WhatsApp Business Bot: What It Does and How to Deploy One
A WhatsApp business bot answers leads and support requests and updates your ERP or CRM. Learn how it works, what it costs and how to set one up.

A WhatsApp business bot is software connected to your company’s WhatsApp number that receives customer messages, answers them and, in its more capable form, acts on them: it qualifies a lead, checks an order, books a meeting or updates a record in your ERP or CRM. The value depends less on the chat window than on what sits behind it. A bot that only returns scripted replies shortens the first response, while an agent tied to your data and rules removes the follow-up work that stalls sales and support. We build the second kind, custom agents with human handoff and auditable logs, on the WhatsApp agent plans that start at $20 per month.
TL;DR
- A WhatsApp business bot connects to the WhatsApp Business Platform, receives messages through a webhook and replies automatically, within a 24-hour window or through pre-approved templates.
- Rule-based bots follow menus, AI chatbots generate answers, and AI agents also take actions in business systems. The last is what removes workload.
- Meta charges businesses per message under the Business Platform pricing, and service messages become chargeable from October 1, 2026, after the first 1,000 per number each month.
- Judge a bot on integrations, human handoff, audit logs and language coverage, not on how fluent its replies sound.
- Our WhatsApp agents are priced at $20 to $499 per month, with one-time setup fees of 1,500 € and 2,500 € listed for the Starter and Growth tiers.
Table of Contents
- What a WhatsApp business bot is
- How a bot connects to WhatsApp
- What it costs: Meta fees plus the bot itself
- What WhatsApp automation can handle
- How to deploy a WhatsApp business bot
- What to evaluate before you choose
- Our view
- Where to start
- FAQ
- Sources
What a WhatsApp business bot is
WhatsApp offers two products for companies. The WhatsApp Business app is a free app for small businesses that answer messages by hand. The WhatsApp Business Platform is the programmatic route: according to WhatsApp’s official FAQ, it lets medium and large businesses chat with customers at scale, with rich messaging such as media and quick-reply buttons. A bot lives on the Platform, because only the Platform exposes an interface (the Cloud API) that software can send messages through.
Three terms are used loosely in this market, and the differences decide what you get:
| Type | How it decides what to say | What it can do | Typical limit |
|---|---|---|---|
| Rule-based bot | Fixed menus, keywords and triggers | Routes customers, sends canned answers, collects form fields | Fails on any phrasing it was not scripted for |
| AI chatbot | A language model writes the reply from a knowledge base | Answers free-text questions in many languages | Talks about your business but does not change anything in it |
| AI agent | A language model plus rules, tools and system access | Reads and writes ERP or CRM data, follows up, escalates with context | Needs design work: data mapping, permissions, logging |
Most searches for a “WhatsApp chat bot” are looking for the first two rows. Most of the operational payoff sits in the third.
How a bot connects to WhatsApp
The mechanics are the same whichever vendor builds the bot, and knowing them helps you ask better questions.
- A business phone number is registered on the Platform. It becomes the number customers message.
- Customers must opt in. Meta’s developer documentation states that you can only message users who have opted in to hear from you, as described on its service messages page.
- A webhook receives each inbound message. Meta sends your server an HTTP request when a customer writes. A technical walkthrough of the Cloud API explains that the server should acknowledge quickly, otherwise Meta retries and the same message can be processed twice.
- The bot decides and replies. The reply goes back through an authenticated request to Meta’s Graph API.
- The 24-hour window limits free-form replies. When a customer writes, a 24-hour customer service window opens and resets with each new customer message. Inside it you can send any free-form message. Outside it, Meta’s documentation says you can only send pre-approved template messages.
The window rule shapes bot design. A lead who asks about pricing on Monday and goes silent can be followed up on Tuesday afternoon only with an approved template, so follow-up wording has to be prepared and submitted in advance, not improvised.
What it costs: Meta fees plus the bot itself
There are two separate cost lines, and comparisons that blur them mislead buyers.
