AI summaries

Any report can be summarized in plain language by an AI model. The summary is a draft you read and sign off. Nothing is written to your data automatically, and the model is told to state only the numbers already in the report.

This is available on the Professional plan and above.

Draft a summary

Open a report (a built-in report or a custom report).
Choose Summarize with AI.
If your organization allows it, pick a model from the dropdown.
Optionally add a short Focus note to steer what the summary emphasizes.
Choose Generate draft.

The draft appears with the name of the model that wrote it. Read it, then Accept it, Edit it and save your version, or Dismiss it. You can also Regenerate to try again. Whatever you choose is recorded, so an accepted summary always names the model that produced it and the person who approved it.

What keeps it accurate

The report is rebuilt on the server, so the model only ever sees the figures the report computed. It is instructed to state those figures and nothing else: no new totals, no estimates, no rounding.

As a safety net, every draft is checked against the report’s figures. If the draft states a number that is not one of them, the summary shows a warning listing those numbers so you can check them before you use the draft.

The summary is written in plain language, the kind a working biologist reads. It does not give advice or make management recommendations; it describes what the figures show.

The focus note

The focus note is optional. Use it to point the summary at what matters to you, for example “what changed at the largest stations”. The note is a hint, not a command: it cannot change the safety rules or make the model state numbers that are not in the report. If a note looks like an attempt to do that, or contains harmful content, it is flagged and recorded, and a clearly harmful note is declined.

Choosing a model

Your organization can offer more than one model. An administrator sets the default and decides whether members may pick a different one per summary (see organization settings). The options can include:

The built-in in-region AI: runs inside the same service in your region, with no data sent to an outside provider.
A managed frontier model, when your deployment offers one.
Your organization’s own provider key (Anthropic, OpenAI, or Azure OpenAI), added by an administrator.

If your organization keeps data in the in-region runtime, only the built-in in-region option is offered, because it does not send data to an outside provider.

Limits

There is a generous monthly limit on how many summaries an organization can generate on the Professional plan; higher tiers are unlimited. If a summary looks cut off because it reached its length limit, a note appears above it; choose Regenerate to produce a fresh one.