To measure software adoption across employees, define what active use means for each application, count the people who meet that definition on a fixed window against the people who should, and track it monthly next to license counts and help desk volume. Logins tell you who showed up. Adoption is who does the work in the tool, and the gap between the two is where the budget leaks.
TL;DR:
Software adoption across employees is the share of people who are supposed to use an application who actually use it for its intended job, measured over a fixed period. That definition has three parts IT teams tend to skip: a denominator (who is supposed to use it), a verb (what counts as use) and a window (over what period). A tool with 1,000 licenses, 700 monthly sign-ins and 200 people completing the workflow it was bought for has 70 percent coverage and 20 percent adoption. Only the second number tells you whether the purchase is working.
The steps below build that measurement from the systems an IT team already runs. None of them needs a new analytics product. They do need an agreed definition for each app, which is step 1 and the step most teams try to skip.
Write one sentence per app that names the audience and the action. For a CRM: "every quota-carrying rep logs at least one activity on an opportunity each week." For an expense tool: "every employee who files an expense does it in the tool, not by email." For a ticketing system: "every service desk agent resolves tickets in the queue, not in Slack." The audience is the denominator and the action is the verb.
Keep the action to the one thing the application was bought to do. A tool with twenty features has one or two that justify the spend, and adoption is about those. Record the definition somewhere the business owner of the app can see it and agree to it, because the number you report in step 7 will be argued with, and the argument is easier if the definition was shared first.
Your SSO layer already knows who signed in to what. Start there, because it covers every app behind single sign-on with one export and no vendor conversation. As of November 2026, Okta's Application Usage report "contains data about who has signed in to an app integration during a specified time period," and Okta notes it "retains events for 90 days, so the earliest available date range for the report is three months prior." Microsoft Entra, Google Workspace and other identity providers expose sign-in logs with the same shape.
Export the report for each app on your list, count distinct people who signed in during the window, and divide by the audience from step 1. That is coverage. Two warnings: a sign-in through SSO is not the same as doing anything inside the app, and apps outside SSO (shadow IT, legacy tools, anything with its own password) will not appear at all. Coverage is the floor of your measurement, not the answer.
Pick a window and never change it between reports. Monthly active users (28 or 30 days) is the common default; weekly active is better for tools people should touch daily, such as a CRM or a ticketing queue. As of November 2026, Microsoft's Adoption Score documentation states the score "is updated daily and reflects user actions completed in the last 28 days (including the current day)," which is a reasonable template for any app.
Count active users against the step 1 definition, not against sign-ins. If the definition is "logs an activity on an opportunity," the active count is the people who did that in the window. Report it as a rate (active divided by audience) and as a raw number, and keep both by team. The team view is where you find the departments that bought in and the ones that are still running the old spreadsheet.
Most enterprise applications ship usage reporting in the admin console, and it is the only place you will see what people did after signing in. Salesforce's Lightning Usage App and adoption dashboards, Microsoft 365's Adoption Score and usage reports, and the equivalents in ServiceNow, Workday and the rest all report feature-level activity. For a CRM-specific version of this step, see how to improve Salesforce user adoption, which covers the Salesforce reports in detail.
Pick two or three depth metrics per app that map to the step 1 verb. Records created or updated, workflows completed, reports run, documents shared in the right place rather than by email. Avoid the trap of tracking every feature: the point is to see whether the core job moved into the tool. One caution on aggregate scores: Microsoft notes its Adoption Score insights "are calculated using data at the organizational level, not the individual level," so a score like that tells you whether the organization is moving, not which teams or people need help. For that you need the team breakdown from step 3 and the signals in step 6.
Put the licensed count, the coverage count and the active count side by side for every paid app. The gap between licensed and active is the cost of non-adoption, and it is rarely small. As of November 2026, Zylo's 2026 SaaS Management Index, based on an analysis of more than 40 million SaaS licenses and $75 billion in spend, reports that "organizations leave an average of 36% of their SaaS licenses unused" when measured against recommended utilization levels.
Reclaiming a license is the easy decision. The harder one is deciding whether a person who has a license and never uses it should be trained or should lose the seat, and that depends on whether they are in the step 1 audience. Someone outside the audience is a reclaim. Someone inside it is an adoption problem, which is what step 6 is for.
