AI chatbot development: website, WhatsApp, support and sales bots

By Ranjit Bhinge, founder of Blur Studio · Updated October 23, 2026 · 6 min read

Short answer: AI chatbot development services build bots that answer customers or staff on your website, WhatsApp or help desk, using AI that reads your own policies and data. If a ready-made tool works with your help desk and your questions are simple, buy it. Build a custom bot when it must connect to your own systems, follow strict rules or run sales steps that no tool supports.

You're looking into AI chatbot development and want to know what actually works, not just what a vendor promises.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build apps, business software, and AI agents and workflows, one project at a time. I also work differently from most of what else is out there, more on that further down this page.

An AI chatbot can answer customers at night, sort new leads and take simple jobs off your team. A bad one can give wrong answers in your company's name. The difference comes down to how it is built, what it can see and how it is tested.

This guide covers website, WhatsApp, support and sales bots. It explains when to buy a ready-made tool, when to build, what to check before launch and how to choose a chatbot development company.

What does an AI chatbot do for a business?

An AI chatbot is software that chats with people in plain language. It uses a large language model, or LLM, which is the kind of AI behind ChatGPT. A business bot also reads your own content, so it can answer from your rules and data.

Most business bots fall into one of five types.

Website bot: Answers visitor questions, collects contact details and books calls.
WhatsApp bot: Replies to customers on your business WhatsApp number, sends reminders and passes chats to staff.
Support bot: Answers common questions from your help center, checks order status and opens tickets.
Sales bot: Asks qualifying questions, suggests the right product and sends hot leads to a salesperson.
Staff bot: Answers employee questions about policies, steps and products from internal files.

How are AI chatbots different from older chatbots?

Older chatbots followed fixed menus. You clicked a button, and the bot showed the next set of buttons. They rarely said anything wrong, but they could not handle a question outside the menu.

AI chatbots understand free text. A customer can type a messy question and still get a useful answer. The risk is that an AI bot can also say something wrong or make something up. So a good AI bot answers from your own content and hands hard questions to a person. My guide to AI that answers from your own data explains the method most good bots use.

Should you build a chatbot or buy a ready-made tool?

Many good chatbot tools exist. For example, Intercom sells an AI agent called Fin. Intercom says Fin works across chat, email, voice, social and other channels. It also says Fin works with help desks such as Zendesk, Salesforce, Freshdesk and HubSpot. Intercom prices Fin on outcomes, so you pay for results.

Here is how the two paths compare.

Question	Ready-made tool	Custom build
Speed to launch	Fast, if it fits your help desk	Slower, since it is built for you
Fit with your systems	Works with popular tools	Works with your own database, app or booking system
Control over answers	Set by the vendor's options	Your rules, your logs, your tests
Cost over time	Ongoing fee to the vendor	Build cost plus AI usage and care
Who owns it	The vendor owns the software	You own the code and accounts

Buy a tool when your questions are common and your help desk is a popular one. Build when the bot must read live data from your own systems. Build when it must follow strict rules, or when it runs sales steps no tool supports. If you are weighing AI builder tools against a developer, see my guide to AI app builders vs hiring a developer.

What should you know about WhatsApp chatbots?

WhatsApp changed its business terms in 2025. From January 15, 2026, the terms bar general-purpose AI assistants from using the WhatsApp Business Solution. TechCrunch reported that this does not affect businesses using AI to serve their own customers.

So a bot that answers questions about your own products, orders or bookings is fine. A bot that acts as an open AI assistant for any topic is the kind Meta now blocks. Keep your WhatsApp bot focused on your business.

A business WhatsApp bot runs on the WhatsApp Business Solution, the business system these terms cover. Check the current terms before you launch, because Meta decides what counts as a general-purpose assistant.

What rules apply to business chatbots?

You are responsible for what your bot says

In February 2024, a Canadian tribunal ruled against Air Canada. Its website chatbot gave a customer wrong details about a refund. The airline argued that the chatbot was a separate legal entity. The tribunal disagreed and said the airline was responsible for all the information on its website.

California bans bots that hide what they are

California has a bot disclosure law. It bans using a bot to mislead a person in California about its artificial identity to sell something. You avoid this if the bot clearly says it is a bot. The law says the notice must be clear and easy to see. A short line at the start of the chat, such as "I'm an AI assistant," makes this clear.

Do not overclaim

The Federal Trade Commission (FTC) has acted against companies that made false AI claims. Its head said in 2024 that there is "no AI exemption" from current laws. Do not tell customers your bot can do more than it can.

What should you check before a chatbot goes live?

Run through this list before launch.

It answers from your content: Test it with real customer questions and compare each answer with your policies.
It hands off to a person: Customers can reach a human, and the person sees the full chat.
It says it is a bot: The first message makes this clear.
It refuses off-topic tasks: The bot stays on your business and does not act as a general AI.
It resists tricks: The OWASP Top 10 for LLM apps lists prompt injection as the top risk. That means messages written to trick the bot into breaking its rules. Ask your builder to test for it.
It protects data: The bot only sees the customer data it needs, and it never shows one customer's data to another.
It keeps logs: You can read past chats, find bad answers and fix them.

How do you choose a chatbot development company?

Ask them to show you a live bot they built, and try to break it. Then ask these questions.

How will the bot learn our policies, products and prices?
How will you test its answers before launch?
What happens when the bot does not know the answer?
Which systems will it connect to, such as our CRM or booking tool?
What will it cost each month to run at our chat volume?
Who owns the code, the AI accounts and the chat logs?

Red flags

No handoff plan: The bot traps customers with no way to reach a person.
No tests: They cannot show how often the bot gets answers right.
Their accounts: The bot runs on their AI keys and their servers, under their name.
Fully automatic sales promises: A bot can qualify leads well. It should not quote custom prices or sign deals alone.

When should you not build an AI chatbot?

Skip it if you get only a few questions a week. A good FAQ page and a contact form may be enough. Skip it if your policies are not written down, because the bot will have nothing reliable to answer from. Start with a ready-made tool if your help desk already offers one that fits.

How does Blur Studio approach chatbot builds?

For chatbots, I start with your real questions. I gather past chats, emails and tickets and turn them into a test set. Then I build the bot to answer from your own content and test it against that set.

I plan the handoff to your team first, so no customer gets stuck. I connect the bot to the systems it needs and nothing more. The bot's code, AI keys and chat logs all belong to your business. After launch, I can take care of the bot, read the logs and improve its answers. If the bot needs to take actions, such as booking or updating records, see my guide to AI agent development.

Questions people ask

How is an AI chatbot different from an AI agent?

A chatbot mainly answers questions in a chat. An AI agent can also take steps by itself, such as updating a record or booking a slot. Many bots start as chatbots and add a few safe actions later.

Can I put an AI chatbot on WhatsApp?

Yes, through the official WhatsApp Business Solution. Since January 15, 2026, WhatsApp's terms bar general-purpose AI assistants. A bot that serves your own customers about your own business is still allowed.

Does my chatbot have to say it is a bot?

California law bans using a bot to mislead people in California about being a bot in order to sell. A clear note that the customer is chatting with AI avoids this problem. It is a good habit in every state.

Am I responsible if my chatbot gives a wrong answer?

You can be. In 2024, a Canadian tribunal held Air Canada responsible for wrong refund details its website chatbot gave a customer. Test your bot well and let it hand hard questions to a person.

Should I build a custom chatbot or use a tool like Intercom Fin?

Use a ready-made tool if your questions are common and it works with your help desk. Build a custom bot when it must use live data from your own systems or follow rules a tool cannot handle.