Is this the same as connecting Discera to Claude or ChatGPT?

No. This is Discera's own assistant, built into the app — nothing to install. Connecting Discera to Claude Desktop, claude.ai, or ChatGPT is a separate setup over MCP. See The MCP server and Claude for that.

Can the assistant spend my quota without asking?

No. Before starting any run it calls a preview — showing the matched-calls estimate and projected quota deduction — and waits for you to say go. Only your explicit confirmation starts the run and deducts quota.

Can it answer questions about my actual calls, not just how the product works?

Yes, where enabled for your workspace. It can search across your Gong call transcripts to answer free-form questions about an account or topic ("what pricing objections came up") and count how many calls mention a term. If that feature isn't turned on for your workspace yet, it will tell you.

Does it change anything in Gong or HubSpot?

No. Like the rest of Discera, the assistant is read-only against your connected tools. The only thing it writes is a Discera analysis run or schedule — and only after you confirm. Those actions are recorded in your workspace audit log.

What is the Discera AI assistant and what can it do?

Getting started·Jul 10, 2026·4 min·By Discera Team

Meet Discera's in-app AI agent: ask questions across your calls, set up and run analyses, and schedule reruns — all from chat, with a preview before any run.

TL;DR

Discera's built-in AI assistant is an agent you talk to in plain English. Ask it questions across your calls, describe an analysis you want and it sets it up and runs it for you, or turn a good analysis into a schedule — all from one chat, and always with a preview before it spends any quota.

How it works

The assistant opens from a launcher in the app and is available on every page once you're signed in (it stays out of the way on the sign-in and onboarding screens). You describe what you want; it figures out the rest.

Under the hood it's an agent, not a scripted menu. When you ask for an analysis, it reads your company profile and prompt library, drafts a real analysis config, and fills in a live config panel on the side of the chat so you can see — and edit — exactly what it's about to run: the question, the date window, the call limit, and any filters.

It can also just answer questions. Ask how a feature works and it searches the help center and cites the article. Ask about your own calls — "what are Acme's key goals," "what objections came up on pricing" — and, where the feature is enabled for your workspace, it searches across your Gong call transcripts and cites the real calls behind its answer.

What you can ask it

·Run an analysis. "Find the deals we lost last quarter and why." It writes the prompt, sets the window, previews the matched calls, and runs it once you confirm.
·Ask about your calls. "What pricing objections came up across our calls?" or "How many calls mention Competitor X?" It searches — or exactly counts — across your mirrored call transcripts and cites the calls. (Available where enabled for your workspace.)
·Look at past work. "What reports do I have?" or "Summarize my last run." It lists and reads your existing reports.
·Ask about your Insights. "What are buyers saying lately?", "Which competitors come up most?", or "What objections are trending?" It reads your rolled-up Insights — themes, competitors, objections, and asks. (Growth and up.)
·Schedule a rerun. After a run, it can offer to make it recurring — weekly or monthly — and set up the schedule as a draft for you to confirm.
·Learn the product. "How do HubSpot filters work?" It answers from the help center with a link.

What happens when it runs an analysis

The assistant never starts a run on a single sentence. Every run goes through the same two steps:

01Preview. It calls a preview that consumes no quota and shows you the verdict — how many calls match, and how much of your monthly quota the run would use. It presents those numbers to you verbatim.
02Confirm. Only after you reply "go," "run it," or similar does it start the analysis. That's the moment quota is deducted.

Once the run starts, the assistant gives you a watch progress link to /runs/<job_id> so you can follow it live, exactly as if you'd started it from /new. If you ask it to cancel a running job, it first warns you that calls already analyzed aren't refunded, then cancels only after you confirm.

Safe by design

·Read-only against your tools. It reads calls from Gong and deal context from HubSpot; it never writes back to them.
·Confirmation before spend. Starting a run and activating a schedule both require an explicit yes from you — the same preview-then-start guard the MCP integration uses.
·Auditable. Runs and schedules the assistant creates are recorded in your workspace audit log, tagged as coming from chat, so admins can always see what it did.

In-app assistant vs. Claude and ChatGPT

This assistant lives inside Discera — there's nothing to install. If you'd rather drive Discera from an outside AI client, Discera also exposes the same kinds of actions over the Model Context Protocol so Claude Desktop, claude.ai, or ChatGPT can read your reports and run analyses on your behalf. That setup is covered separately in The MCP server and Claude.

Availability

The in-app assistant is rolling out and may not be enabled for every workspace yet. Answering questions across your own Gong calls is a further step that's turned on per workspace — if it isn't available to you, the assistant will say so and fall back to what it can do. If you don't see the assistant and want it turned on, email hello@discera.ai.