Eighteen months. That is the median time I have observed between an SMB owner signing the initial AI platform contract and the same platform being quietly deprecated — sometimes by the same team that championed it, sometimes by a successor hired specifically to clean up what went wrong.
In roughly nine out of ten of those deprecations, the failure is not technical. The software works. The API calls return. The model produces output. The pilot hits its stated KPIs. And then the thing just stops being used, drifts quietly into the graveyard of tools nobody opens anymore, and the organisation returns to the workflow it had before — sometimes with a residual line item in the budget and a quietly bruised leadership team.
The gap between the technology working and the business being transformed is the single most underpriced variable in AI consulting right now. I call it the Readiness Gap, and I want to explain what is in it, because closing it is most of the work.
The question is almost never "does this technology work." It is "does this organisation, right now, have the conditions under which any AI initiative would work."
Five dimensions, not one
When I run a readiness assessment for a client, I am not looking at one number. I am mapping five dimensions, each of which can individually tank an initiative regardless of how the other four score. They are:
- Data foundation. Where does the data live, how trustworthy is it, and is it structured in a way that an AI system can actually consume?
- Strategic clarity. Is there a specific, named problem this is meant to solve — or is the strategy just "do something with AI"?
- Technical infrastructure. Can your current stack support what you are about to bolt onto it, and do you have the security posture to do so without creating a new risk surface?
- People capability. Do the humans who will actually use this thing have the skills, time, and — critically — the willingness to change how they work?
- Governance. Who owns the initiative, who decides when it is or is not working, and what happens when it produces an output nobody anticipated?
Here is what a real readiness matrix looks like for a recent client — a 40-person distribution business that came to me convinced they were ready to buy an inventory-forecasting AI:
| Dimension | Score | Brief |
|---|---|---|
| Data foundation | 🔴 Red | Inventory data lives across three spreadsheets and one desktop SQL instance. Three of four SKU categories use inconsistent unit conventions. |
| Strategic clarity | 🟡 Yellow | "Reduce stockouts" is the stated goal but no baseline has been measured, no target set, and no owner named beyond the founder. |
| Tech infrastructure | 🟢 Green | Modern ERP, clean cloud footprint, no legacy blockers. |
| People capability | 🔴 Red | Ops lead has never worked with a predictive system. No process for handling model outputs that contradict intuition. |
| Governance | N/A | No framework exists yet. Not a red flag — a blank page. Addressable as the first workstream. |
Two of the five dimensions are red. The tech is fine. If this client had bought the forecasting platform their competitor had just recommended, they would have spent something in the range of $60–90k on a tool their data could not feed and their people could not operate. The pilot would have worked. The platform would have been abandoned. The founder would have concluded AI was overhyped.
What we did instead was three months of unglamorous work on data and change management, then a targeted automation on a smaller slice of the problem. The headline outcome: one actual stockout category reduced by 31% in the first quarter after implementation, and — the part that matters more — the ops lead now trusts the system enough to let it make decisions she would previously have intervened on.
The technology was never the bottleneck. The bottleneck was the conditions the technology was being asked to work in.
Why the gap is getting wider
Three forces are actively widening the Readiness Gap right now, and SMB owners are disproportionately exposed to all three.
1 · Vendor urgency is at an all-time high.
Every AI platform vendor is compensated on speed of deployment. Their discovery calls are engineered to minimise the time between "interesting conversation" and "signed contract." A readiness assessment that concludes "you are not ready yet" is a vendor's worst outcome, so vendors do not run them.
2 · "AI-powered" has become a meaningless label.
The market is saturated with tools that bolt a large language model onto a workflow that did not need one. These tools work for demo purposes and then create more friction in production than they remove. Telling them apart from genuinely useful systems requires domain judgement most SMB buyers have not yet had time to develop.
3 · Leadership is under pressure that is new and specific.
SMB owners in 2026 are being asked, by boards and peers and investors, to have an AI answer. The pressure to have one is, often, exactly what produces the wrong one. The most expensive mistakes I have seen in the last eighteen months were decisions made under this specific pressure — not due diligence failures, but urgency failures.
The fastest way to make a bad AI decision is to feel you are behind. The second fastest is to hire a consultant who agrees with you.
What to do instead
If you are reading this and recognising your own organisation in the description, the corrective is not dramatic. It is three steps, none of them expensive, executed in order:
- Score yourself honestly across the five dimensions. The assessment is free and takes 5 minutes. Take it here. It will not sell you anything. It will tell you where you are.
- Fix the reds before you buy anything. Red dimensions are not gating conditions to be bulldozed — they are predictors of failure. Invest the comparatively small amount of money it takes to move them to yellow before you sign a platform contract.
- Choose narrow first. Pick the smallest possible slice of a real problem, solve it well, and use the experience to build the internal muscle that will make the next slice easier. Transformation compounds; it does not arrive.
If you are already past the buying decision and the initiative is showing early signs of drift — low adoption, mysterious technical debt, quiet complaints from the people using it — the same three steps apply. Stop, score, fix the reds. The cost of pausing a stalled initiative is almost always lower than the cost of continuing one.
And if you want someone who will tell you honestly which column you are in: that is what the first conversation is for. Book a discovery call here. If we are not the right fit, I will say so.
— Tom