I have sat in a lot of rooms where an AI initiative was approved with real conviction, real budget, and a roadmap that looked serious. Months later, the same initiative is quietly absent from the conversation. Nobody killed it. It simply stopped moving, and everyone learned to stop asking about it.
Over time I stopped trusting the launch. The launch tells you almost nothing. What an initiative looks like on day one — the deck, the vendor, the ambition — has very little to do with whether it survives. So I built a habit of judging differently. I watch the first ninety days, and I look for a small number of signals that, in my experience, predict the outcome far better than any business case.
Why the first ninety days decide everything
Ninety days is long enough that the novelty wears off and short enough that the original sponsors are still paying attention. It is the window where an initiative either becomes part of how people actually work, or becomes a thing people are politely working around. After ninety days, that pattern is very hard to reverse. Habits have set. The initiative is either load-bearing or it is decoration.
What I am really testing for is not technology. It is whether the organization has built the conditions for the technology to compound. The model was never the hard part.
The five signals I look for
When I want to know whether an AI initiative will ship, I ask five questions. None of them are about the model.
- Does someone own the daily output? Not the project — the output. A name, a person who notices on the day the quality drops, not at the quarterly review.
- Is there a failure protocol? Everyone plans the happy path. The initiatives that survive have already decided what happens the first time the system is confidently wrong.
- Are real users touching it weekly? Not in a pilot cohort that was hand-held into success. In their actual work, on their actual deadlines, without supervision.
- Is the input being refreshed? Every AI system depends on data, context, or instructions that quietly go stale. If nobody owns the refresh cadence, the system degrades on a schedule.
- Has anyone said no to scope? An initiative that is still trying to do everything in week ten has not been forced to confront reality. The ones that ship got narrow on purpose.
If most of these have clear answers with names attached, the initiative is almost certainly going to ship. If they are vague, no roadmap will save it.
An AI strategy without an operational owner is just a forecast. The work is in who shows up on day forty-five.
What the first ninety days should actually produce
I am not looking for scale in the first ninety days. I am looking for evidence of operational gravity — that the initiative has started to pull real work toward it. One workflow that genuinely depends on it. One team that would complain if it disappeared. One measurable signal that someone checks without being reminded. That is worth more than a polished dashboard nobody opens.
The initiatives that go on to scale almost always look modest at ninety days. They have done one thing completely instead of ten things partially. That restraint is not a lack of ambition. It is the discipline that makes ambition survivable.
The discipline underneath
None of this is exotic. It is operational hygiene applied to AI: clear ownership, honest failure handling, real adoption, fresh inputs, and the courage to narrow scope. The reason it is rare is that it is unglamorous, and it has to happen after the exciting part is over.
This is the layer I care most about in my own work and in the businesses I advise — the operating system that turns an AI decision into a durable capability. It is also the work my team at Webspot does with organizations across the region: not just choosing the technology, but building the operational layer that makes the technology stick. The 90-day test is simply how I check, early, whether that layer is being built — or whether we are admiring a plan that has already started to drift.
If you are about to greenlight an AI initiative, do not wait ninety days to apply the test. Ask the five questions now, before launch. If they do not have owners with names, you do not have an initiative yet. You have an intention.
