Pensero connects AI activity to each person's profile using the identity provided by each AI tool.
If the organization-level integration has a problem, the person's AI identity is missing or incorrect, or the AI tool sends data differently than expected, AI adoption and AI Impact may look incomplete.
This article walks you through the checks to run when a teammate's AI data looks missing or incorrect.
Who this is for
This article is for Managers and Directors without Pensero admin permissions.
You can complete the user-level checks yourself. When a problem requires organization-level changes, we'll show you when to contact your Pensero workspace admin or Pensero Support.
AI integrations supported by Pensero
Pensero supports multiple AI coding assistants and agents, including:
GitHub Copilot
Cursor
Claude Code
Gemini Code Assist
OpenAI Codex
Devin AI
AWS Bedrock
The available integrations can change as we add support for more tools. See the current list at: https://pensero.ai/integrations
If your organization uses an AI tool that isn't currently supported, email [email protected] and tell us which integration you'd like Pensero to support.
Using Claude Chat or Cowork?
Pensero tracks Claude Code only: in the terminal, in your code editor, and in the Code tab of the Claude desktop app.
The Chat and Cowork tabs of the Claude desktop app are not tracked. Work done there does not show in AI impact or AI adoption, so your AI % can be 0% even if you use Cowork every day.
To have your AI work counted, use Claude Code (terminal, editor, or the Code tab with Local selected).
Diagnosing AI Data Issues
Follow these steps in order.
Step 1: Confirm the AI integration is connected
First, make sure Pensero has a working organization-level connection to the AI tool.
Find the AI tool you're troubleshooting under Coding AI Assistants.
What to look for
Working
The integration is connected and operating normally. Continue to Step 2.
Needs attention
There is a problem with the organization-level integration.
Because this article is for managers without admin permissions, ask your Pensero workspace admin to open the integration, review the warning, and fix the connection.
After the connection is fixed, allow up to 24 hours for Pensero to process new or reprocessed data before checking the teammate again.
If your workspace admin can't resolve the integration issue, feel free to contact us through the Product Chat. Our team will be happy to investigate and help you resolve it.
Step 2: Confirm the teammate's AI user mapping
If the organization-level integration is healthy, check whether the affected teammate is mapped to the correct identity.
Open the Contributors Page: https://pensero.ai/hr/people
Find the teammate whose AI data looks missing or incorrect and open their profile.
Go to Integrations tab.
Shortcut: the same Check is available on Data health. Use the My data tab for your own accounts, or the Git identities column on My reports data for your team.
You'll see the AI identities currently connected to that person, such as:
GitHub Copilot user
Claude Code user
Cursor user
Other supported AI identities
If the mapping is missing
Your own AI accounts: open Data health → My data, click Link more accounts, pick the AI tool (for example GitHub Copilot) and enter your username or email. No admin access is needed. If you changed your username, add the new one and keep the old account: it holds your past usage.
A teammate's accounts:
Click Add integration user.
Select the AI provider.
Enter or select the teammate's correct username or email.
Click Add integration user.
Run Check against the new mapping.
Once the mapping is added successfully, allow up to 24 hours for processing.
Then review the teammate's AI Impact again.
Check the existing mapping
Confirm that the username or email shown for the AI integration belongs to the correct person.
Then click Check next to the integration.
Pensero will run a healthcheck against that mapping.
If the check passes
The identity is recognized and Pensero can process data associated with it.
Continue to Step 3.
If the check fails
Review the healthcheck result or error message.
Review the error message and update or manage the username or email as needed.
If the mapping looks correct but the check still fails, contact us through the Product Chat with the user and integration details.
Step 3: Confirm which AI tools are actually contributing data
Once the organization integration and user mapping are validated, open the teammate's AI Impact tab.
Open the Contributors Page: https://pensero.ai/hr/people
Find the teammate whose AI data looks missing or incorrect and open their profile.
Go to AI Impact tab.
Review the AI activity for the affected period.
At the bottom of the page, review the activity table to confirm which providers are sending AI data for this teammate.
For example, you may see activity attributed to:
Claude Code
GitHub Copilot
AI co-author
Other connected AI providers
This is an important check because a teammate can have several AI identities mapped while only some of those tools have recorded activity during the selected
period.
If AI co-author doesn't appear for a teammate
AI co-author shows up only when a teammate's merged PRs have commits signed by an AI tool (a Co-authored-by line in the commit message).
If you don't see it in their AI Impact raw data, it means one of two things:
The teammate isn't using an AI tool that signs commits.
The teammate uses AI, but the co-author line is removed before merge. For example, they delete it from the commit message or squash without keeping it.
In both cases Pensero can't track that work as AI-assisted.
We recommend that engineers keep the Co-authored-by line in their commits. If your team uses AI and you still see no AI co-author data, ask them to check that the line is kept when they merge.
If the teammate's AI tool is connected to Pensero (for example Claude Code telemetry), their AI activity still shows under that tool.
If the expected provider appears
Pensero is receiving activity from that provider. If a number still looks wrong, or older activity is missing, the cause is usually how that data is collected or calculated rather than a broken connection — see Why is AI usage % lower than expected? below.
