Complete systemBuilt to be run. Everything below is yours to use, with or without us. The last section is a prompt you paste into any AI assistant along with your own admin and finance exports.

The Technology Waste Audit

Find out what you’re actually paying for, whether it’s being used, where two tools are quietly doing the same job, and what to do about each kind of finding. This is the complete system in one place: the checklist, the worksheets, the finding format, and the exact prompts, so you can run it on your own software stack this week. One example runs through the whole page, a 40-person marketing agency with five paid tools, so every blank has a filled-in version next to it.

You need seat or user exports from your paid tools, a current employee roster, and the willingness to check a plausible-looking finding against the people who actually use the tool before recommending anything about it.

What you are looking for

Software waste isn’t one thing. It’s at least three different problems that happen to show up on the same spend report, and each one needs different evidence and a different decision.

Reconcile. What’s assigned and billed, against who actually still works here. The one directly measurable finding in the whole audit.

Map. Where two or more tools do the same job, and whether that overlap is real waste or a legitimate difference in team, workflow, or migration stage.

Trace. Where the same fact gets typed into more than one system, back to the specific missing or broken connection causing it.

Decide. Cancel shelfware, merge confirmed redundancy, and for anything unsanctioned, ask why it’s there before deciding what to do about it.

The output is a set of findings sorted into separate, labeled buckets — not one headline percentage. No published figure for “how much SaaS spend goes to waste” clears the bar of an independently verified market statistic; every one comes from a SaaS-management vendor’s own managed portfolio, with its own definition of “unused.” A number built from your own seats, your own roster, and your own contracts is worth more than any of them.

Pick the scope

Name every tool in scope and the decision this is for, before pulling a single export.

Example

Scope: all paid software tools with more than five seats, plus anything found on the corporate card that isn’t already tracked

The decision this is for: whether the software line item can absorb 20% headcount growth without a proportional cost increase

Write these two lines before anything else

Scope: ______

The decision this is for: ______

An audit that doesn’t end in a decision is an inspection, and inspections get filed. Name the decision now.

Evidence checklist

Gather it before you interpret any of it. No single source is enough on its own — that’s the reason this checklist has four parts.

From the systems

Seat or user list from every paid tool in scope, with whatever activity or last-login field the vendor exposes
Identity-provider access logs, if you centralize through Okta, Google Workspace, or Entra — note each one’s retention limit before relying on it
AP, corporate-card, and expense exports covering the review period
Current employee and contractor roster, with start and end dates
Contract terms and renewal dates for anything you might cancel or downgrade

From the people using the tools

Who owns each tool, and why they chose it over an existing option
What they re-key from one system into another, and how often
Any tool they pay for or use that IT doesn’t know about, and why
What they’d change first if it were entirely up to them

How to talk about this with the team

A software audit reads as “we’re about to take your tools away,” and people respond to that by hiding the tool they’re not supposed to have instead of explaining why they needed it. Say this, or something like it, before you pull a single export.

Say: “I want to understand what we’re actually paying for across our software, including anything that was never officially approved. This isn’t about catching anyone out. If you’re using something IT doesn’t know about, I’d rather hear why you needed it than find out from a card statement — there’s a real chance the fix is making it official, not taking it away.”

Then do this: review the overlap and shadow-IT findings with whoever owns each tool before any decision is final, and treat “why this one” as a real question deserving a real answer, not a formality on the way to a predetermined cancellation.

The findings are about the stack, not about who bought what without asking. If one turns out to be about a specific person’s judgment, that’s a management conversation, and it doesn’t go in the brief.

Know what the evidence will not show you

Whether a login means value. Identity logs show access, not whether the person used a feature that mattered. Corroborate with the vendor’s own activity data where it exists.

Anything outside the identity provider. A tool paid on a personal card, or used without SSO, is invisible to Okta, Google Workspace, and Entra alike. Only finance data and asking people directly will surface it.

Why a tool was chosen. A spend report shows that a purchase happened, never why. That answer only comes from the person who made the call.

A finding sourced only from a spend or login export is worth less than one confirmed with the person who actually owns the tool.

Getting your evidence into an assistant

The step nobody writes down, and the one that stops most people before they start.

Getting the files out

Most systems have an Export or Download button. Your admin consoles (Okta, Google Workspace, Entra), each vendor’s billing page, and your accounting or expense system will each have an export.
Email: copy the thread, or forward it to yourself.
Can’t export it? Screenshot it — these assistants read images.
Nothing exports? Write down dates, counts, amounts, who did what.

Which assistant, how many chats

ChatGPT, Claude, Gemini, or Copilot — pick one that takes file uploads.
One conversation per process. Upload the evidence first, then the prompts, in order.
New process, month, or account? Start a fresh chat.
Upload failing or replies drifting? Send less at a time.

Before any of it leaves your building

Check first: client material may be covered by your engagement terms or privacy law. Sending it to a third-party assistant is your call, with your lawyer if you have one.

You probably don’t need the names: dates, counts, and who did what are usually enough. Swap in Client A, Client B before you upload.

Check the setting once: consumer tiers may train on what you send; business tiers usually don’t. Worth checking before you start.

None of this is a reason to skip the exercise. It’s the reason to spend ten minutes on the first upload deciding what needs to be in it.

Reconcile what you pay for

Full method in System 01. Start with the one signal that’s hard to argue with.

One row per seat

Tool, assigned to, still employed here (yes/no), last login, annual cost, contract lock-in

Merge every seat list against the current roster once. A seat assigned to someone no longer on the roster is the closest thing in this audit to an unarguable finding.

