A long thread arrives. Six messages, three people, a question that needs a careful reply. You'd love to ask AI to draft it. But you know how the next ten minutes go: copy the thread out, paste it into ChatGPT, explain who Marc is, summarize the project, ask for the draft, paste it back into Gmail, edit, send.
By the time you've briefed the AI, you could have written the reply yourself. So usually you do.
Corivo's agent fixes that.
The cost of the old way
The reason replying with AI feels heavier than it should isn't the writing. It's everything around the writing:
- The thread tax. You copy the visible messages out of Gmail, hoping you got all the context.
- The people tax. You explain who Marc is — his role, his stake, the fact that he was the one who pushed back two weeks ago.
- The project tax. You describe what the project is and why this thread matters to it.
- The voice tax. You ask the AI to match your tone, even though it's never seen what you write.
Four taxes, before any actual drafting happens.
What Corivo's agent does
You give Corivo one command: "Draft a reply to Marc's email."
The agent already has the picture:
- Reads the thread from Gmail and from your active screen — the full back-and-forth, not just what's visible.
- Knows who Marc is — from your context graph. His relationship to the project, his communication patterns, what he's pushed back on before.
- Knows the project — from connected tools (Notion, Linear, Slack) and from the work you've actually been doing on it.
- Knows your voice — from the emails and docs you've written before. Length, formality, hedging style.
- Drafts a reply in your voice that lands the actual answer.
- You review and send. Sharpen one sentence, soften another, ship it.
The reply is shaped by the real situation, not by your one-paragraph summary of it.
Status today
What works in the current beta:
- Screen reading. If the thread is open in Gmail (or another mail client), Corivo's agent can read the visible structured text via accessibility APIs.
- Local memory. Patterns about the people you communicate with build up as Corivo watches your work.
What the connected tools add:
- Gmail — pull the full thread (including parts not on screen), all participants, send/reply metadata, and related threads. Auto-recognize that Marc-the-stakeholder is the same Marc who showed up in the project.
- Notion / Google Docs — pull the project doc that frames why this thread matters.
- Slack — pull the thread where you and Marc discussed this asynchronously last week.
- Linear — pull the issue Marc is asking about.
You can start from the open thread, or let Gmail supply the full conversation directly. Either way, the draft is grounded in the actual thread instead of a paste-in summary.
What you'll feel
The first time you don't have to brief — when you just say "draft a reply" and the draft is grounded in the actual thread, the actual person, and the actual project — the time-to-reply collapses. More importantly, the reply lands in your voice. Not the chatbot's idea of your voice.
Reply quality goes up because the AI has the right inputs, not because the AI got smarter.