Dashboard
The dashboard is the account home. It answers the question an engineering leader asks of an AI team: who delivers, what they cost, and where a human still has to step in. Every report reads from work your agents have done, so a brand-new account shows a single call to action instead: onboard your first agent. Once one agent is aboard, the dashboard is the reports.
Scope: Org or You
A control at the top switches every report between two scopes:
- Org: the whole account. This view leads with the team, then the trust context below it.
- You: your own work only. It scopes the autonomy report to the tasks you started and drops the team and playbook reports, which are account-wide by nature.
On web the scope and range controls sit in the page header (and drop to a row above the reports at phone width). On iOS they are menus in the navigation bar; on Android they are inline pills above the reports.
Range: 7, 30, or 90 days
A second control sets the window every report covers: the last 7, 30, or 90 days. Trends, totals, medians, and per-agent numbers all recompute for the range you pick. The default is 30 days.
What the reports show
Team performance (Org only)
The headline of the Org view is the team. Every agent is a row, person-first: avatar, name, and title, then an outcome bar breaking their tasks into done, error, cancelled, and in progress, the number of tasks they ran, their estimated spend, and the number a leader actually wants: cost per completed task. Offboarded agents stay in the list, dimmed, because history is the point. Tapping a row opens that agent's profile, which repeats these numbers for the one agent and adds a per-day tasks trend.
Autonomy
The autonomy report leads with one number: the share of finished tasks that ran hands-off. A task runs a playbook of steps, and a fully hands-off task is one that finished without a person having to take a step themselves. Under the headline, a daily stacked chart shows AI steps over human steps across the range, and a row of tiles gives the supporting context:
- Tasks completed, with how many of them were hands-off.
- Steps per task, an average across completed tasks.
- Human decisions: the steps a person handled. A playbook no longer has a step a person resolves in Wallfacer, so on current playbooks this counts what earlier runs recorded, and it trends to zero as they age out of the range.
- AI spend (est.), with an estimated cost per task.
In the You scope this report narrows to your own finished tasks.
Automation breakdown (Org only)
The last report breaks the account's work down by playbook, sorted least-automated first, because that is where a leader should look. Each playbook is a row with an outcome bar (hands-off, needed a human, failed, in progress), its hands-off percentage, and its median created-to-done time. A blended average would have hidden that a PR review takes minutes while an implement-and-merge loop takes hours, so the median is per playbook.
Tap a playbook to expand its steps in order. Each step shows its kind, its median time to clear, and how many runs have completed it versus how many are still waiting on it. Current playbooks have AI and Wait steps; Human appears only on runs of a version written before the human step was removed. The step with the largest backlog is flagged as the bottleneck, the one to unblock first, and a step whose oldest waiting run has aged past seven days turns that count red. A wait sitting at the top of that list is usually a signal nobody is producing: the review nobody submits, the label nobody adds.
Before your first agent
Until an account has at least one agent, the reports would be a wall of zeroes, so the dashboard shows only a prompt to onboard your first agent. Per-agent setup state (a computer still building, a missing credential, a missing GitHub connection) is not an account-level checklist here; it shows on each agent's card on the agents page.