FAQs

Usage numbers only tell you who opened a tool. To know whether people use AI well, look at the work itself: give them tasks like their real job, have senior people review the output, and check whether they catch the mistakes AI makes. That is what Osava measures, alongside what the firm spends and how much each tool is used. More on this: Why usage data won't tell you if AI is paying off.

Compare what you pay for AI with what it changes. In practice that means four things: cost (what you pay against what gets used), time (hands-on time against total time to finish), quality (senior-reviewed scores on real tasks) and risk (whether flawed AI output is caught before a client sees it). Time saved is the minimum. The other three show whether AI is improving the work. More on this: How to measure the ROI of AI in a professional services firm (2026).

It depends on how many of the seats are used, and for what. Start by comparing seats paid for with seats in use, because unused seats are pure cost. Then look at the jobs people use the tool for and whether time or quality improves. You may also find people using tools the firm hasn't licensed. More on this: Paying for AI nobody uses: unused licences and unlicensed tools in small firms.

It depends on how much of a task is hands-on work. AI speeds up the part where someone is actively working. It doesn't shorten time spent waiting for reviews, approvals or client input. That's why Osava measures both hands-on time and total time to finish.

Ask them, routinely. Osava asks people at the end of every task which tools they used, so unlicensed tools show up without scanning devices or installing anything. Because it relies on what people report, it can only undercount, but it shows where client data might be going. More on this: Paying for AI nobody uses: unused licences and unlicensed tools in small firms.

Test the skills on work that looks like their real job, not with a quiz or a self-rating. The skills that matter most are checking AI's output, briefing it clearly and deciding what to hand over. Firms use Osava to test everyone on tasks from their sector. Individuals can try the free Osava challenge in about five minutes. More on this: How to check AI output before it reaches a client: a checklist.

The core ones are checking AI output for errors, briefing AI clearly, researching with reliable sources, and knowing which work to hand over and which to keep. They sit on top of skills firms already value, such as judgement and clear writing. The mix keeps shifting as the tools change, which is why it's worth measuring more than once.

No. Osava runs in the browser, and your people keep using the AI tools the firm already provides. There is nothing for IT to deploy.

About an hour. Add the AI tools you pay for and what they cost, invite your people, and confirm which AI jobs your firm does from a ready-made list for your sector.

About two hours each, across six tasks built from real work in your sector.

Short pieces of work like the ones your people do every week, built for your sector. They test judgement with AI: what to hand over, how to brief it, and how to check what comes back. We walk through them in a demo.

Against fixed criteria for each task. Some answers are first marked by an AI model, and Osava staff check a sample. Your senior people decide who they would trust to work unchecked. Anyone can ask for a mark to be reviewed.

Any. Your people use the AI tools the firm already provides, and you record which tools you pay for. Osava doesn't connect to your systems.

No, it's your choice. With a comparison group, the report separates the effect of AI from everything else going on in the firm. Without one, it shows how results changed from one round to the next, and says clearly that there is no comparison. More on this: How to prove AI training worked: use a comparison group.

A firm report that updates with each round of testing. It shows spend against use, the tools people use that the firm hasn't licensed, hands-on time against waiting time, and capability scored out of 100, with skills gaps by team. Each round shows what has changed since the last, so you can see whether training and new tools are paying off.

We build the tasks for each sector, so they reflect the work your people actually do. If your sector isn't covered yet, a demo is the quickest way to check the fit.

Yes. The free Osava challenge takes about five minutes and covers checking AI's work, briefing it and deciding what to hand over. Results are private and never shared with an employer. Take the free challenge.

Your firm's Osava administrators see each person's results by task. Free challenge results are never shared with employers.

In London, with our database provider. Our privacy policy lists every provider we use and where they are.

No. Firms are never named. Comparisons across firms only use combined figures, and only when at least five firms are in the group.