What you can do with In Parallel's AI
Once you connect an AI assistant to In Parallel over MCP, it can read your execution data and act on it in plain conversation — so the question becomes what to ask it for.
What it is
In Parallel has two kinds of AI, and they do different jobs:
The AI built into In Parallel keeps your execution plan current — it turns meetings into structured reports, surfaces drift, and proposes priority changes for you to confirm. You don't drive it directly; it runs as part of the meeting → review → confirm loop. That layer is covered in AI in In Parallel.
Your own AI assistant, connected over MCP — Claude, ChatGPT, or Cursor — reads your In Parallel data on demand and acts on it when you ask. This is the one you talk to.
This article is about the second one: the practical things a connected assistant can do for you. If you haven't connected a tool yet, start with Connect your AI tool to In Parallel.
What you can ask for
You don't need to know tool names — describe what you want and your assistant picks the right tool. These are the patterns that work well today:
You want to… | Ask something like… | What it draws on |
Prep for a meeting | "Catch me up on the Back-to-School Workspace before my 2pm — recent decisions and open actions" | Recent meetings, decisions, action items |
Catch up after time away | "What changed in my Workspaces while I was out last week?" | The plan diff and changelog |
Run a weekly review | "Run a weekly review of the Acme Workspace" | The |
Triage your tasks | "What's overdue across my Workspaces, and help me reschedule it" | Open action items, owners, due dates |
Turn a discussion into goals | "Read last Thursday's planning meeting and propose goals from it" | A meeting transcript → goal candidates |
Check execution health | "Run a drift report on the Migration Workspace — what's slipping?" | The execution-health diagnostic |
Record a decision or question | "Log that we decided to ship the beta on the 15th in the Launch Workspace" | Writes a decision back into In Parallel |
Keep your delivery board in sync | "Turn the new decisions from In Parallel into Jira epics" | In Parallel + a second connector (see the Jira how-to) |
Read requests — looking things up, summarising, reviewing — run freely. Requests that create or change something (logging a decision, creating a goal, rescheduling a task) go through your client's approval step first, so nothing is written without your say-so. See What the AI can and can't see for how those limits work.
Two patterns worth knowing
Pre-wired routines. Beyond one-off questions, In Parallel ships a couple of ready-made multi-step routines your client can run by name — look for them in your client's prompt or slash-command menu:
weekly_workspace_review— a full weekly review of a Workspace: health check, then open decisions and questions, then recent meetings.triage_action_items— lists a Workspace's open work (optionally for one owner) and helps you assign and schedule it.
Working across two tools at once. Connect In Parallel and a delivery tool like Jira to the same assistant, and it can reconcile the two — turning In Parallel decisions and goals into Jira epics, and reflecting delivered Jira work back into your plan. The execution layer stays in In Parallel; the delivery work stays in Jira; the assistant bridges them. See Use the In Parallel MCP with Jira in Claude for a worked example.
A faster route from a single Finding
If you're already looking at one observation — an action item, decision, or risk — in In Parallel, you don't have to switch to your assistant and re-explain it. Send to AI hands that Finding to your AI tool with a pre-rendered prompt that already carries the context, so the assistant starts from the work rather than from a blank chat.
Related
Connect your AI tool to In Parallel — set up Claude, ChatGPT, or Cursor over MCP.
What the AI can and can't see — the data boundaries and permissions behind these actions.
In Parallel MCP tools reference — the full set of tools a connected assistant can use.
Use the In Parallel MCP with Jira in Claude — a worked two-tool example.
Send to AI — hand a single Finding to your assistant with its context attached.
AI in In Parallel — the AI built into the product, and what it doesn't do.