When “not yet” is the right AI decision
In AI discussions, “not yet” can sound negative. It can sound like delay, caution, or a lack of ambition.
In regulated organisations, it can be something more useful: a decision to sequence the work properly.
The point is not to wait for perfect data, perfect governance, or perfect confidence. Those conditions rarely arrive. The point is to avoid treating every AI idea as if it carries the same risk, requires the same controls, or is ready for the same level of reliance.
Some AI use can start now. Some should be narrowed. Some should wait until the organisation has stronger foundations in place.
That is not an anti-AI position. It is a practical one.
Start where AI can safely help now
Generative AI can already support useful, lower-risk work in many teams. It can help draft internal communications, summarise non-sensitive material, improve templates, prepare meeting notes, structure options, support learning, and reduce blank-page admin.
Used carefully, these activities can save time and build confidence without pretending the organisation is ready for high-reliance AI decision support.
That distinction matters.
A team may be ready to use AI to improve a first draft, but not ready to use it to shape a regulated decision. It may be ready to summarise public or internal guidance, but not ready to process sensitive operational data. It may be ready to help staff explore ideas, but not ready to automate workflow steps that affect service delivery, assurance, or accountability.
The weak decision is not “not yet”. The weak decision is pretending all of these uses are the same.
Move beyond the yes-or-no argument
This is where many organisations get stuck. One group wants to move quickly, because the tools are already available and staff are experimenting. Another group wants to slow everything down, because the data is messy, governance is unclear, or the risks are not fully understood.
Both concerns can be valid. The mistake is turning the discussion into a simple yes-or-no argument.
A more useful question is: what kind of AI use is sensible now, and what needs to be true before more ambitious use can proceed?
There are good reasons to slow down larger AI plans. If the source of truth is unclear, if ownership is weak, if no one knows who checks the output, if workflows depend on undocumented judgement, or if the organisation cannot explain what happens when something goes wrong, then scaling AI too quickly may simply scale the confusion.
That does not mean the organisation should do nothing. It means the immediate work should be different.
Use “not yet” to define the next useful step
The right next step may be to clarify ownership, improve data definitions, map the real workflow, agree approved-tool boundaries, train staff on safe use, or identify lower-risk use cases that can build confidence.
It may mean using generative AI for drafting and support while keeping decision-making, approval, and assurance firmly with people.
This is where “not yet” becomes useful. It creates space to separate immediate opportunity from longer-term readiness.
A mature AI decision is not always yes or no. Sometimes it is: use it here, narrow it there, prepare this area next, and do not scale that use case until the foundations are clearer.
That kind of decision may feel less exciting than launching a broad AI programme. But in regulated environments, it is often more valuable.
Ambition still needs sequencing
The goal is not to block sensible experimentation. It is to stop ambition running ahead of the organisation’s ability to govern, explain, and sustain what it is doing.
“Not yet” should not mean fear. It should not mean drift. It should not mean another year of waiting for perfect conditions.
Used properly, it means something much more practical: start where AI can safely help now, while doing the work needed for more ambitious use later.
That is often the better AI decision.