In Israel, WhatsApp is often not "another channel". It is sales and support. A good bot is not a cute three-button menu. It is a flow that respects Meta's limits, the person in the field, and the customer who already typed "hi?" for the fourth time.
I build WhatsApp bots, and Discord or Telegram when that is your model. Business API, message templates, approvals, a hook into your site or CRM. If you searched WhatsApp bot developer and got a prompt workshop, that is a different job.
What it does once it is live
Answers that repeat. Booking. Orders. Delivery status. Payments when the business is ready. And a way out to a human when the bot should not guess. Without that exit, people get stuck, and you get yelling instead of leads.
Wiring inventory, a calendar, or a sheet is usually the expensive and important part, not the "hey, how can I help" line. I wrote about those connections in integrations and APIs.
Before you ask what a bot costs
The same word can mean a simple menu, or a system with billing, a desk, and three vendors behind it. Price jumps because scope jumps. The breakdown is in the WhatsApp bot pricing article. Bring to the call what already happens in your chat. We start there, not from a slide.
Who it fits
It fits when the same questions keep coming: clinic, shop, club, field service. It is a poor fit if you want "AI that sells by itself" with no sales process. A language model can draft a reply. It does not replace knowing who the customer is and what you are allowed to promise.
Questions from real builds
Do we need WhatsApp Business API?
For a serious bot, usually yes. A home hack is not a Meta-compliant business number. On a call we pick what matches your volume, without selling the expensive package because it sounds impressive.
How long does it take?
Menus and one clear flow: days to weeks. Payments, a desk, and three systems behind it: depends on what you already have, and how fast someone on your side can approve templates and test on a real phone.