A clear decision on every AI project you’ve funded.
The projects are under way and the money is committed. We read each one against the knowledge it depends on, and the financial case it was funded on, then give it a verdict: stop, continue, or reclassify.
Some of these cancellations are coming either way. The review makes them deliberate, and frees the budget before more is spent on the ones that won’t pay back.
From late 2026, Australia’s Privacy Act introduces transparency obligations for automated decisions that affect people. The AI you can’t account for is the AI you can’t defend.
Every project, on one picture.
Two axes: the return each project is expected to deliver, and whether the knowledge it needs is written down where AI can use it. The verdict follows from both.
Illustrative. Two projects in Context Debt look right, but the knowledge they need isn’t written down (Stop). Three are on track (Continue). One is sound but mislabelled (Reclassify).
The cost arrives before the value.
The market has moved from “can we build it?” to “can we afford to operate it?”. The CFO sees the running cost rise before the return shows up. Freed time is not money in the bank: spend rises, people feel productive, and none of it shows up on the bottom line.
The review catches it before the running costs compound.
AI doesn’t remove scarcity from knowledge work. It relocates it.
One decision on every project.
A portfolio map with every initiative on it, each with a verdict and the reasoning behind it. The verdict is one of three.
The knowledge it needs isn’t written down where AI can use it, and no path to fix that in time. Stopping it frees the budget and avoids the sunk cost.
On track, and measured for what it actually does. Leave it alone. Naming what works matters as much as naming what doesn’t. Some projects just do a simple job well.
Sound, but mislabelled: booked as growth when it really just holds you level, or sold as a cost cut when it’s really cutting risk. Reclassify it, and it stops looking like a failure when it isn’t one.
“We’re stopping the copilot. It can’t handle the cases that matter most, because the judgement they need was never written down. We’ll move the budget, and write that knowledge down first.”
Illustrative, from a simulated engagement.
Two questions on every project.
First, whether the knowledge it needs is written down where AI can use it. Where it matters, we don’t just say there’s a gap, we prove it: we take one of its answers and check it against a source you can rely on. The vendor owns the model. You own the verification. Second, what it’s really funded for.
Funded for growth
Someone owns a lift in the numbers, and the spend is justified by growth.
Funded for savings
Someone owns a reduction the spend pays for.
Funded to hold the line
Neither growth nor a cut. It keeps you level. The right home for a lot of AI spend, and the hardest to fund.
The failure isn’t funding a project to hold the line. It’s funding it while calling it growth. The review names the mislabelled ones, so each is measured for what it is.
A defined piece of work, not an open engagement.
Fixed scope, a set shape, a named deliverable. The number of projects sets the length, so the exact size is agreed up front.
We go through every funded initiative, test the most important one against where its answers really come from, then give you a portfolio map with a verdict and the reasoning on each.
About two weeks for a typical portfolio. Agreed up front against the number of initiatives in flight.
A portfolio map, a verdict on every initiative, and the reasoning behind each one.
It tells you where each project sits and what the gap is. It doesn’t close the gap. That is separate work.
It names the mislabelled projects. It doesn’t restructure your capital allocation or advise on accounting. That is your call, with your accountant.
It reads a portfolio at one point in time. It isn’t ongoing governance, though it shows where that discipline is missing.
The limits are what make the rest credible.
Maps the stack first. The review runs best against it, and can run first on its own when spend is already in flight.
You are here. A verdict on every project you’ve funded.
When one decision needs deeper work, this takes it on, and codifies the knowledge AI couldn’t use.
Start a conversation.
Have AI spend in flight and the running costs arriving before the returns? Tell us what’s on the roadmap, and we’ll look at which of your investments will pay back. No deck, no pitch.
Start the conversation→Every engagement is led hands-on, and we take on a small number at a time.