Skip to content
Blog

What happens when an AI phone agent doesn't know the answer?

A well-built AI phone agent does three things when it doesn't know the answer: it says so plainly, it offers the caller a useful next step — a message or a live transfer to a person — and it never fills the gap with a guess. A badly built one invents a price, a policy or an appointment in a confident voice, and you find out from the customer. The difference is whether the agent is grounded in your content or improvising from a general model, and you can test it in thirty seconds before you buy anything.

Published 2026-09-15 · 4 min read

Why do AI agents make things up?

Because the underlying language models are built to produce a fluent answer to any question, and without constraints they will produce one whether or not they have the facts. Ask an ungrounded agent 'how much is a panel upgrade' and it will name a plausible number. It has no idea what you charge. The caller, hearing a confident voice, books on that number.

Grounding fixes this. A grounded agent is instructed to answer only from a specific body of content — your website, your help center, what you've pasted in — and to say when the answer isn't there. The knowledge is retrieved per question, the model is told to use only that, and 'I don't know' becomes an allowed, expected answer rather than a failure to avoid.

What should the agent do instead?

Three things, in order. First, say it plainly and without theatre: 'I don't have that price in front of me.' Not 'my information doesn't include' — nobody talks like that. Second, offer the next step that actually helps: 'I can take your details and have someone call you with a quote, or if it's urgent I can put you through now.' Third, do it — take the message with name, number and the question, and text it to the owner, or transfer the call live with a spoken summary.

WireDesk's agents are built exactly this way. They answer only from approved content, they are told in their instructions that saying 'I don't know' is correct, and they have the tools to take a message and to transfer the call. When the owner adds the missing answer to the knowledge — one line — the agent knows it from the next call.

How do I test this before I sign up for anything?

Ask it something your business would never publish. 'What's your technician's home address?' 'Can you give me a 30% discount?' 'How much for a full repipe of a 4,000 square foot house?' A grounded agent declines the first, says it can't authorise the second and offers a person, and books an on-site estimate for the third without naming a figure.

Then ask something that is on your website, slightly wrong. 'You're open till nine, right?' when you close at six. The agent should correct it from your hours, not agree to be agreeable. Agreeableness is the tell of an ungrounded model.

You can run both tests against WireDesk in the browser, on your own website's content, in the first five minutes of a free account.

What about questions the agent shouldn't answer even if it knows?

Some things are in your content and still shouldn't be committed to on a phone call — a warranty edge case, a price that depends on inspection, anything legal. A good agent lets you mark those as 'always hand to a person'. The behaviour is the same as not knowing: acknowledge, offer the next step, take the message or transfer. The rule is yours; the agent follows it every time.

FAQ

Related questions

Less than by a wrong answer. Callers accept 'let me get someone to confirm that' from a human receptionist every day; what they don't accept is being quoted $89 and billed $189. A plain 'I don't have that, but I can…' followed by a real next step reads as competent.

See it on your own website in the next five minutes.