Technology Waste Audit · System 01 · the one signal that isn’t arguable
Reconcile What You Pay For Against Who’s Still Here
A login proves someone signed in. It doesn’t prove the seat was worth what you paid for it, and it definitely doesn’t prove nobody signed in through a different door. No single system in a typical business holds the whole truth about what’s actually being used: identity logs show access, not value; finance shows a charge, not who touched the product. One cross-check clears that fog almost entirely — whether the person the seat is assigned to still works here.
The idea in one line: triangulate identity logs, vendor admin exports, and the HR roster before concluding anything is unused. An assigned seat belonging to someone who left three months ago needs none of that triangulation — it’s the one finding in this whole audit that’s hard to argue with.
No single source tells you the whole truth
Four kinds of evidence, each with a real gap the others happen to fill.
What it tells you
- Identity logs (Okta, Google Workspace, Entra): who authenticated to what, and when
- Vendor admin exports: seats purchased, seats assigned, and often a vendor-defined “active” flag
- Finance and expense data: what’s actually being paid for, including tools nobody told IT about
- HR roster: who currently works here, by name
What it can’t tell you on its own
- A login is access, not feature use or business value
- “Active” means something different at every vendor — logged in once, opened a page, or used a paid feature
- A merchant name on a card statement can be opaque, and one charge can bundle several products
- None of the first three knows, by itself, whether the assigned person still works here
The HR roster is what turns the other three from suggestive into conclusive. Everything else in this audit needs judgment. Cross-referencing a seat against an employee who no longer appears on the roster doesn’t.
What each source gives you, and its limits
The one signal that’s hard to argue with
Not every finding in this audit needs the same amount of proof.
Shelfware — a paid seat with no meaningful recent use — is directly measurable in a way that overlap, integration debt, and shadow IT are not. And within shelfware, one version of the finding is nearly unarguable: a seat still assigned and billed to someone the HR roster says no longer works here. No judgment call about what counts as “active enough” is required. The person is gone. The seat isn’t.
Everything else — a current employee who hasn’t logged in for 60 days, a seat that’s technically active but barely used — is real evidence, but it needs the context a departed-employee seat doesn’t: is this person on leave, did their role change, is the tool seasonal. Start with the unarguable list. It’s usually large enough on its own to be worth the audit.
Build the license ledger
Merging a vendor export against an HR roster by hand, tool by tool, is exactly the kind of matching an AI assistant does well — as long as you keep it matching rather than deciding what to cancel:
Export a seat or user list from each tool, and your current employee roster from HR. Upload them together and send these two messages in order.
- “Here’s a seat list from [tool] and our current employee roster. Match each seat to a name on the roster. List separately: seats matched to a current employee, and seats whose name doesn’t appear on the roster at all.”
- “For the seats that matched, add whatever last-login or activity data is in the export. Don’t recommend cancelling anything yet — just organize what we have.”
A worked example
A 40-person marketing agency pulls seat lists from its five paid tools and merges each against the HR roster.
What the merge found
- 11 seats across the five tools belong to people who left in the last 14 months — three of them left before the current office lease started.
- A project-management tool shows 34 assigned seats against 29 current employees; the other 5 are contractors whose engagements ended.
- One design tool’s admin export has no usable last-login field at all — a gap to note, not a finding to guess at.
What’s immediately actionable
- 11 departed-employee seats, annualized cost reconstructed from the vendor invoices: $6,400/year, cancelable this week on 3 of the 5 tools and at the next renewal on the other 2.
- The design tool’s missing activity data becomes a System 02 question, not a System 01 answer.
None of this required a debate about what counts as “active.” It required one list of current employees and five exports, merged once.
A clean license ledger tells you who’s paying for what. It doesn’t yet tell you whether two tools doing similar jobs are actually redundant, or whether the re-typing someone complains about is a training problem or a missing integration. System 02 covers how to tell the difference.