Productivity prompt

The Productivity prompt card is where you tell the AI what counts as productive for your organisation. Whatever you put here replaces the default 1–5 rubric for everyone in scope.

When to customise it

The default rubric is generic — "deep work = 5, entertainment = 1". It works fine for a mixed team, but it tends to misclassify specialised work. Examples:

• A software engineer staring at a stack trace looks idle to a generic rubric, but is doing the most productive thing they can do.
• A call-centre agent on the phone with an inactive screen looks idle, but is the highest-productivity state of their job.
• A designer with Figma + Pinterest open looks distracted, but Pinterest is research for them.

If your team plays one of these roles, customise the prompt.

How to write a good prompt

A useful prompt has three sections:

1. EMPLOYEE CONTEXT & ROLE — one paragraph that says who this person is and what they do.
2. TOOLSTACK — the apps / sites / files this role lives in.
3. SCORING RUBRIC — explicit rules for what each of the five scores means in this role.

The clearer you are, the better the AI scores. The default rubric is intentionally vague so it works for everyone; your customised rubric should be specific to your job functions.

The built-in templates

The Use template dropdown gives you three starting points:

Minimal (default)

The same generic 1–5 rubric the tracker uses out of the box. Useful when you want to write your own prompt and the default isn't tilted toward any role.

Software engineer

A rubric tuned for full-stack engineers / tech leads. Treats:

• IDE coding, terminal debugging, AWS console, GitHub PRs as 5.
• Jira tickets, Slack architecture discussions, docs writing as 4.
• HackerNews, file organisation, personal inbox as 3.
• X/Twitter, crypto / stock checks, casual browsing as 2.
• YouTube entertainment, games, personal shopping as 1.

Includes a note that "staring at an error log is highly productive, even if it looks like the user is stuck" — which is the single biggest miss the generic rubric makes for engineers.

Call centre / Virtual assistant

A rubric tuned for customer-support roles. Treats:

• Live softphone / CRM updates / ticket processing as 5.
• Knowledge-base research, internal escalation, ticket queue management as 4.
• Standby in the queue, organising workspace as 3.
• Casual browsing during downtime as 2.
• Extended off-task time (entertainment, shopping, games) as 1.

Org vs team prompts

• At org scope, leaving the prompt blank falls back to the desktop's built-in productiveScore rubric.
• At team scope, leaving it blank means the team inherits the org-level prompt. To override, type a different prompt in.

A Reset to desktop default / Use org default button next to the textarea lets you go back to the default without manually clearing the field.

Picking a template overwrites your text

If there is already text in the prompt field, picking a template asks you to confirm before replacing it. Templates are designed to be complete drop-in rubrics, so the editor doesn't try to merge them with whatever you'd already written.

Tips

• Iterate on real data. Pick a template, run for a day, open a timeline that scored "wrong", then adjust the rubric.
• Be specific about tools. Listing the actual apps your team uses ("VS Code, Linear, Notion") consistently produces better scores than vague rules ("development tools").
• Don't punish thinking time. Most missed scores come from the AI undervaluing time spent reading / thinking. Add an explicit line in your rubric about it.