This week I ran a LISA live session on how to compare enterprise datasets using AI. To make it work, I built some free sample ITAM data, and I’m sharing it here. It covers ServiceNow asset data, Intune and Active Directory.
I chose those three because they’re so common. ServiceNow is a popular enterprise tool. It’s got your service desk, incident, problem and change. You might be using it for ITAM and CMDB too, so it’s a good source of intel. Intune is your device management. It pushes out software and patches and looks after devices. Active Directory is your permissions: access, sign-on and so forth.
So between those three datasets, you should have a good picture of what’s going on in your environment. “Should” being the operative word.
Why use sample ITAM data?
Playing with enterprise data isn’t always easy, because of security concerns. If you’re sharing data about individuals, there’s GDPR. If you’re sharing MAC addresses, or anything that looks like a gateway into your organisation, there’s security. This is dummy data with spoofed email addresses. As a result, you can play with it on any AI system or ITAM tool.
It’s in Google Sheets. Duplicate it, then download it as Excel, CSV or whatever you prefer. There are 10,000 devices and about 30,000 line items in total across the three systems. There’s also an export of around 1,100 Adobe users from the Admin Console.
I’ve deliberately loaded it with inaccuracies. In other words, it looks like real life: duplicates, missing devices and inaccurate records all over the place. That makes it a great testing ground for trialling things.
Download the sample ITAM data
The sample ITAM data belongs to a fictional company, Kestrel Group. Here are the four datasets:
- Adobe: 1,129 user accounts
- ServiceNow: 11,661 assets
- Intune: 12,712 devices
- Active Directory: 10,041 records
Two ways to use it
First, start testing your AI skills for ITAM, today.
Second, try it with ITAM tools. Upload it and see what they can do.
Over to you
Let me know how you get on. Tell me how accurate you think the data is, and why.
I’ve also shared the Adobe licence information, so you can try an Adobe optimisation. Tell me how much you estimate you can remove based on the data. Or, more realistically, tell me what you’d need to do next to feel confident about removing that Adobe spend.
Three takeaways from the session
1. Make the AI show its decisions before it shows you the numbers.
Don’t just ask AI, “How accurate is this hardware data?” It will give it a good go. I’ve used ChatGPT and Claude, and Copilot would work for the same exercise. The files aren’t big, even with 10,000 devices. So it doesn’t take long or chomp through many tokens.
However, AI can make very persuasive arguments about information it knows nothing about. It will make assumptions. In the live session, the simple prompt told me the hardware data was 87% accurate. In rehearsal I never got the same answer twice, and every answer was wrong. I know exactly how accurate the data is, because I curated it.
For example, it counted Macs as missing because it couldn’t see them in Active Directory. But Macs typically don’t use Active Directory. The AI did its maths correctly and its judgements wrong.
So test it first. Ask it how it’s going to handle the data, then do the actual analysis.
2. Treat the reconciliation as a structured, auditable process.
Show your working. You don’t need to when you’re practising. In a live environment, though, you need to be able to replicate it. There are two reasons for that:
- You can pick up what the heck you did last time and repeat it. You can also build it into a skill, a repeatable process or a system.
- Somebody else can pick up your work and understand what you did and how. That adds to its credibility.
3. Tie the data to commercial impact.
We’re not doing this as a housekeeping exercise. Keeping things tidy is a benefit, but the goal is commercial impact. There is significant opportunity in messy data. This is where your human judgement as an asset manager really comes in.
You need to be able to show the discrepancies in the data, and the cost-saving opportunities they reveal. Once your data is trustworthy, there are an enormous number of those.
Have fun with the sample ITAM data. Let me know how you get on.
