In Parallel uses AI to keep shared Workspace context useful as work changes, then makes that context available to people and authorized AI tools.
AI works from shared context
Shared context is more than a list of goals or one execution plan. It combines:
Durable narrative in maintained Workspace pages
Goals and intended outcomes
Source-linked signals and observations from meetings, documents, connected conversations, and other authorized sources
The current execution plan
Topics and briefings that interpret what is moving
AI uses different parts of this context for different tasks. It does not need to flatten the whole Workspace into one permanent prompt.
How AI builds and maintains context
1. It captures sourced change
The Recorder turns meeting conversation into Notes and source-linked observations. Decisions, actions, risks, questions, and learnings retain a relationship to their source instead of becoming unsupported prose.
When Slack or Microsoft Teams is connected and conversation access is allowed, authorized conversations can contribute context and signals in the same way. Workspace observation kinds define what AI should look for, including default kinds such as decisions and customer-defined kinds such as customer need.
2. It compiles durable narrative
Page types tell AI what a maintained page is for. A Charter, Objectives page, Runbook, Retro Log, and meeting page need different structure and maintenance behavior.
In Parallel can prepare a first version and update compiled pages as relevant Workspace context changes. This reduces the burden of manually keeping a second documentation system synchronized with the work.
3. People edit and review
People can edit Workspace pages directly. About this page explains the page's purpose, perspective, source inputs, and refresh behavior.
The page header shows when the page was updated and whether its latest refresh succeeded. A Refresh failed status is a reason to check important details against their sources.
4. AI reuses the maintained context
Once context is maintained, it can support many surfaces without asking the team to explain the Workspace again:
Home and Workspace briefings
Page-aware Chat through Ask
Meeting talking points, preparation, and follow-up
Topic storylines
Execution-plan and drift analysis
Authorized connected AI tools such as Claude, ChatGPT, and Microsoft 365 Copilot over MCP
Reports, documents, presentations, and other deliverables prepared in those connected AI tools
How maintained pages and source records differ
A maintained Workspace page is a narrative and handover layer. It helps a new reader become competent about the current situation.
It is intentionally not the only source of truth. Signals and observations remain linked to the meeting, conversation, document, or other authorized source behind them. A summary can omit or compress detail, so high-impact claims should remain traceable to the underlying record.
This separation lets AI provide a readable overview without pretending that generated narrative is equivalent to the evidence behind it.
How AI uses context safely
Per-reader access: Chat and connected tools receive only context the reader is authorized to access.
Per-task selection: Ask can ground one turn in the page on screen; a briefing can use a broader Workspace view.
Freshness signals: Source timestamps and the page's update or refresh state help a reader weigh the material.
Reviewable actions: Connected AI tools can use supported In Parallel actions, but consequential writes and deletions require confirmation.
No silent authority: AI-generated wording does not make a decision, assign accountability, or prove that a claim is correct.
AI may refresh a maintained page. That is maintenance of a narrative layer, not an unreviewed business decision. Review consequential changes against the source records and responsible owner.
Improve the context AI works from
Keep each Workspace bounded by a coherent area of accountability.
Use Workspace pages for durable context, not every transient update.
Keep goals specific enough to express the intended outcome.
Review important signals, observations, and ownership changes.
Review high-value Workspace pages after major changes and check their refresh state.
Open source links before acting on a surprising or consequential claim.
The richer the shared context becomes, the less often people and AI need to reconstruct the basics. Quality comes from useful structure, visible sources, and deliberate review—not simply from adding more text.
What AI does not know automatically
AI cannot recover information that was never captured, infer context hidden by permissions, or reliably resolve a conflict that the team has not resolved. Well-maintained Workspace pages make gaps easier to see; they do not make missing evidence disappear.