It's totally normal that you got the AI "hot potato" and got no hot AI data with it. The reason: the person who bought the AI tool is usually not the person who gets tasked with AI adoption, which leaves a gulf between the person who can get the usage data and the person who needs it.
If your first readout is next week and all you have is a seat count, meet with your IT person and call in a favor. Ask what they can pull from the admin console: seats, usage, and a few basics. On a week's notice that's the best you'll do.
Given more time, the raw usage data takes you much deeper, and here's how.
1. Congrats! You get to move AI usage (and business results)
When you get handed the AI adoption hot potato, it means the senior folks trust you to tackle poorly defined projects that span people, tools, and technology, the kind no one else in the org can. You might not be the first person to get it: a CTO who didn't do too much with it, or somebody in IT, or somebody in HR may have held it before you.
What the business wants in terms of results can mean a lot of different things. Typically it's one of the following:
- Enabling people. Improving individual usage sophistication, token efficiency, process, and more.
- Improving the way your tools work. Building skills, centralizing skills, economizing usage and seats.
- Measuring business results. Using AI to improve the KPIs the business cares about. Things like: the number of accounts a CSM can manage, sales pipeline velocity through a certain deal stage, or the enrichment quality of a new lead.
You should ask the person who gave you the project what goals they have in mind. But, to be honest, more often than not you'll hear "you're smart, you'll figure it out." So the first imperative is to set a plan. (Your first 90 days as AI owner walks through that plan.)
2. What you can still do without AI usage data
When people who own AI adoption don't have data, they tend to do three things, and all of them are well intentioned.
- Field a survey. Ask what people like about the AI experience, where AI could be more useful, and which manual tasks they could target with AI. It's self-reported, with <100% response rate, and people are often unaware of their own cleverness in how they use AI (or overindex on it). But it's far better than nothing, so a survey is a good idea if you have no other internal AI adoption data.
- Run enablement. Trainings where power users from the org teach their peers how they use AI. But what works for one person may not work for everybody, and to meet people where they are you need data beyond self-assessment.
- Cap spend (optional). The AI project often comes with a CFO requirement to "limit our spend," so the owner gets a little budget data from finance (not IT) and rolls out a spend cap. That's a speed limit, and it keeps folks from creating surprise runaway costs. But it's a blunt tool in a nuanced space: it caps a great power user who happens to be a frontline employee, it needs a new process for lifting the cap, and it still allows wasteful use by anyone under an arbitrary monthly figure.
3. Ask IT for data from your Lab of choice first (e.g. Claude)
We recommend starting with the lab your people already use, whether that's Claude, ChatGPT, Gemini, or Copilot, rather than a tool that happens to have AI in it, like HubSpot, Salesforce, or Glean. After a year or two, people are probably in it far more than you think, which means there's a rich history to look at and a live stream worth capturing from today forward.
As an example, to get any data on Claude usage, you'll want to talk to your Primary Owner (you can see who this is in your own Claude settings) and ask them to screenshot some admin views for you.
If that's the first data you've ever seen, it's fine for your very first executive readout. Beggars can't be choosers, and we take what we can get. It's also an incredibly high-level roll-up that omits the richness and cleverness of your team having used these tools for months now, and it's nowhere near enough to drive AI adoption the way you're expected to.
The fuller version, telemetry, exports, and a cost API pulled into one place (we have a different post on how to build it), is real development work, and IT will probably tell you it's more than they budgeted for.
4. Five AI usage metrics to track first
For a moment, let's imagine IT can build the real thing so you can get your metrics wishlist.
As a thought experiment, picture a spreadsheet of AI usage, a simple frame of reference for these data.
| Metric | What it is | What to look for |
|---|---|---|
| Users | Who is using AI (the rows) | How far adoption has spread, team by team |
| Usage volume | How much each team uses, over time | Heavy users, and new heavy users |
| Topics | A tag on each conversation for the work being done (prospecting, demo prep, QBR prep, renewals) | How each team's time splits across its core work |
| Skill usage | Runs per skill, over time | What's used, what's shared, whether central skills get used |
| Spend | Tokens and dollars by team, topic, and skill | Skills that are used a lot and cost a lot |
With those five data points you can do a heck of a lot! Run a pivot table on that spreadsheet to learn:
- The skills that get used
- How people spend effort and time with AI
- Where teams use AI for work outside their core function
- Who the power users are that go deep in particular parts of their job
- How usage is changing over time by topic
The very first step is to get the five metrics, start charting them, and make a set of five charts.
5. The choice: be the data-driven leader you are...or wing it
Remember, you got the AI adoption charter because you're a senior person who can get things done with people and technology. So you have a choice: operate in the way that earned you the AI adoption charter, or... wing it.
Naturally, we recommend you be true to yourself. If you owned any other number in the business, you'd start with the data analysis to understand where you are.
The move is to talk to your IT team and see whether they're able to get you this data. Very likely they won't take on the size of build you need, even if the links above make it easier.
So we at SightLift are always happy to sit down with you and your IT team to guide them through it and discuss the build. Of course we're biased, but we think SightLift is a much faster way to get you where you need to go, operating the way you know how to best. We'd be thrilled to explore what that looks like with you.