A thought paper for the business that upgraded the weapon and still can’t see the target.
— THE WEAPON GOT SHARPER. THE JUNGLE DID NOT
The best hunters in the world cannot see in the dark. Decades of experience, deep knowledge of the terrain, strong instinct — none of it grants night vision. It’s not a failure of preparation. It’s the nature of the jungle.
So someone hands the hunter a better weapon.
Sharper scope. Faster reload. A readout on the stock that shows shot accuracy in real time — ninety-four percent, green light, performing beautifully.
You are still in the dark. The weapon just got better at doing the thing that was never the problem.
This is what deploying AI can feel like. Not a failure. An upgrade to the wrong part of the problem.
— HAVING THE WEAPON IS NOT THE SAME AS BECOMING THE HUNTER
There’s a belief sitting underneath most AI deployments, rarely said out loud: that having it will change things. That your business becomes something different simply by acquiring the capability — the same way acquiring a website once meant you were “online,” and acquiring a CRM once meant you were “data-driven.”
The hunter who buys the sharper weapon does not become a better hunter the day it arrives. You become someone carrying a better weapon. Whether you hunt better than before is a separate question, answered later, in the jungle — not at the point of purchase.
Equipped means you have the weapon. Prepared means you know whether carrying it is changing anything you actually hunt for.
And the change is never the weapon’s to deliver alone. It shows up when you notice which signal is moving in the right direction, and make the deliberate call to strengthen exactly that — more of what’s working, less of what isn’t. That call is yours. The AI can point at the signal. It cannot make the decision for you, and it will not make it while you wait.
You’ll find out whether you were equipped or prepared — after the money is spent.
— THE DASHBOARD THAT SHOWS THE WEAPON, NOT THE JUNGLE
If you’ve deployed AI, you likely have a vendor dashboard. Model accuracy, prediction volumes, system uptime. Green across the board. The model is performing.
You hired carefully once too — good people, the right roles, real effort spent. That was work being done. It was never the same question as whether the strategy was working. The model is the same kind of work. Deploying it, watching it perform — that’s effort, and it’s real. It has never been the same question as whether the business is better off.
What the dashboard doesn’t show is whether any of that performance is changing your business. Whether the decision the AI was supposed to improve is actually being made better. Whether the shot is landing where it needed to.
Three people, three different deployments, found this out the hard way. None of them had a model problem.
A CFO, after a board question that used to end in a story rather than a number, built the measure that let this year’s answer be a number instead. Decision quality up. Defaults down. The board asked two follow-up questions instead of twenty. What stays with him isn’t the win — it’s the two years before it, spent describing what the system was doing instead of knowing whether it worked.
A supply chain director watched a demand model run green on accuracy for six months while she still couldn’t tell her CEO whether stockouts had actually improved — because nobody had defined “improved” before go-live. What she carries from that deployment isn’t the budget line. It’s the memory of empty shelves on lines the model was supposed to be covering, while the accuracy metric never blinked.
A chief experience officer discovered — eight months in — that customers who followed their AI’s recommendations ended up buying less than customers who ignored it entirely. Confident, wrong, and invisible, for eight months. The click-through rate never told her. Nothing was watching the actual outcome, so nothing caught it drifting.
Read the three again. If one of them already sounds like you, that isn’t a coincidence.
Three green dashboards. Three businesses that still couldn’t answer the only question that mattered.
— THE COMMON THREAD — AND IT ISN’T THE AI
None of those three outcomes came from a worse model. The models performed exactly as advertised in every case — the work got done, competently, on schedule. What was missing was whether any of that work changed the business. Not intelligence. Sight — a way to watch the business outcome independently of whether the model felt like it was working.
A vendor dashboard tells you the weapon is firing correctly. It cannot tell you whether you’re still hunting the right thing.
And a signal still has to be read, not just glanced at. A line drifting for three straight weeks and a line that dipped once look identical on a dashboard built for a glance. You have to actually notice which one it is.
That’s a different instrument. It isn’t a better model, and it isn’t more AI. It’s the thing that watches whether your outcome is arriving at all — regardless of how green the model’s own report card looks.
Most don’t have this. You have performance. You don’t have proof.
— THE QUESTION WORTH SITTING WITH
Your board will ask: is the AI investment working?
Most answer with a story. The dashboard is green. The team is enthusiastic. The model is accurate.
If your honest answer includes “the model just needs more time,” or “the vendor says it’s on track,” or “we’re still in the tuning phase” — you already know. You are describing the weapon. You have stopped asking about the target.
And the ammunition isn’t free while that gets sorted out. The subscription renews whether the shot lands or not. The token bill arrives whether the business moved or didn’t. Every month spent unsure is a month spent paying for a weapon whose only certainty is the invoice.
The harder admission is rarely that the AI isn’t working. It’s that some part of you already suspected it months ago, and found it easier to keep glancing at the model’s green light than to ask the question that might come back red.
A sharper weapon in a dark jungle is still a hunter who can’t see. The upgrade was real. It just wasn’t the thing that was missing.
And seeing the gap is not the same as closing it. The signal shows the drift. Closing it is still a separate day’s work — and it’s yours, not the model’s.
The hunters who get there fastest are rarely the ones who moved first. They’re the ones who didn’t have to retrace the same ground twice.
You already know which of the three voices in this paper sounds closest to yours. That is the discomfort worth sitting with — not the paper.
If that recognition creates even a moment of discomfort, it’s worth a conversation.
The Prepared Business
We help organisations see whether their AI investments are changing the business — not just performing well on the vendor’s terms.
Leave a Reply