Your software portfolio grows faster than your view of it. Here is the evidence a CFO wants.

There is a meeting most finance chiefs know. Someone suggests "looking at the software costs". Everyone nods. An analyst spends three weeks gathering contracts, invoices and license counts into a spreadsheet. The spreadsheet is impressive for two months. Then it is history.

The problem is not willingness. The problem is that the software portfolio in a mid-sized company is a moving target: every department buys on its own, AI tools arrive at a pace no procurement process was built for, and every license renews on its own date, usually automatically, usually without anyone having looked at usage first.

Sprawl is not a scandal. It is the natural state of any organisation that moves quickly. The interesting question is not "how did we avoid it?" but "how do we see it, continuously, without three weeks of archaeology every time?"

Three questions a spreadsheet cannot answer

A static overview can tell you what you pay. It cannot answer the three questions a real rationalisation decision requires:

Do we even use it? License count is not usage. Usage lives in the login signals, and they have to be held against the contract, continuously, not once a year.

Does it overlap with something we already have? That requires tools to be connected to the capabilities they support. Three systems on the same capability are not necessarily a mistake, but they are always a question that deserves an answer.

When do we have to decide? The best rationalisation case in the world is worthless if the renewal went through last week. The decision window is the date, not the report.

Notice the pattern: all three questions are relationship questions. Tool to usage, tool to capability, tool to calendar. That is why the spreadsheet loses: it can hold the rows, but not the relationships.

From overview to evidence

In Atlas the software portfolio lives in the same living model as capabilities, owners, contracts and the AI registry. That changes what a cost review is:

The utilization picture appears on its own. Usage signals settle as a layer over the portfolio: here are the tools with many licenses and few logins. Not as an accusation, but as a list of questions.

Overlap becomes visible through the capability map. Because every system is connected to what it supports, the Duplicate Hunter agent can point to candidates: three tools, one job. The human decides whether it is redundancy or a deliberate choice.

The renewal radar turns the calendar into a control. "You renew three tools on the same capability within 60 days" is the kind of sentence that turns an abstract sprawl problem into a concrete decision with a date and a budget frame.

The case writes its own draft. When the evidence lives in the model, the rationalisation case can be generated: affected systems, usage picture, overlap, notice windows, dependencies and risks. People qualify and decide. Nobody starts over in a blank document.

The side benefit: the AI portfolio is the same exercise

Here is the observation that ties the cost agenda to the governance agenda: shadow AI is found with exactly the same signals as unused SaaS. Login patterns, expense entries and new integrations reveal both the tool nobody uses and the AI assistant nobody has registered.

That is why cost evidence and the AI registry in Atlas are one model and not two products. The CFO and the CAIO look at their own layer of the same landscape. And the CIO, who has to stand on both agendas at once, finally gets one place to stand. Relevant context from Gartner's CIO agenda: AI spend grows around 35 percent year over year, while only about a third of CIOs can consistently document financial outcomes. The evidence discipline is not decoration. It is the currency of the budget conversation in 2026.

Start with one question

You do not need a programme to get going. Pick the question that hurts the most:

• Which three renewals are closest, and what do we really know about usage?
• Where do we have several tools on the same capability?
• Which AI tools are we paying for that nobody has registered anywhere?

One honest answer usually funds the rest of the journey, so it holds together.

Try Atlas at atlas.spekir.com, or talk to us about EA.

Sources: Gartner CIO Agenda 2026: AI spend around plus 35 percent year over year; about a third of CIOs consistently prove financial outcomes. Industry observations about SaaS sprawl are deliberately rendered qualitatively without percentage figures.