Standing still is also a decision.
Two businesses start in the same place and drift years apart without a single dramatic event. Neither makes a bad call. Only one makes a call at all.
Picture two mid-sized contractors in the same city, in the same year. Same headcount, same margins, same Microsoft 365 tenancy, same pressure from the board to “have an answer on AI.” Call them A and B.
Business A decides that the safe move is to wait. Not to ban anything, just to make no decision. Business B decides that the work its people already do with AI deserves a governed place, and puts a foundation under it: evidence, permissions, approval, a record.
Nothing dramatic happens to either of them. That’s the point.
Year one: indistinguishable
In year one, you couldn’t tell them apart. Both quote work, win some, deliver it. In A, staff quietly use personal AI accounts; drafts improve; nobody asks where the words came from. In B, the same staff do the same work, but the drafts are grounded in the company’s own evidence, and every answer they rely on is one they’re entitled to see.
A’s leadership reads B’s position as bureaucracy. B’s leadership reads A’s as luck that hasn’t run out.
Year three: the quiet compounding
By year three the differences are structural, though still invisible from the street. B’s tenders reuse the ratified judgement of its best estimators, including one who retired last autumn. Its month-end reconciliations close in hours because the ledger only ever grew; nobody argues with a record both parties watched being written. Its answers to a client’s “where did this figure come from?” take minutes.
A is not in crisis. It’s just slower in a hundred small ways. The quoting spreadsheet has more versions. The person who understood wet-hire pricing has resigned, and the new estimator is reconstructing the logic from old PDFs. An auditor’s routine question takes three weeks and produces a shrug.
Capability compounds quietly. So does entropy.
Year five: years apart
By year five, B bids faster than A with fewer people on each tender, defends its documents without a scramble, and adopts each new generation of AI models the week they arrive, because the gate was always the permanent part, and the model behind it was always replaceable. A, meanwhile, is finally commissioning a review into “AI readiness,” and discovering that its real problem was never AI. It was two decades of ungoverned work, now moving at machine speed.
No dramatic event separated them. One decision did, and one of them never noticed making it.
Responsible adoption is the safe move. Doing nothing only feels like one.
The honest version of “wait and see” is “let ungoverned work keep accumulating, faster than before.” The honest version of caution is a foundation: permissions that hold, evidence that attaches itself, approvals that leave a trace, and AI brought in only where reasoning genuinely adds value. That is what a governed operating model is for, and it is available before, not after, you scale AI on top of it.