Explore Work Intelligence
See what kinds of work your team does with AI, which Linear projects the spend lands on, and where to find engineering output.
Work Intelligence connects AI spend to the work it went into. It sits under Observe in the sidebar and has three views: Projects, Workflows, and Engineering. The page title reads Work Intelligence / Workflows, and so on, so you always know which view is open.
Who can see this: Everyone can open Work Intelligence, but the sidebar lists only the views your role can read:
| View | Admin | Manager | Member |
|---|---|---|---|
| Projects | Whole organization | — | — |
| Workflows | Whole organization, with a team filter | Their own work | Their own work |
| Engineering | Whole organization, with a team filter | Their reporting line | — |
Clicking Work Intelligence opens its first view. Admins can filter by team with the picker at the top right, and the team you pick stays selected when you switch views.
If you have an old link to the Workflows or Engineering pages, it opens the matching Work Intelligence view.
Workflows
The Workflows view groups sessions into kinds of work, such as a recurring task your team does with AI. Tokenize builds these groups by analyzing sessions, so they appear once enough sessions have been analyzed. Until then the view says "No workflow report is available."
Pick a period
Use the picker at the top right to choose a report:
- A rolling window, such as Last 30 days.
- Week, Month, or Quarter. Use the arrows to step through earlier periods, or click the period to open a calendar and pick one. A period you can't pick says why: "Not over yet", or "No report: too few sessions, or older than a year".
Tokenize keeps a report for each past week, month, and quarter, so you can compare how work changed over time.
Categories
A Categories bar at the top splits the period's sessions by what they were for, with the largest categories named and the rest grouped as Other. If no session has a category yet, it says "No categorized sessions in this window."
Workflow cards
Each card is one workflow, with the people involved. Click it to see Spend in this window, Conversations, Avg. per conversation, and Share of spend, plus the workflow's Labels, Sub-workflows, and People. When some of the period's spend couldn't be priced, the share is labeled Share of known spend instead.
When several people do similar work, the details say "N people doing similar work — check for duplicate effort."
Admins see the whole organization. Everyone else sees their own work. Workflow names and descriptions describe the organization's work as a whole, but the numbers on each card only count sessions you're allowed to see.
Projects
The Projects view ties AI spend to projects in Linear. It needs a Linear workspace connected to Tokenize. Until then it shows "See what your AI spend built" with a Connect Linear button, which opens Settings → Integrations. See Connect Linear.
Who can see this: Admins.
If the view says "Reconnect Linear", the Linear connection has expired. Click Reconnect Linear. "Linear is syncing" or "Linear sync queued" means Tokenize is still reading your workspace; projects appear when the sync finishes. Once Linear is connected and synced, "No Linear projects yet." means the workspace has no projects Tokenize can see.
Projects are listed in a table grouped by their Linear status, with in-progress projects open first. Each row shows Estimated cost, Estimated tokens, People, and Sessions, and a last No project group lists issues that aren't in any project. A line at the top says when Linear was last synced. If it says "Linear data is out of date." or "Linear sync failed.", the page is showing the last good sync.
Use the date range picker to choose a range of up to 30 days. Picking a team narrows the list to projects with activity from that team; it then says "No project activity for this team in this period." when there is none.
Click a project to open it. The detail page shows the same four stats, then tabs for Pull requests, Sessions, Figma files, Notion and Drive, Tools, People, and Teams. Tabs with nothing in them are left out. The Open button at the top right opens the project in Linear.
- Pull requests are cards that show the same tokens and cost as the pull request's row on Engineering. That is the pull request's whole attributed cost, not this project's share of it, so the cards don't add up to the project's total. Hover the cost to see the split by provider and the Attribution confidence. A pull request that isn't in the Engineering cost report for the same range shows no cost; its link still opens its row.
- Sessions read like a Sessions row: the person, the model they used, and the session title, such as "Ava used Claude Opus 4.5 for Fix the login redirect". The tokens and cost on the row are the project's share of that session. Open a session to see what ties it to the project, such as the issues, pull requests, Figma files, and documents it touched.
- Figma files are cards with the file name, how many nodes the session touched, and what the session did to the file, such as Created. Figma files show no cost.
- Notion and Drive lists the Notion pages and Google Drive files the project's sessions touched, the ones touched by the most sessions first, each with a badge for what sessions did: Created, Modified, Commented, or Viewed. A document with no title says what it is, such as "Google Doc" or "Notion page". "+12 more documents" means the tab lists only the first ones.
"Report too large" means the range covers too much activity. Pick a shorter date range.
Spend on Projects is an estimate of what the sessions tied to each project cost. See How cost is calculated.
Engineering
The Engineering view connects coding spend to pull requests, with code output, languages, reasoning effort, and skills alongside. See Track engineering output.
Related
- Open any session from a workflow or project on Sessions.
- Find the changes most likely to lower spend on the Opportunities board.