Key Terms

Key Terms

Your agent listens to your guests through speech recognition, and speech recognition is very good at ordinary English and much worse at your words. Brand names, product names, feature names and industry jargon are exactly the words it has never seen — so it guesses, and it guesses wrong.

Key Terms is where you tell it what those words are.

Why it matters

When your agent mishears a word, two things go wrong, and the second one is worse.

The obvious problem is that it sounds like it doesn’t know your product. A guest asks about your platform, the agent hears something else entirely, and asks the guest what they meant — about your own product name. It’s a bad moment in an otherwise good conversation.

The subtler problem is that a misheard word can change the answer. Your agent decides what it’s confident enough to answer directly and what it should defer on, and that decision is based on what it heard. If an industry term comes through as a different word, the same question can get answered one time and deferred the next. Key Terms removes that inconsistency.

It’s already on

You don’t need to switch anything on, and there’s nothing to configure before it works. Every workspace starts with a set of key terms already in place, built from your workspace and agent names.

This page is where you improve that list — not where you enable it.

Open Settings → Personal Meeting Link → Key terms. The terms you’ve added appear at the top; below them, under Already recognised, no action needed, are the ones you get automatically.

What to add

Add words your guests actually say out loud, and that aren’t ordinary English:

• Your product and feature names — especially invented words, compound words, or anything with unusual capitalisation
• Your company name, if it isn’t a common word
• Industry jargon your guests use — the vocabulary of your market, not just your brand
• Names of things you sell or integrate with, if they come up in conversation

Don’t add ordinary words. “Pricing”, “demo” and “support” are already recognised perfectly, and adding them makes things worse — see below.

Why the list is capped

You can have up to 25 key terms, and this is a real limit, not an arbitrary one.

Key terms work by telling the recogniser “listen extra hard for these”. Give it a handful of distinctive words and it gets them right. Give it a long glossary and it starts forcing matches — hearing your key terms in places where the guest said something else entirely. Accuracy goes down, not up.

So the list is a curation exercise. Twenty well-chosen words beat a hundred, every time. If you hit the cap, remove something that rarely comes up before adding something new.

Teaching it what it hears instead

Sometimes you already know exactly what the agent mishears. If your product is called “BuzzTrail” and the transcript keeps saying “bus” or “post trial”, you can say so directly.

Click a term to expand it, then list what it’s being heard as, separated by commas:

bus, bus trail, post trial

Your agent will treat any of those as the real term and answer the question, instead of stopping to ask what the guest meant.

This is different from the term itself. The term is the correct spelling and is what the recogniser listens for. The mishearings are the wrong versions, and they’re used to help your agent interpret what it heard — they’re never fed back into the recogniser, because teaching it to listen for “bus” would cause the exact problem you’re fixing.

Who can edit

Any host can edit the key terms for their own workspace. Key Terms are workspace-wide: they apply to every meeting and every agent in the workspace, not to a single link.

Related

• Branding — your company name, which also seeds your key terms
• Behavior — choosing your agent’s face, voice, and framework