, a phone call feels like effort. The problem is that WhatsApp messages pile up in an inbox, and a question that arrives at midnight waits until morning.
Pick the right API first
Meta offers businesses three different things and they get confused with each other:
| Suits | Automation | Multiple users | |
|---|---|---|---|
| WhatsApp Business (app) | One-person business | No | No |
| Business Platform — On-Premise | Large enterprises | Yes | Yes |
| Business Platform — Cloud API | Most SMBs | Yes | Yes |
For nearly every small and mid-sized business the right answer is Cloud API. Meta hosts it, you do not run a server, and it is built for automation.
What you need
Before starting, have these ready:
- A Meta Business account. Free to create if you do not have one.
- A phone number. The critical part: it must not already be registered to another WhatsApp account. Do not use your personal number.
- Business verification. Not mandatory at low message volumes to begin with, but required to raise your limits.
- A display name. The name customers see; it goes through Meta's review.
Point two trips people up constantly. If your number is already on WhatsApp, you have to delete that account first, then connect the number to Cloud API.
The 24-hour rule
This is the most misunderstood part of WhatsApp integration, and it affects your costs directly.
The moment a customer messages you, a 24-hour service window opens. Inside that window you can reply freely, in your own words, as much as you like.
Once the 24 hours are up, you cannot send free-form text. To restart the conversation you must use a pre-approved template message.
The practical consequence: replying quickly is not only a customer satisfaction matter, it is a cost matter. Conversations that stay inside the window need no templates.
This is where AI has a concrete benefit — a message arriving at 2am does not wait until 9am, it gets answered before the window closes.
Bringing AI into it
The logic is the same as the website widget: build the knowledge base once, use it across channels. In Aihelpbot the site widget and WhatsApp share the same knowledge base, so you are not writing two answers to the same question.
The setup order:
- Connect your WhatsApp number through Cloud API.
- Choose which agent answers on that number.
- Once the webhook is verified, incoming messages start flowing.
- The AI replies; when a handoff condition is met, an operator is notified.
Things to watch
Keep replies short. The WhatsApp chat pane is narrow. A five-paragraph answer that reads fine on your site will not be read on WhatsApp. Add "answer briefly and directly" to the system prompt.
Formatting differs. WhatsApp does not support markdown; it has its own formatting rules. Long text with headings and lists renders badly.
Be honest about who is answering. Hiding the fact that a customer is talking to an AI may work short term, but it is a trust risk. Keep the handoff option visible.
Prepare template messages in advance. Get a few templates approved up front for when the 24-hour window has closed — approval is not instant.
Where to start
Rather than making WhatsApp your first channel, it tends to go more smoothly if you run the widget on your site first and let the knowledge base settle. Once it is mature, connecting WhatsApp is just adding a channel — the content is already there.