Send to AI
Send to AI takes an observation from your meetings — an action item, a decision, a risk — and hands it to your AI assistant with a pre-rendered prompt that already knows the context: what was said, who's responsible, what the next step is. One click in In Parallel becomes a working session in your AI client of choice, with no copy-paste in between.
What it is
After a meeting, In Parallel turns the discussion into a changelog of structured observations. Some of those — the actionable ones — are work you'll do outside the meeting: draft a doc, refine a spec, sketch a plan. Send to AI is the bridge to that next step.
For each actionable observation, In Parallel pre-generates a prompt that captures the relevant context from the meeting — the goal, the decision, the constraints, the next step the owner agreed to. When you click Send to AI, that prompt opens in your chosen AI runtime, ready to run.
You stay in control: the prompt is visible before you send it, the AI client is your choice, and the work happens in the tool you already use.
How it works
Step | What happens |
1. Observation gets a work-item | When a changelog observation is actionable (action item, decision, risk, opportunity), In Parallel pre-renders a prompt tailored to that kind of work. The prompt embeds the meeting context, the observation's content, and the implicit next step. |
2. You see the Send to AI affordance | On the action-item card (and in Findings detail), a sparkle icon appears next to the title. Hovering it reveals a tooltip with the receiving AI client's name. |
3. It goes to your AI client | Your Preferred AI client for Send to AI setting (in Personal settings) controls this. On Always ask (the default) a chooser lets you pick a client per send; set a specific client and the click skips the chooser and opens that one directly. |
4. The prompt opens in your AI client | In Parallel deep-links to the receiving runtime with the prompt pre-filled. You review, edit if you want, and start the conversation. |
Choosing your AI client
Send to AI can open the prompt in any of these clients: ChatGPT, Claude (Web), Claude Desktop, Cursor, Perplexity, and VS Code + Copilot.
Set which one to use under Personal settings → Preferred AI client for Send to AI. Leave it on Always ask (the default) to choose per send, or pick a specific client to skip the chooser.
To let an AI client also read your In Parallel data — not just receive a one-off prompt — set up the MCP connection first; see Connect your AI tool.
What's in the prompt
The pre-rendered prompt isn't just a description of the action item. It includes:
The meeting context — what was being discussed when this came up
The observation itself — the action, decision, or risk as captured
The owner's commitment — what they said they'd do, in their own words where possible
The implicit next step — what kind of work the receiving AI should help with (draft, refine, plan, summarise)
You can review and edit the prompt before it runs in your AI client. The goal is that the default prompt is good enough to use as-is for most observations.
When to use Send to AI
Send to AI fits the in-between space: too much for a one-line task tool, not yet shaped enough for a clean handoff to a teammate. Good fits include:
Drafting a follow-up document from a meeting decision
Refining a spec sketched during the discussion
Working through a risk's possible mitigations
Turning a vague action item ("look into X") into a concrete next step
It's not the right tool for everyday task tracking (the action item itself lives in In Parallel) or for work that requires a teammate's input — hand that off to the teammate directly.
Privacy and scope
The prompt is sent to the AI client you choose. Each one has its own privacy and data-retention policy; In Parallel doesn't store the conversation that happens after the handoff.
The prompt includes only the observation and the meeting context already visible to you in In Parallel. It does not include other observations, organisation-wide data, or other members' personal data.
Related
Connect your AI tool — set up the MCP connection to Claude, ChatGPT, or Cursor before using Send to AI
Findings — where you act on observations across Workspaces
Changelogs — where Send to AI most commonly appears, in the action items list
Connect your calendar and notification tools — companion article for setting up Integrations