Help CentreAI signals

AI signals

Overview

TeamPlot detects AI-related signals from real activity: patterns in messaging, code metadata, or both that suggest someone may be adopting AI tooling, shipping faster, or changing how they commit. On the TeamPlot homepage, we group the headline examples under From AI tooling—Exploring AI Tooling, Output Above Usual Pace, and Velocity Step-Up. Those map to the same detection described below (AI channel activity, high code output, and AI velocity).

Signals are heuristics: they highlight patterns worth a conversation, not proof that a specific tool was used. Like the rest of TeamPlot’s signal set, they are prompts to check in, not a verdict—detected from activity data (counts, timing, titles, and stats), not gut feeling.

For code-backed signals we use GitHub, GitLab, or Azure DevOps metadata only—for example commit and PR titles, merge rates, and line additions. We do not read source code, diffs, or file contents. For chat, we use channel activity and timing (including which channel names someone is active in), not message content.

Admins can tune attribution patterns and the lines-per-commit threshold under Settings → Organization so detection matches how your team writes commit messages and ships code.

TeamPlot also raises AI Attribution and AI Commit Size signals; they are part of the same AI-tooling family in the product even though the homepage highlights the three patterns above.

Exploring AI Tooling (AI Channels)

What it is: The person started posting in Slack or Microsoft Teams channels they had not used in the prior baseline window, and those channel names look AI-related.

How we talk about it on the homepage: Active in AI-related channels for the first time — worth a check-in on what they are finding useful.

How it is calculated:

  • We compare unique channel names where the person was active in the last 7 days with channels they used in the 8th through 21st day before today (the baseline window).
  • Among channels that are new in the recent window, we keep only those whose names contain one of these substrings (case-insensitive): ai, claude, copilot, chatgpt, gpt, llm, cursor, openai, anthropic, gemini, codex.
  • If at least one such channel appears, we raise a signal. Severity is low.

What to use it for: A lightweight nudge to ask what they are exploring (new Copilot channel, internal AI guild, support thread). Good for onboarding, community participation, or making sure people know where to get help.

Output Above Usual Pace (AI Output)

What it is: The person’s average daily lines added in version control jumped compared with their own recent baseline.

How we talk about it on the homepage: Code additions significantly above baseline — could reflect AI-assisted development. Worth a conversation about approach.

How it is calculated:

  • Requires a connected code identity (GitHub, GitLab, or Azure DevOps).
  • We compare the mean of daily “additions” over the last 7 days to the mean over the 14 days immediately before that (days 8–21 ago).
  • We only evaluate if the baseline average is at least 10 lines per day, so new or quiet contributors do not trigger noise.
  • If recent output is at least 60% higher than baseline, we raise a signal.
  • Severity is medium if the increase is 150% or more; otherwise low.

What to use it for: Opening a discussion about review load, how larger changes are broken up, pairing, and whether the pace is sustainable—not to accuse anyone of using AI. The jump may also reflect a big feature, a refactor week, or fewer meetings.

Velocity Step-Up (AI Velocity)

What it is: Both merge velocity (PRs merged per day, on average) and code output (lines added per day, on average) increased together versus the person’s baseline.

How we talk about it on the homepage: Both PR merge rate and code output up significantly — could be AI tooling, fewer meetings, or a great stretch of focus.

How it is calculated:

  • Requires a connected code identity.
  • We compare the last 7 days to the same 14-day baseline window as above (days 8–21 ago) for:
    • average daily PRs merged, and
    • average daily additions.
  • We require a positive baseline PR rate and baseline additions of at least 10 lines per day.
  • The signal fires only if PR throughput is up at least 30% and output is up at least 40% versus baseline.
  • Severity is medium if PRs are up at least 75% and output is up at least 100%; otherwise low.

What to use it for: Understanding what changed this week (new tooling, focused sprint, reduced context switching). Useful for celebrating momentum or checking that quality and review still match the faster pace.

AI Attribution

What it is: Recent commit or PR titles contain phrases your organization treats as AI attribution markers (for example Copilot or “generated by” lines).

How it is calculated:

  • Requires a connected code identity.
  • We scan commits, PRs opened, and PRs merged from the last 14 days and match the title text against your org’s attribution patterns (substring match, case-insensitive).
  • Default patterns include common markers such as co-authored-by: claude, generated by copilot, generated by cursor, and similar—you can customize or extend the list in organization settings.
  • Severity is low for one or two matching events and medium for three or more.

What to use it for: Natural conversation starters about which tools people use, commit message conventions, and how the team wants attribution to appear in history and reviews.

AI Commit Size

What it is: The person’s average lines added per commit this week is high compared with both a configurable threshold and their own prior pattern.

How it is calculated:

  • Requires a connected code identity.
  • Over the last 7 days, we compute total additions divided by total commits (minimum 3 commits in that window).
  • We compare that ratio to an organization threshold (default 200 lines per commit; adjustable in settings).
  • We also look at the same ratio for the baseline window (days 8–21 ago). If the person already averaged near or above the threshold in the baseline (specifically 80% of the threshold or higher), we do not signal—they may simply work in large commits normally.
  • If the recent ratio is above the threshold and the baseline was clearly lower, we raise a signal. Severity is medium if the recent ratio is at least double the threshold; otherwise low.

What to use it for: Discussing incremental commits, reviewability, and whether large batches align with team practice—without assuming every large commit is AI-generated.

Summary

In the app Homepage-style name (where applicable) What it approximates
AI Channels Exploring AI Tooling New AI-named chat channels vs prior activity
AI Output Output Above Usual Pace Sharp rise in daily lines added vs your baseline
AI Velocity Velocity Step-Up PR merge rate and output both up vs baseline
AI Attribution — Commit/PR titles matching org-defined phrases
AI Commit Size — High lines-per-commit vs threshold and your baseline

Together, these help managers notice change early, ask better questions in 1:1s, and align on tooling and delivery norms—always as prompts for dialogue, not automated judgments.

Back to Help Centre Something unclear? Contact us →