Map the overlap

Full method in System 02. A category match on a spend report is a candidate, not a finding.

Legitimate reasons for overlap

Genuinely different team workflows
A client or compliance requirement one tool meets and the other doesn’t
A migration already underway

Check for duplicate entry too

What gets re-typed between two disconnected systems
Why there’s no connection: no API, nobody owns it, it broke
Cross-link the Process Waste Audit for the general pattern; trace this one to its specific technical cause

Decide: cancel, merge, or ask why

Full method in System 03. Shadow IT gets a different first question than the other two.

Shelfware: cancel or downgrade

Redundancy: merge, after pricing the migration

Shadow IT: ask why, then sanction, replace, or shut down — in that order

Add it up without lying to yourself

This is the document people argue with, so keep the labels honest about what kind of number each one is.

Four buckets, never combined into one figure

Verified recoverable: a cancellation or downgrade you can execute this week.

Renewal-period savings: real, but locked until a contract date. Log it, don’t forget it.

Potential exposure: looks like shelfware but needs a workflow or contract check before you can be sure.

Productivity cost: labor and error cost from fragmentation. Real money, but it’s not a software invoice saving, and it goes in a different line than the other three.

What the finished brief contains

If it’s missing any of these, it isn’t finished.

In it

The scope and the decision this was run to inform
The license ledger, with departed-employee seats called out separately
The capability map, with a reason recorded for every tool that stays
Duplicate-entry findings traced to a specific missing or broken connection
A decision for every finding, using the right ladder for its kind
The four savings buckets, kept separate

Not in it

A borrowed “X% of SaaS spend is wasted” figure from a vendor report
One combined dollar figure mixing verified savings with potential exposure
A redundancy finding based only on category match, with no capability map behind it
A shadow IT finding shut down without asking why it existed first
Any vendor’s “active user” count treated as if it meant the same thing as another vendor’s

Example — the technology waste brief, on one page

Scope, decision: all paid tools over five seats, plus card-expensed software; whether the line item can absorb 20% headcount growth without a proportional cost increase.

Shelfware: 11 departed-employee seats. $1,900/year verified recoverable now, $4,500/year at renewal, remainder pending a vendor usage export.

Redundancy: merge Asana into ClickUp. No contract lock-in, two afternoons of migration work.

Duplicate entry: connect Monday to accounting directly — the API already supports it. Removes a 40-minute weekly re-entry task.

Shadow IT: an unofficial file-conversion tool, sanctioned rather than shut down — two key clients require the export format it provides, which the sanctioned design tool doesn’t.

One page, four kinds of finding, no borrowed percentage. The headcount-growth question gets a real answer: yes, once the confirmed shelfware and the Asana merge are both actioned.

The master prompt

Paste this into any AI assistant along with your seat exports, roster, and spend data. It runs the whole method in order and will not hand you a borrowed waste percentage.

Run the full diagnostic

I want you to help me find where our software spend is going
to waste. Your job is not to quote me an industry percentage
for how much SaaS spend typically goes unused. Your job is to
reconcile what we're actually paying for against who's using
it, map real overlap against legitimate differences, and
recommend the right decision for each kind of finding.

I will give you seat or user exports from our paid tools, our
current employee roster, and spend or contract data.

Work through these stages, in order, and do not skip ahead:

1. SET THE FRAME
Read back to me, in one sentence each: the scope of tools in
review and the decision I said this is for. If I haven't told
you one of these, ask before continuing.

2. RECONCILE
For each tool, match every seat to a name on the roster. List
separately: seats belonging to someone no longer on the
roster, seats with no recent login where the export supports
it, and seats you can't evaluate because the export lacks
usable activity data. Do not guess at usage where the data
doesn't support it.

3. MAP THE OVERLAP
Group tools by capability category. For each group with more
than one tool, ask me who uses each one and why, rather than
assuming overlap is waste. Only flag a group as redundant once
you have an answer, or once I tell you there is no defensible
answer.

4. TRACE DUPLICATE ENTRY
For anything I describe as re-typing the same information into
more than one system, ask what's causing the disconnection: no
API, no owner, a broken integration, or a connection nobody
built. Size it as minutes per occurrence times occurrences per
year times a fully loaded hourly rate.

5. DECIDE, PER FINDING
For shelfware: recommend cancel or downgrade, and say whether
it's executable now or only at renewal. For redundancy: only
recommend a merge after I've confirmed the migration cost and
contract lock-in. For anything unsanctioned or unofficial: ask
why it's being used before recommending sanction, replace, or
shut down, in that order — do not default to shut down.

6. SIZE IT HONESTLY
Report four separate figures: verified recoverable savings,
renewal-period savings, potential exposure needing more
verification, and productivity cost from fragmentation. Never
combine them into one number, and never cite an external
"percentage of SaaS spend wasted" statistic as if it applied to
us.

7. WRITE THE BRIEF
The scope and decision this was run for. The license ledger,
with departed-employee seats called out. The capability map,
with a reason recorded for every tool that stays. Duplicate-
entry findings with their specific cause. A decision for every
finding. The four savings buckets, kept separate. And the
claims you are refusing to make.

The refusals matter as much as the findings. An assistant asked to find software waste will happily produce a confident percentage and a cancellation list, because that’s what the request sounds like it wants. Most of the stages above exist to make it check with you before it decides something is waste, which is the part you’d otherwise have to catch yourself.

Run it once and you’ll have a clean license ledger, a capability map with reasons attached, and a decision for every finding. Run it again at the next renewal cycle and you’re mostly checking whether the decisions held, which is a shorter exercise than the first pass.