Usage: how many people use AI and for what

Insights > Usage answers one question: how many people use AI, in which tools and for what. The screen measures use, not productivity and not business value. Everything is a count; what someone typed appears nowhere, and the AI-usage figures carry no user id.

#Period and filters

The bar at the top is the same as on Compliance and Sensitive data. Period picks a window, from Last 7 days to Last year or your own start and end date. With Department insights on, you pick a department next to it; in Standard mode that asks for a period of at least 7 days. Under Filters you pick Tools: at most 8 at a time, with a search field once more than 8 have been seen. The blocks then show only those tools, or say they cannot follow the filter.

#The four numbers

• AI adoption: how many of your seats were AI-active this month, "12 of 40 seats AI-active this month".
• Prompts sent: every message sent to AI tools in the period, with the change against the period before.
• AI tools in use: the number of distinct tools, with how many were newly discovered this period.
• Use in allowed tools: the share of AI use that runs through allowed tools.

AI adoption and usage frequency are monthly figures for the whole organisation; they do not follow the period or the filters, and the screen says so. If you allowed only the business variant of a tool and the extension cannot tell which account is in use, those visits do not count in the share; the figure says how many visits that concerns.

#Where AI use goes

Per allowed tool its share of all AI visits and the number of visits, largest first: the first ten at once, the rest behind Show all tools. At the bottom one quiet row, Not allowed or no decision yet, which leads to Compliance. Every other row opens the tool page; a decision is taken there, not here.

#How often AI is used

Your team by usage frequency this month: Daily, Weekly, Occasionally and Not active, as one bar.

The counting happens on the device. The extension keeps track itself of the days AI was used, and once a month reports only the outcome, as a bare +1 per bucket. The dashboard receives counts per bucket, never who is in one. The breakdown appears from 5 active users; below that the card says only how many team members were AI-active this month. Someone with the extension in two browsers may count twice; the card says so in its footnote.

#Usage over time

Prompts per week as bars and the number of distinct AI tools used as a line. With a tool filter it becomes visits per week, per selected tool.

#What AI is used for

Seven use cases: Writing communication, Marketing & content, Summarizing, Rewriting & editing, Translating, Coding & technical and Analysis & knowledge. Click a use case to see which tools carry that work, with each tool's prompt count.

The label comes from the first message of a conversation: it is classified on our own servers at European providers and discarded at once. Only the label counts; the prompt itself is never stored. When classification fails, the prompt counts as Unclassified, and the totals count every prompt.

#Per tool

Click a tool and you are on its page in the register, with under Usage the figures you want first: Visits with the period before next to it, Share of all AI visits, Days active, Last seen and Installs this month. The last one counts browser extensions that used the tool this month, not people; the desktop app is not counted. With one to four installs there is no number but "fewer than 5", so that a figure like "1 of 38" at a disallowed tool does not point at one colleague.

During a trial you can switch on demo data from that card; the screen then shows the invented figures of an example organisation, with a banner saying so.