Meta’s messaging fees. Per the WhatsApp FAQ, Meta has charged businesses per message since July 1, 2025, with rates that depend on the category: marketing, utility, authentication, service or Meta Business Agent. The same page states that from October 1, 2026 service messages are also charged per message, with the first 1,000 service messages free per business phone number each month and charges from the 1,001st. Rates vary by the recipient’s country, so the current rate card on WhatsApp’s pricing page is the place to check before you forecast a budget. Meta updates it on its own schedule.

A quick way to estimate exposure: count monthly conversations by category. Order status replies and support answers are service or utility traffic. Promotions and re-engagement campaigns are marketing and cost the most per message. A bot that resolves issues in fewer messages also lowers the Meta bill, which is a practical argument for good conversation design.
The bot or agent. This is the build and run cost: platform subscription, setup, integrations and maintenance. Our listed WhatsApp agent plans are:
| Plan | Price |
|---|---|
| Starter | from $20 per month |
| Plus | $99 per month |
| Growth | from $249 per month |
| Max | $499 per month |
| Enterprise | no published price |
One-time setup is listed at 1,500 € for Starter and 2,500 € for Growth on our voice and WhatsApp agent page, and Enterprise setup is scoped individually. Pricing for custom work depends on workflow complexity and usage, so integrations into ERP or CRM systems are scoped in a consultation, not read from a rate card.
What WhatsApp automation can handle
Cover the tasks by department, since that is how requests arrive.
Sales follow-up
Speed to first reply is the sales use case. An agent answers an inbound enquiry within seconds, asks the qualifying questions you define (budget, timeline, product line, location), writes the answers to the CRM and hands a prepared summary to a salesperson. After the first contact it sends the follow-ups that people forget: a quote reminder, a meeting confirmation, a document request. The Voice AI versus chatbots guide on our site helps you decide when a call suits a lead better than a message.

Customer support
Repetitive questions are the obvious target: opening hours, delivery status, returns policy, invoice copies. The best-known example is “where is my order?”, which an agent can answer by reading order status from the system of record, as we describe in our WISMO article. The same pattern works for appointment changes and warranty questions. Anything the agent cannot settle under its rules goes to a person with the conversation attached.
Operations and back office
This is where an agent differs most from a chatbot. Examples include:
- Collecting missing documents during onboarding and chasing them on a schedule, covered in our piece on automated document collection.
- Updating order status, delivery dates or contact fields in an ERP or CRM after a customer confirms them.
- Extracting data from images or files customers send, such as invoices, with OCR.
- Producing daily reports on open conversations, response times and unresolved cases.
Sector examples
The patterns are shared, the rules are not. A jewellery retailer needs product enquiries and appointment booking, described in our guide to WhatsApp automation for jewellery stores. A property firm needs lead qualification and viewing scheduling, which our real estate chatbot page outlines. Online sellers have their own order, return and delivery flows, covered in AI agents for ecommerce.
How to deploy a WhatsApp business bot
The sequence below is what a competent deployment follows. At each step we note where projects commonly go wrong and what our approach takes off your team.
1. Pick the process, not the tool
Start with a measurable problem: first-response time on leads, share of support tickets that are status questions, hours spent keying orders into the ERP. Projects that begin with “we need a chatbot” end up with a bot that talks and changes nothing. We begin with an assessment of which processes suit automation, offered as AI automation consulting.
2. Secure Platform access and a number
You need a Meta business account, a verified business and a phone number that is dedicated to the Platform. The common failure is attempting to reuse a number already active on the consumer or business app, which complicates migration. Plan the number early because it is the one thing customers will see.
3. Map the data and the rules
Decide which systems the agent may read, which it may write to, and which actions need human approval. Rules mapped to your actual data are what keep an agent from inventing a delivery date. Our agents act on rules tied to your own data and processes, with every action recorded in an auditable log.
4. Prepare templates and opt-in
Write the message templates for follow-ups, reminders and notifications and submit them for approval, since they are the only way to reach customers after the 24-hour window closes. Record how each contact opted in. Templates and opt-in wording are the usual cause of delays at launch.
