Deciding in minutes: what changes when the number arrives before the meeting
The cost of a slow decision rarely shows up in a spreadsheet. How the time between doubt and number sets how many decisions a company can make per month — and why that is the real gain of connected AI.
There is a number almost no company measures that explains much of the pace difference between competitors: how long it takes from someone raising a doubt to someone holding the number. Call it time-to-data. In many companies it is two to five business days, and nobody treats that as a problem because it is normal.
Companies rarely lose by deciding wrong. They lose by deciding late and by not deciding — which is what happens when checking is too expensive to be worth it.
What the delay produces
- Decisions by opinion. When data does not arrive in time, the most confident person in the room decides.
- A request queue. The data team becomes a bottleneck, and the small request waits behind the big one.
- Questions that die. The doubt that would have changed the week never becomes a request because it is not worth the hassle.
- Budget stuck in a mistake. A bad campaign keeps running until month close, because that is the only time anyone looks.
A simple calculation, with conservative numbers
Suppose a company spends 100k a month on media and discovers at month close that 15% of the budget sat in a campaign with no return. Finding that on day 3 instead of day 30 does not save 15% of the month: it saves 15% of 27 days. That is 13,500 a month, on the same budget, with no strategy change.
The figure is illustrative; the mechanic is not. Every postponed decision has a daily cost, and the daily cost is invisible precisely because it never appears as a line in a report.
Where the time hides
| Step | Typical time | Where it ends up |
|---|---|---|
| Framing the request | Minutes | Unchanged |
| Waiting in the data team’s queue | 1 to 3 days | Disappears |
| Writing and checking the query | Hours | Becomes seconds |
| Coming back with the second question | Another full cycle | Becomes the next sentence |
| Deciding | Minutes | Unchanged — and the only one that should cost |
AI does not speed up the decision. It erases everything that sat between the doubt and the decision.
The risk of deciding fast
Deciding faster on wrong data is worse than deciding slowly. So the first month should be spent checking: always ask for the filter used, compare with a number you already trust, and treat a discrepancy as information rather than failure.
Once the base earns trust, the rule gets simple: reversible decisions get made fast, irreversible ones still go through people. That distinction is worth far more than any uniform approval process.