Custom AI chatbot or off-the-shelf chatbot?

The short answer

The first thing to ask is what the chatbot is actually for. If the answer is "answer the same ten questions on our website", an off-the-shelf tool will do that for a monthly fee and you can have it live this week. If the answer is "turn enquiries into booked calls with people who fit, without my team typing anything up", you need something connected to the rest of your business, and that's where a custom build pays for itself.

Side by side

Compared	Off-the-shelf chatbot	Custom AI chatbot
Time to launch	Days	Weeks, set out in the written scope
Upfront cost	Low	Projects start at £4,500
Ongoing cost	Monthly subscription, a lot of the time by conversation or seat	AI usage paid to providers, optional support
Answers from	The pages and FAQs you load in	Your approved documents, services and rules
Qualification	Basic form-style questions	Your real qualifying questions, and it sends people different ways based on the answers
Booking	A lot of the time just a link to a calendar	Books inside the conversation
Channels	Usually website chat	Web chat, SMS and social messages
CRM connection	Depends on the tool's integrations	Every conversation logged against the contact
Handover to a person	A general "contact us"	Rules you set, with the full conversation passed on
Ownership	You rent it	Built in your accounts

When off-the-shelf is enough

• Your questions are general and the answers are already on your website.
• You don't get that many chats and a person reads them all anyway.
• You don't need the chat to book people, check them or update your CRM.
• You want to see if visitors even use chat before you spend money on it.

If that's you, start with a tool, watch the conversations for a month, and you'll learn exactly what a better bot would need to do.

When a custom chatbot is worth it

• Enquiries come in from a few places. Website chat, SMS, Instagram and Facebook messages all need the same fast, right answer.
• Checking people matters. You need to know budget, timing, location or the service they want before a call gets booked, and send people different ways depending on what they say.
• The next step is a booking. A bot that sends a link loses people. One that books them while they're chatting keeps them.
• Your CRM is where the truth lives. Every conversation should land against the contact with a summary, so your team starts the call already knowing the details.
• A wrong answer costs you. If getting price, eligibility or policy wrong could cost you money or a complaint, you need tight control over what the bot is allowed to say.

The part both options get wrong: accuracy

A chatbot speaks for your business. In Moffatt v Air Canada (2024), a tribunal held the airline liable for wrong fare information its website chatbot gave a customer. So whichever way you go:

• Give the bot a small, approved set of information to answer from.
• Get it to say it doesn't know and pass the question to a person, and never guess.
• Set clear handover rules for complaints, sensitive topics and anything to do with money.
• Read a handful of conversations every week for the first month.

We build these rules into every chatbot we make, and they're on our conversational AI and chatbots page.

Questions to ask before you choose

1. What should happen after a good conversation? A booking, a quote request, a call back?
2. Which channels do your enquiries actually come in on?
3. What does your team ask before a call is worth booking?
4. Which answers should never be made up on the spot, like prices, policies or eligibility?
5. Where should the conversation be saved so your team can see it?

If your answers are short and simple, a tool will do you well. If they involve your calendar, your CRM and a few channels, talk to us. Our approach page shows how a build goes from audit to launch. And if most of your enquiries come in by phone, read about our AI receptionist instead.