Decision engineering for technology leaders
Decisions that survive being questioned.
Give DECIDE the problem in your own words and everything you already have. It builds the record behind the decision — what the evidence supports, what is being assumed, and who authorised what. Permanently.
Three decisions free, for one person, with no card.
The sample decision that ships with the product. Any figure an executive reads can be opened.
Most technology decisions are defended after they are made.
The reasoning lives in a spreadsheet, three email threads and somebody's memory. By the time a board asks why, what was assumed is indistinguishable from what was checked.
of CIOs report breaking even or losing money on their AI investments
Gartner, 2025 · 506 CIOs
the average cost overrun on large IT projects
McKinsey, with the University of Oxford, 2012 · 5,400+ projects
less value delivered than predicted, on those same projects
McKinsey, with the University of Oxford, 2012 · 5,400+ projects
Every figure on this site carries its source and its year, including where that year is 2012. A claim without provenance is not evidence — here is how we hold ourselves to that.
Seven stages. It assembles; you inspect.
You do not fill this in. DECIDE builds the record from what you already have, and you correct it.
Replace, renew or exit the virtualisation platform before general support ends in October 2027.
Unsupported estate carrying production workloads, with renewal leverage falling every quarter.
How does the migration window survive a failed host evacuation on the first weekend?
Move the weightings yourself. Prioritise strategy and the recommendation flips to Option C by 1.7 points — which DECIDE explains rather than hides.
Option B approved — £380,000 against a delegated limit of £500,000.
Detailed delivery stays in your delivery tool. DECIDE holds the conditions the authorisation was given under.
What actually happened, against what was predicted — feeding the next decision rather than a drawer.
AI does the analysis. You own the judgement.
DECIDE reads, structures, compares, challenges and recommends. It cannot approve expenditure, accept residual risk or authorise go-live — and it cannot mark its own output as verified. That is not a policy; it is a constraint in the database.
Every statement carries one of four labels, and the label travels with it into the executive pack. Monochrome on purpose — provenance is a fact, not a status to celebrate.
The incumbent is a spreadsheet and a meeting.
Five tools currently do this job, and every one of them was built for something else.
A number in a cell has no provenance. There is no answer to “where did this come from?”
No line between known and assumed. No adversarial step. No executive layer.
Records what people are doing. Never why the organisation decided to do it.
Risks after somebody accepted them — at £10k to £150k a year.
Every one of them makes you book a demo. Not one publishes a price. Ours is published.
Your next decision is the pilot.
Deciding what to do about VMware? End of general support is October 2027 — that is exactly the shape of decision DECIDE is for.