These are the least glamorous AI features in Gleap and among the most useful. They don't answer anybody — they make the queue readable, so your team spends less time working out what each conversation is about.
You'll find most of them under Settings → AI → Assist. Automatic tagging is set up separately, as a workflow action — see below.
Conversations arrive without one. Left alone, a queue of twenty reads as twenty first lines, several of which are "Hi".
Generated titles turn that into something you can scan and search. This is the one to switch on first — it changes how the inbox feels more than anything else on the page.
A summary of the conversation so far, so someone picking up a long thread doesn't have to read twenty messages to find out where it stands.
It earns its place on threads that get handed over, or that ran through several people. On a two-message exchange, it adds nothing the messages don't already say.
A proposed priority based on what the customer wrote. Useful for surfacing the genuinely urgent thing that arrived politely worded.
Incoming messages can be translated automatically into your browser language, which lets a team answer languages nobody on it speaks. When you reply, use the "Translate before sending" option in the composer to send your answer back in the customer's language.
You give a list of tags and conversations get tagged as they arrive. It's set up as an AI-based Tagging action inside a workflow, not in the Assist panel — add the action, define the tags you want the AI to choose from, and save. This is what makes reporting by topic possible at all — manual tagging works for a week and then quietly stops.
Keep the list short and unambiguous. Ten well-separated tags get applied consistently; forty overlapping ones produce noise, because "billing" and "payment" will be assigned more or less at random.
Each of these runs on incoming conversations, so they add up on a busy inbox. If you're watching AI spend, they're a reasonable place to be selective — titles and tagging give the most back for what they use, and summaries are worth having only where threads genuinely get long.
