Voice agents and chatbots are not the same thing

Both are powered by the same kind of language model. Both answer customer questions. On a slide they look interchangeable. In practice they solve different problems for different customers, and cost very different amounts to run.

The customer is in a different mood

Someone typing into a web chat is browsing. They are relaxed, they are comparing options, and they will tolerate a pause while the system thinks. If it is unhelpful they shrug and close the tab.

Someone on the phone is usually trying to resolve something now. They are frequently mildly irritated before you even answer. They will not tolerate a three-second silence, and if the experience is poor they will remember it and tell people.

That difference in expectation drives almost everything else.

Voice is technically much harder

Text arrives cleanly. Speech does not. A voice agent has to cope with regional accents, a lorry going past, a toddler, someone on a hands-free in a tunnel, and people who start a sentence three times before committing to it.

It also has to handle interruption. Real conversations involve people talking over each other, and an agent that ploughs on regardless feels immediately robotic.

And it must be fast. The comfortable gap in human conversation is a few hundred milliseconds. Exceed about a second consistently and the whole thing feels broken, no matter how good the answers are.

The cost difference is real

Web chat is inexpensive to run. Costs are measured in pennies per conversation and it scales almost for free.

Voice involves converting speech to text, generating a response, converting text back to speech, and holding a telephony connection open throughout. It typically costs several times more per conversation.

That said, the comparison that matters is not chat versus voice. It is voice versus the cost of a missed call. For a business where an average job is worth several hundred pounds, one recovered enquiry a week pays for the whole thing several times over.

When voice is the right answer

Choose voice when your customers phone you, when calls are being missed, when enquiries arrive outside working hours, or when your audience simply does not use web chat. Trades, clinics, lettings, restaurants and anyone with an older customer base usually fall here.

When chat is the right answer

Choose chat when your customers are researching before they buy, when questions are detailed and benefit from links and images, when you want the assistant to work alongside a browsing experience, or when volume is high and margins are thin.

Retail, software and considered-purchase businesses usually fall here.

The uncomfortable truth about doing both

Most businesses eventually want both, and the mistake is treating them as two projects. Build the knowledge base, the integrations and the escalation rules once, then put two front doors on it. Otherwise you end up maintaining two systems that gradually diverge, and answering the same question differently depending on which channel the customer used.

Start with whichever channel your customers actually use today. Add the second when the first is genuinely working.