If the expected provider doesn't appear
Confirm:
The organization integration is healthy.
The correct user identity is mapped.
The healthcheck passes.
You're reviewing the correct date range.
You've allowed up to 24 hours after changing a mapping or connection.
If all five are true and activity is still missing, contact Pensero through the Product Chat.
Why is AI usage % lower than expected?
Pensero calculates AI usage % in weekly windows. AI lines generated in a given week are credited only against lines merged in that same week.
If AI lines are generated in one week and merged in another, they are never matched to each other. The result is 0% — in the week the code was written, in the week it was merged, and in any wider range that contains both.
Example
1,000 AI-generated lines written on Monday 10 August
The same 1,000 lines merged on Tuesday 25 August
Date range selected | AI usage % |
9–15 August | 0% |
20–26 August | 0% |
9–25 August | 0% |
Widening the date range does not recover the number. Pensero sums the weekly results across the range rather than dividing total AI lines by total merged lines over the whole period, so two weeks that each return 0% still return 0% combined.
The gap does not need to be large. Code generated on a Friday and merged the following Monday falls into two different weeks and will read as 0%.
Why does Pensero count AI lines by week?
AI usage % is built from two separate data sources, and the weekly window is how Pensero reconciles them. Generation activity comes from the AI tool itself, through OpenTelemetry or the provider's API. Merge activity comes from your git provider. Neither source carries an identifier that links a specific generated line to the specific pull request it eventually lands in.
Pensero therefore reconciles the two on a weekly window: AI lines generated in a week are matched against lines merged in that same week. Work that crosses a week boundary between generation and merge cannot be reconciled, and does not appear in AI usage %.
This means AI usage % is best read as a directional adoption trend across a team over time, not as a complete count of AI-assisted work. Teams whose work routinely takes more than a few days to reach merge will see the metric understate their actual AI usage.
Does Pensero count AI lines from repositories that are not connected?
Yes. Claude Code telemetry does not say which repository the work was done in, so all AI lines from the company account count for that person, personal projects included. AI adoption compares these lines with the lines the person merged in connected repositories, and it never goes above 100%. Work in other repositories can make AI adoption look higher. To keep the number fair, connect every repository your team works in.
Checks to run before reporting a low % AI usage problem
Confirm the merge dates of the pull requests in question, not the commit dates.
If generation and merge fall inside the same week, this is not the explanation — return to Steps 1–3 above.
A team with long review cycles, a weekly release train, or a Friday code freeze will hit this consistently. That pattern is expected behaviour, not a broken integration.
If generation and merge are inside the same week and you are still seeing 0%, contact us through Product Chat with the teammate and the affected pull requests.
How much history is available, and which metrics arrive at all, differ by integration. If activity is missing from before a certain date, or a teammate shows fewer metrics than expected, see How Pensero collects AI data in Understanding AI Economics.
Missing data that is expected
Some AI tools don't share every type of data with Pensero. If a tool doesn't send a metric, Pensero can't show it.
This is expected. It doesn't mean the integration is broken or that the teammate is set up incorrectly.
The table below shows what we can't get today for specific integrations:
AI tool | Data Pensero can't get today | What you'll see |
OpenAI Codex | Usage from before the setup | No historical data. Codex data starts on the day each developer's computer has the Pensero configuration (otel). |
Claude Code | Usage from before the setup | No historical data. Claude data starts on the day each developer's computer has the Pensero configuration (otel). |
GitHub Copilot | Cost | No cost for Copilot in AI Economics. |
Custom integrations | Tokens and cost | No token or cost data for that tool. |
Claude desktop app: Chat and Cowork tabs | All usage | Nothing from these tabs. They don't count in AI impact. Only the Code tab is tracked. |
Custom integrations are AI tools connected through a custom setup built for your organization, instead of one of the standard integrations listed above.
We update this list as AI providers share more data. If you see missing data that isn't in this table, follow the steps in the next section.
Claude Code data stopped after a plan change
If Claude Code data stops after someone changes plan (for example, from API keys to Pro or Max, or from a personal plan to a Team or Enterprise seat), the setup for the old plan no longer applies. Apply the setup path for the new plan, restart Claude Code, run /login and verify. See Choose your setup path and Using Claude Code with more than one plan.
What if the integration is healthy but data is still missing?
If you've completed the checks above and a specific teammate is still missing AI activity, Pensero Support can investigate the underlying data.
Before contacting us, confirm that:
The organization's AI integration is connected and doesn't show Needs attention.
The teammate has the correct AI identity mapped.
The mapping passes Check.
You've selected the correct date range in AI Impact.
You've confirmed whether the expected provider appears in the teammate's activity.
You've allowed up to 24 hours after fixing or adding a connection or mapping.
If everything looks correct but data is still missing, contact us through the Product Chat with the teammate, integration, and missing activity details.
Advanced troubleshooting
This step is especially relevant when historical AI data is missing, even though the integration and user mapping are correct.
Pensero has an additional diagnostic snippet that can help identify and recover missing historical activity.
Contact us through the Product Chat with the affected AI provider and teammate, and we'll guide you through the process.