Task Board

The Task Board organizes work visually (To do → Doing → Done). The differentiator: people and agents share the same board. AI tasks run on their own at the time you set; human tasks come with acceptance, deadlines, action checklists and evidence-backed completion — and everything is recorded in an auditable history.

​

The lanes

​

Every task type has its own look

Glance at the board and know who handles what:

• 👤 Person — blue border and an avatar with the assignee’s initials.
• 🤖 AI — purple border, executed by an agent or team.
• 🌐 API — orange border, calls an external system (webhook).

​

Creating a task

​

Human tasks — acceptance, delivery and evidence

When you assign a task to a person, it doesn’t just land on their pile:

1. Acceptance. The person gets notified (bell + email) and decides: accept (the task moves to To do) or decline — declining requires a reason, which goes back to whoever asked. Only the assignee can decide; not even the company owner can accept on someone else’s behalf.
2. Actions. Inside the task lives an action checklist — the steps of the work. Every checked item records who did it and when.
3. Completion with proof. Completing requires telling how it was done and having at least one action recorded (you can type them right there). Attach evidence (spreadsheet, screenshot, contract). The result becomes a complete block — note + steps + evidence + delivery time — in the same place where AI writes its answers.

​

Sub-tasks and Roadmap

Break big work into sub-tasks (up to 4 levels) — each can go to a person or an agent, mixed in the same tree. The parent’s progress adds up on its own as children finish. On the Roadmap tab, create initiatives (the layer above: “Launch the new portal by September”) and tie tasks to them. The initiative shows a real progress bar computed from actual work.

​

Visibility — who sees what

Owners and admins always see everything (governance). The filter is enforced server-side: non-participants can’t see the task even with a direct link.

​

Filters and saved Views

• Quick filters: Mine × All · 👤 People / 🤖 AI / 🌐 API · ⏳ Expiring / 🔴 Overdue — resolved server-side, across all lanes.
• Views: save a named filter (“My criticals: high priority + mine + overdue”) and apply it with one click.

​

Dashboard — your board’s metrics

The Dashboard tab works like the Pipes dashboard: pick what to measure (lead time, cycle time, weekly throughput, load per person, awaiting acceptance, time to accept, declined, overdue with targets…), how to break it down (person, project, initiative, priority, week…) and the chart type. Drag, resize and save the layout. The dashboard obeys the active filter: filter “Mine + People” and the charts compute over exactly that slice — with a badge showing the applied filter.

​

History — everything is recorded

Every task carries its own auditable history: who created it, who assigned it, who accepted or declined (and why), every move between lanes (from → to), every edit with the changed fields, every checked action, AI executions (success/failure) and the completion. The record is immutable — editing a task never rewrites its past.

​

Your Personal Assistant knows your tasks

• In the workspace, the “My work” card highlights what awaits your acceptance — accept/decline right there, and clicking any task opens the full modal (actions, completion, sub-tasks, history) without leaving the screen.
• By conversation, the assistant lists, details, accepts, declines, completes and checks actions on your tasks — and can retrieve any task by name, even completed or archived, always with a direct link.
• The Cortex sees the whole board within each person’s clearance: overdue, awaiting acceptance, declines with reasons, load per person and recent completions with the how-it-was-done note.

​

Recurrence — run on its own

When marking a task as recurring, pick the interval: Also set the time and the timezone (from the company Settings). On every completion, the next occurrence is created automatically.

​

Assigning to an agent

1. On the card, under Execute with, pick Agent and choose one.
2. Make sure the agent has access to the data/tools the task requires.
3. When the task kicks in, the agent executes and records the result on the card.

​

Task, process or team?

​

Troubleshooting

​

Next steps