Skip to main content
The Engineering Overview tab is for engineering managers. It shows one team at a time, or all teams together, in enough detail to act on: who is spending, what it bought, and which PRs or people need a look. Who sees it: executives, and engineering managers (a role with the ai-roi:manager permission). Before teams are set up, the approver of a Wallets group also sees that group here. See Who sees what.
The Engineering Overview needs teams. If none are set up, the tab shows a single “set up teams” section. See Teams and people.

Pick a team

Choose a team, or All teams, at the top of the tab. Executives and engineering managers open on All teams; others open on their own team. All teams covers everyone who is on a named team. People on no team (“Unassigned”), PRs that matched no one, and spend from automated agents are left out of this tab. The Executive Summary still counts them.

What’s on the tab

Engineering overview

The team’s three scores (Velocity, Productivity, Efficiency), compared with the organization. A team with fewer than 5 engineers shows no scores to people outside it. Below the scores, Where the team stands shows five tiles and a spend chart: The chart shows daily AI coding spend for ranges up to about three months. Switch it between Tool, Model, Person and, on All teams, Team. The largest groups get their own color; the rest fold into “Other”.

Where the PR cost goes

The AI cost of the team’s merged PRs, grouped into workstreams (a Linear project or Jira epic when the PR names a ticket, else the repository). Toggle between Map, a treemap sized by cost, and List, a ranked list. Every PR behind it is in Engineering & PRs.

Where the spend goes

A flow from a source to a work type (Build, Fix, Understand, Docs, Ship, Other) to an outcome:
  • Merged: spend tied to a merged PR.
  • No merged PR: exploration, research, unfinished work, or spend not linked to a PR.
Switch the source between Repository and Team (it opens on Repository for one team, Team for all teams). All the team’s AI spend is in the flow, so it adds up to the team’s spend. Use the Flow / Table toggle to see the same numbers as a table. Under the chart, a bar shows where each work type came from: the PR title or branch, a linked ticket, the AI ROI Agent, or “Not categorized”. The larger the agent’s share, the more the split rests on labels chosen on developers’ machines. See Work types.

People

One row per person on the team, flagged first:

Flags

Flags are prompts for a conversation, never a verdict. A person can be flagged for:

Engineering and PRs

Every PR merged in the range, framed as return on AI:
  • Return on AI-assisted work shows the AI cost of the PRs shown, the share that is AI-assisted and the cost per output point, and compares AI-assisted PRs with the rest on median AI cost per PR, output per PR, median time to merge and estimated rework.
  • By type of work shows each work type’s AI cost, PR count against the previous period, cost per PR and rework. Click a type to filter the list.
  • Pull requests lists every PR. Search, show All, AI-assisted or Other PRs, and sort by Costliest, Newest, Most output or Slowest. The figures above follow the filters.
A cost shown with ~ is an estimate (no agent data for that PR). Arriving means the PR merged after the last day with final spend, so its cost isn’t counted yet. — means the author isn’t matched to anyone. See How the numbers work. This section follows the page’s scope: an executive’s focused team, a manager’s chosen team, or a developer’s own PRs.

Teams setup

A summary of who is on which team, with a Manage teams button. Shown to people who can manage teams. See Teams and people.

Use cases

Three ranked lists of where AI work goes:
  • Top product domains: the Linear projects and Jira epics of merged PRs.
  • Top technical areas: frontend, backend, infrastructure, testing and so on, from changed files and agent edits. Tests count toward the change they test.
  • Top repositories: the repositories of merged PRs.