Usage data tells you who is not adopting. It does not tell you why. Three sources do. First, the help desk: tag tickets by application and count them per active user per month, and watch which apps generate the most "how do I" questions. Second, short surveys to the step 1 audience, asked at rollout and again at 90 days. Microsoft's own account of measuring its internal Copilot rollout, published on its Inside Track blog, describes starting with monthly active usage and then moving to weekly and daily frequency metrics alongside satisfaction surveys and listening sessions, because usage alone did not explain what was happening.
Third, in-app signals. An AI adoption agent such as Worknet sits with employees inside each application, answers their questions in context and records where they stall, which gives IT a per-app, per-workflow view of friction without waiting for a ticket. If you already run a digital adoption platform, its guide completion and drop-off reports serve a similar purpose for the workflows you have built guides for; the guide to measuring digital adoption platform ROI covers how to read those numbers without overstating them.
Each app gets one page with five lines: audience size, coverage rate, active rate, two depth metrics and tickets per active user. Add a target for the active rate, the name of the business owner, and the single biggest reason people get stuck from step 6. Review it monthly with the owner, and keep the same window, the same definitions and the same page layout for at least two quarters so the trend is readable.
Set targets per app, not one number for the whole estate. A system of record that every rep must use should be close to 100 percent active among its audience. A specialist tool used by one team a few times a quarter should not be, and judging it on a monthly window will produce a false alarm. The scorecard is also where you decide: train, change the process, reclaim licenses, or retire the tool. For how the broader discipline is defined, see what digital adoption means, and for the vendor landscape if you are considering tooling, the digital adoption platform roundup.
Expect two to four weeks to get the first scorecards out for the five to ten applications that carry most of your spend, most of it spent on step 1 definitions and on finding the right admin report in each vendor console. After that it is a monthly routine of an hour or two per app. The first report will be rough. The third one, with a trend line and a stable definition, is the one leadership will act on.
Worknet is an AI adoption agent for the software your customers use and the software your employees use. It sits with every user inside the app, offering proactive support and onboarding. It answers with context, acts on the user's behalf within the permissions you set, and brings in the right person from your team when needed. For SaaS companies, what it learns becomes expansion signals for sales, risk alerts for success, and fixes for product. For IT, it rolls out across the apps you buy with no vendor changes and shows where employees get stuck. An AI-native alternative to Pendo, WalkMe and Appcues.
For the measurement described here, the agent adds the layer that usage reports cannot: it sees the question an employee asked, the step they could not complete and the moment they gave up, inside the application, and surfaces that back to IT by app and by workflow. It runs alongside your identity provider and the vendor admin reports rather than replacing them. Pricing is quote-based.
Define what active use means for each application, then count the people who met that definition on a fixed window (28 or 30 days is the usual default) and divide by the number of people who should be using the tool. Pull sign-in data from your identity provider for coverage, feature-level data from each vendor's admin reports for depth, and help desk tickets per application for friction. Report the result on one scorecard per app every month.
The core metrics are coverage (share of the intended audience who signed in during the window), active rate (share who completed the application's core job in the window), depth (two or three feature or workflow counts that map to that job), license utilization (active users divided by paid seats) and tickets per active user. Together they show whether people showed up, whether they did the work in the tool, and whether it cost the help desk anything.
It depends on the application and the audience, so set a target per app rather than one number for the whole estate. A system of record that every member of a team must use should be close to 100 percent active among that audience on a weekly or monthly window. A specialist tool used occasionally by one group should have a lower target and a longer window. There is no published benchmark that applies across applications, and a single average hides the few apps that matter.
Only partly. Identity providers such as Okta and Microsoft Entra report who signed in to each application, which gives you coverage across every tool behind single sign-on in one export. A sign-in does not show what the person did inside the application, and tools outside SSO do not appear at all, so SSO data is the floor of the measurement. Add each vendor's admin usage reports for depth and your help desk data for friction.
An AI adoption agent such as Worknet sits with employees inside the applications they use, answers questions in context and records where people stall, which gives IT a per-app, per-workflow view of friction without waiting for a ticket. It complements identity-provider sign-in data and vendor usage reports rather than replacing them. Worknet is quote-based, so its cost has to be weighed against the licenses and support time it recovers in your own estate.
.png)