AI Checker

Every answer checked before it's sent.

AI Checker compares each answer with your knowledge, catalogue and live tool results. Wrong facts are corrected, and anything it can't back up is swapped for an honest reply, before your customer ever sees it.

Targeted corrections

Wrong price in the draft? Fixed before sending.

The checker scores how much of a draft reply isn't supported by your knowledge. If a fact is wrong, it swaps just that fact, so the customer gets the right answer with no awkward re-ask.

  • Skips the small talk

    Short greetings and "terima kasih" replies aren't checked, so tokens go to real answers.

  • Patches only the wrong fact

    A wrong price or delivery term is replaced in place. The rest of the reply stays as written.

  • Checks against everything it knows

    Your knowledge documents, product catalogue, Google Sheets or API results, what the customer told you and the ad they tapped.

Honest fallback

Can't back it up? It won't send a guess.

When a draft makes a claim nothing in your knowledge supports, and there's no single fact to swap, the checker holds it back. The customer gets a short, honest reply instead.

Checking uses a few extra tokens per message, in exchange for more reliable replies.

  • Guesses are held back

    A made-up price, promise or policy never reaches the customer.

  • Honest, in your tone

    The fallback reply follows your chatbot's personality and the customer's language.

  • Failed lookups say so

    If a Google Sheets or API call failed, the customer is asked to try again instead of hearing a dead end.

Want a person to step in when the AI isn't sure? That's AI Feedback. More Info

Accuracy level

Balanced for most shops. Very Strict for clinics.

Slide between Moderate, Balanced and Very Strict. Balanced catches real errors and ignores noise. Very Strict rejects anything not fully backed by your knowledge, which is what healthcare, legal and finance businesses want.

  • Very Strict

    Rejects anything not fully grounded in your knowledge. The pick for healthcare, legal and finance.

  • Balanced

    Catches real errors and ignores noise. Recommended for most businesses.

  • Moderate

    Allows moderate inaccuracies and only steps in on clearer mistakes.

Off is also an option: replies go out unchecked.

Never let a wrong price reach a customer

Turn on AI Checker, pick Balanced, and let it double-check your chatbot's answers before they go out.

AI Checker is included in the Business plan and the free trial.

Chat with us