Picrowd AI

Picrowd AI summarizes, answers questions, and reasons over the project context you can access. When allowed, it proposes changes, runs on a schedule, uses tools, and remembers.

Where AI appears

  • As its own AI workspace inside a project, with sessions, the action center, memory, schedules, and tools
  • As the floating AI pill on every workspace; press the backquote key (`) to open or close it
  • Inside note or Board conversations when mentioned
  • As the AI tools bar at the bottom of every workspace

What AI is useful for

The workspace offers starter prompts such as Project health check, Spot blockers, Summarize recent activity, Find missing info, Prioritize next steps, and Review people and ownership.

  • Summarizing a project or record
  • Finding stale follow-up or unclear next steps
  • Drafting a clean update from messy notes
  • Comparing linked records
  • Explaining recent changes in a workspace
  • Reading a file you attach: images, PDFs, Word documents, and text
  • Generating images, video, and audio through connected media tools

Talking to it

The composer is the same one used for notes. Type ^ to reference a project, ^^ to reference a record, @ to mention a person, # for a hashtag, and / to insert an AI tool or a command such as summarize, find, draft, compare, or memory. While the AI works you see its plan as a checklist and can pause, answer its questions, or ask for a change.

Access matters

AI only uses context the current user is allowed to see. Hidden or inaccessible work never becomes AI context, and per-member hidden columns stay hidden.

Project admins turn AI read access and AI write proposals on or off per project, and can require a confirmation before any execution. Write mode is off by default, and templates do not silently enable it. AI privacy settings and consent live in Settings → AI.

AI write proposals

  • Picrowd AI proposes draft actions first instead of silently changing data
  • You can edit a draft before approval, and editing resets approvals
  • Actions and batches can be approved, rejected, executed, and undone when the action supports undo
  • Destructive actions require two confirmations
  • Tool use is approved per task, or you can trust specific tools, or all tools, in Settings → AI

Automations and scheduling

You can hand the AI a standing instruction and a schedule so it runs on its own: a morning digest, a follow-up sweep, a weekly summary. Automations keep the instruction, run within your permissions, and still propose changes rather than writing silently. See AI automations & scheduling for the full picture.

Tools it can use

While it works, the AI can reach for tools: reading a web page is built in and free, and with your own accounts it can search the web, use mail, documents, calendars, task apps, media generators, and the Apple apps on your Mac. See AI tools & connectors for how to set these up.

What it remembers

The AI keeps durable facts in a private memory that lives in your AI Vault, so it does not need everything re-explained each time. Memory is auditable and editable, and it never becomes a way to see work you could not otherwise access. See AI memory and the AI Vault.

Staying informed

When the AI has finished a run or an automation has something for you, the AI pill glows and shows a nudge card so results are easy to find. Nudges are capped so they stay useful rather than noisy, and the bell can send them as device notifications.

Good prompts

  • Summarize this record and list the next actions
  • What changed in this project recently?
  • Which rows look blocked or stale?
  • Draft a short client update from these notes
  • Compare these records and point out missing information