5. Build the conversation flows and handoff
Define what the agent does when it is unsure, when the customer is upset, and when a request is outside its rules. Good handoff passes the full conversation and any prepared draft to a named person or queue. The agent can reply in over 70 languages, so language does not need a separate team.
6. Test on real conversations, then roll out in phases
Replay historical enquiries, including the messy ones with typos, mixed languages and voice notes. Start with one process or one team, review the logs weekly and widen the scope once the numbers hold. Pilot and phased rollouts are part of how we scope projects.
7. Operate and measure
Track first-response time, resolution without handoff, escalations by reason and conversions from lead to meeting. We stay accountable for the workflow’s outcome after launch, which means the agent is reviewed and adjusted against those figures and not left running unattended.
What to evaluate before you choose
Use the criteria below to compare any approach, from an off-the-shelf tool to a custom agent. Our guide How to Choose an AI Agent goes through seven questions in more detail.
| Criterion | What to look for |
|---|---|
| Integration depth | Reads and writes ERP or CRM fields, not just sends a notification |
| Human handoff | Passes the full context and a prepared draft to a person |
| Auditability | A log of every message, decision and system change |
| Rules over improvisation | Actions limited by rules mapped to your data |
| Language coverage | Native-quality replies in your customers’ languages |
| Channel coverage | WhatsApp plus voice and email when customers switch channel |
| Accountability | Someone responsible for outcomes after go-live |
A bot that scores well on the first row and poorly on the last six is a chatbot with a database lookup. The quality of the answer is the cheapest thing to get right, while integration, handoff and logging are where deployments succeed or fail.
Our view
We build WhatsApp agents as part of a business workflow, not as a separate chat product. The question we ask first is which process is slow, and what a person would need in front of them to finish it faster.
That is why our agents are wired into the ERP or CRM, act on rules mapped to the company’s own data, and hand over to people with the context attached. They reduce workload without replacing the people who hold customer relationships.
The measure of success is outcome, not message volume: faster lead response, fewer status requests reaching staff, cleaner records. Each deployment is therefore scoped around a workflow, logged so that every action can be audited, and reviewed after launch.
Suyash
Where to start
If your team already knows which workflow it wants to automate, the fastest route is a scoped build. Our WhatsApp plans run from Starter at $20 per month to Max at $499 per month, with Enterprise scoped individually, and our AI agent development service covers custom integrations with your ERP or CRM.

If you want to see how an agent sounds and behaves first, the live demos let you try one in the browser. For a recommendation tailored to your process and systems, we offer an assessment through our automation consulting service, and the full range is on voxdonna.com.
FAQ
Can I use a WhatsApp business bot with the free WhatsApp Business app?
The free app is built for answering messages by hand. Bots and agents connect through the WhatsApp Business Platform, which gives software programmatic access to messaging. If your volume or your need for CRM integration has outgrown manual replies, the Platform is the route.
Why can the bot not message a customer whenever it wants?
Meta limits free-form messages to the 24-hour customer service window that opens when a customer writes. After it closes, only pre-approved templates can be sent, and only to users who have opted in. Prepared templates for reminders and follow-ups solve this in practice.
How much will Meta charge for the messages?
Meta charges per message by category and recipient country. Its FAQ states that service messages become chargeable from October 1, 2026, with the first 1,000 per business phone number each month free. Check the live rate card for your target countries, since it changes on Meta’s schedule.
Will a bot replace my support or sales team?
No. Our agents are designed to reduce workload, not to replace people. They take repetitive questions, follow-ups and record updates, and pass anything outside their rules to a person with the conversation context.
Does the bot support more than English?
Our agents work in over 70 languages and run around the clock, so customers can write in their own language at any hour. Rules and handoff paths are defined per workflow so behavior stays consistent across languages.
Sources
- WhatsApp Business Platform FAQ
- WhatsApp Business Platform pricing
- Meta developer documentation: service messages and customer service windows
- Building a WhatsApp chatbot with the Cloud API: a Node.js guide