The approval nobody questioned before the invoice landed.

An early Samplify customer ran the numbers last month and sent them across. Six people with access to the system. Fifty-five questions asked in thirty days. No mandate. No target. They just kept using it because requests kept coming in.

When he walked through the month, the pattern was simple. Tools people claimed they urgently needed. Renewals about to roll over automatically. Each time, before approving or pushing back, the team fired off a question first.

The total came to three hundred hours of research time recovered in a single month. At least two purchase requests blocked. One of those alone would have cost around two million pounds.

The cost of asking nothing

Enterprise software waste is widely discussed. The mechanism gets less attention. It is not usually recklessness. Companies approving software they already own are not being careless. They are working without a fast, confident answer to the most basic question in software governance: do we already own something that does this?

That question comes up every time a request lands. Someone in marketing needs an e-signature tool. Someone in operations wants a new project management platform. The person reviewing the request has, on average, no reliable way to answer it quickly. So friction fatigue sets in, the request gets approved, and the money moves. Or someone says no on gut instinct and the requester finds another route anyway. Either way, the outcome was not chosen. It happened by default.

Why the check cannot be done

The gap is structural. ChatGPT can reason about software categories, but it hallucinates on feature data fifteen to thirty percent of the time. That is not an acceptable confidence level for a procurement decision that needs to be defensible. SAM tools track entitlements, not capabilities. They know you own a licence. They cannot tell you whether two tools in your estate overlap at the feature level, or whether the tool someone is requesting covers the same ground as something bought eighteen months ago.

Procurement systems sit in the workflow but carry no product intelligence. You can try to remember a vendor briefing from six months back. The answer rarely surfaces that way.

Governance at the point of decision

Most software governance happens in retrospect. Annual audits, rationalisation reviews, spend analyses run by finance after the year closes. These are necessary. They are not sufficient. By the time the audit runs, the duplicate licence has already renewed. The tool nobody uses has been embedded in three workflows. The spend has already happened.

The best time to stop unnecessary spend is before the invoice lands. Not six months after.

What those fifty-five queries represent is something different. Governing at the actual point of decision. Each time a request came in, an answer was available in under twenty minutes. Feature-level analysis, cross-referenced against everything the company already owned. Not from memory. Not from a spreadsheet. From something that knows what each piece of software actually does, held alongside the full context of what the organisation has already licensed.

What two million pounds actually proves

The number that stays with you from that conversation is not the three hundred hours. It is the simplicity of the mechanism. A question was asked. An answer came back. A decision was made differently than it would have been otherwise.

No transformation programme. No replacement of existing SAM tools or procurement systems. One layer added above them, so the work already done on cataloguing the estate could influence decisions as they happened, rather than after they had already gone wrong.

That is what the decision layer workflow is built for. Not as a reporting dashboard nobody checks. Not as a background audit nobody reads. As something that sits directly in the approval flow, so the right answer is available at the right moment, sourced and auditable, in the workflow the team already uses.

If you want to see what this looks like in practice, a thirty-day proof of value requires no integrations and no upload. Value is visible in days, not months.