AI chatbot vs AI voice agent for customer enquiries: which do you need?

The short comparison

Compared	AI chatbot	AI voice agent
Where it works	Website chat, forms, messages, email	Phone calls, inbound and outbound
Best for	Enquiries that arrive in writing, often out of hours	Businesses where customers prefer to call
Handles	Questions, qualifying, booking, follow-up	Questions, qualifying, booking, routing calls
Hands to a person	By passing the conversation to your team	By transferring the call or taking a message
What the customer needs	To be willing to type	Nothing. They just call your number
Running costs	AI usage, billed by the provider	AI usage and phone minutes, billed by the providers
Extra rules	Tell people it is automated	Tell callers it is automated. Calls the agent makes have a consent rule, covered below

The two are closer than they look. Underneath, both work from the same approved information and the same booking rules. The difference is the channel, and the channel is your customers' choice, not yours.

How do you decide which to build first?

1. Count one normal week. Write down every call, form, message and email. Start where most of them are.
2. Count what you miss. Missed calls and voicemails point to a voice agent. Forms answered the next morning point to a chatbot.
3. Note when they arrive. Evening and weekend enquiries are the ones nobody is there to answer, so they weigh more.
4. Ask what your customers prefer. Some customers always call. Others never will.
5. Pick the bigger leak. Build that one first, then add the second once the first is in use.

A worked example

This is an illustration, not a client. A clinic counts one week and finds 40 forms and messages and 12 calls. Most of the forms arrive in the evening and are answered the next morning. Three of the calls are missed. The written enquiries are the bigger leak, so the chatbot goes first.

Turn the numbers round. If the same clinic had 12 forms and 40 calls, with 15 calls missed while the team were with clients, the voice agent would go first. The count decides, not the technology.

When is each one the wrong choice?

A chatbot is the wrong choice when your customers mostly phone. A chat window on a website does nothing for someone who is driving, or who simply prefers to speak to someone.

A voice agent is the wrong choice when few calls come in and someone already answers them. It is also the wrong tool when most calls are complaints or sensitive conversations. Those need a person from the first word.

Neither is worth building yet if you get a handful of enquiries a week and answer each one within the hour. The same goes if your prices, services and policies are not written down anywhere. An AI can only repeat information you have approved, so write that down first. It costs nothing, and your own team will use it too.

What do both need to work well?

• Approved information only: prices you are happy to share, services, opening times and policies.
• A clean handover: anything sensitive, unusual or upset goes to a person, with the conversation attached.
• Your calendar and CRM connected: so bookings and notes land where your team works.
• Honesty: both should say they are automated.
• A person on the decisions that matter: money, legal commitments, complaints and regulated advice stay with your team.

You are responsible for what an AI says in your name. In 2024 a Canadian tribunal held Air Canada liable for incorrect fare information its website chatbot gave a customer, and rejected the argument that the chatbot was responsible for its own actions (Moffatt v Air Canada). Test properly before customers use it.

What are the rules for automated calls?

In the UK, automated marketing calls need the person's specific prior consent. General marketing consent is not enough (ICO guidance on telephone marketing).

That is why we treat calls the agent makes as a separate decision in the scope, not a switch to turn on later. If you want a voice agent to ring people, sort out the consent before anything is built.

What does each cost?

Projects start at £4,500. The price depends on the channels covered, how many systems it connects to and, for a voice agent, whether it makes calls as well as answering them.

Running costs are separate and are paid directly to the providers from your own accounts: AI usage for both, plus phone minutes for a voice agent. We estimate them before building. Ongoing support is optional, as a monthly retainer priced separately.

How we approach it

1. Audit. We map how enquiries arrive, the tools you use and where they get missed.
2. Design. You get a written scope: channels, integrations, handover rules and how success is measured.
3. Build. Inside your own accounts, keeping the tools that already work.
4. Test. Real questions, edge cases and failure paths before a customer talks to it.
5. Launch. In stages, with your team trained and the documentation handed over.
6. Improve. We review the numbers and agree the next changes.

Read more about how we work, our AI receptionist, and custom vs off-the-shelf chatbots. If you are not sure enquiries are the right place to start, what to automate first has a simple scoring test.