A decision is only as good as what nobody checked.
Most technology decisions are reasonable at the time and indefensible six months later. The analysis was sound; it is the record of it that has gone. What survives is a recommendation, a number, and no way to tell which parts were established and which were assumed by someone who has since left.
Facts are not assumptions
Every material claim in a DECIDE decision is classified as a fact, an assumption, an unknown or a constraint, and carries an owner and a source. An assumption stays visibly an assumption until somebody verifies it against something. Nothing is promoted by being repeated, by being in a slide, or by being typed confidently.
That includes anything the AI produces. A suggestion is recorded as unverified until a named person checks it — enforced by a constraint in the database rather than by a setting, so no future feature can quietly turn it off.
A challenge from the author is not a challenge
The person who wrote the plan cannot be the person who tests it. In DECIDE that is not guidance — the database refuses to let the same person hold both roles on one decision. It is the difference between a challenge stage that means something and one that is performed.
Where two of your own documents disagree, DECIDE says so rather than picking one. A conflict between sources is among the most valuable things in a folder, and the usual fate of it is to be resolved silently by whoever read the second document last.
The method
You do not fill this in. DECIDE builds it from what you already have, and you inspect and correct it.
What exactly must be decided, and what happens if nothing changes.
Genuine alternatives, developed to the same depth. Including doing nothing, priced.
How does this fail? Not “does it work?” — that question has never found a weakness.
Compare under weightings you can move, and see whether the answer survives them.
A named authority approves against a delegated limit. The record cannot be edited.
Outcome milestones and go/no-go conditions. Detailed delivery stays in your delivery tool.
What actually happened, against what was predicted.
If you are deciding what to do about VMware, this is the shape of decision DECIDE is for.
VMware 8.x reaches end of general support in October 2027. That is a public, dated deadline that forces a choice with a large number attached, several credible alternatives, and a board that will ask how the answer was reached.
of VMware users plan to reduce their usage by 2028
Virtified, 2026 · 450 users across 14 countries
of workloads are expected to move off VMware by 2028
Gartner, 2026
AT&T alleged in a New York Supreme Court filing that Broadcom proposed a 1,050% increase. We describe that as an allegation because that is what it is — a claim in a court filing, not a finding. Elsewhere you will see “70% of VMware customers will have left by 2028”; that figure comes from a vendor blog and contradicts Gartner's own 35%, so we do not use it.
The cost of getting this wrong is already well documented.
of managers’ time is spent making decisions
McKinsey, 2019
of transformations fall short of their objectives
BCG, 2020