The debate about slowing frontier AI is not a reason for your business to stop, or to rush. The right pace is set by your readiness: clear priorities, sound data and controls, and a way to measure value. Boards should move fast where they are ready and fix the gaps first where they are not.
AI leaders and frontier labs have recently called for the development of the most advanced AI to slow down. This article answers the question boards are actually asking: what should that debate mean for the pace of AI adoption in our own business?
Should a business slow down because the AI industry is debating it?
No, but it should listen to why. The debate about the pace of frontier AI is not new. In March 2023, an open letter published by the Future of Life Institute called on AI labs to pause the training of the most powerful systems for at least six months (Future of Life Institute, Pause Giant AI Experiments: An Open Letter, 22 March 2023). The debate has continued since, and it concerns the frontier: the most capable models and the risks they may carry.
Most businesses are not building frontier models. They are adopting tools built on them. The concerns in the debate, about control, misuse and systems that act in unexpected ways, are still relevant, because they show where controls matter. But they are a reason to adopt carefully, not a reason to stop. Stopping has its own cost: competitors keep learning while you do not.
What should set the pace instead?
Your own readiness. An organisation that knows what it wants AI to achieve, has data it can trust, controls it can evidence and a way to measure value can move quickly with confidence. An organisation missing those foundations will find that speed produces pilots that stall, costs without return and risks nobody owns. We set out the foundations in AI readiness starts before AI.
That means the right pace is rarely uniform. It varies by use case and by how much harm a mistake could cause.
| Use case profile | Suggested pace | Condition to move faster |
|---|---|---|
| Internal, low risk, reversible: drafting, research, summaries | Move now, with policy and training | Usage guidance and data rules in place |
| Operational, moderate risk: forecasting, case handling, internal automation | Pilot with a baseline, then scale | Named owner, clean data, measured result |
| Customer-facing or financial decisions | Move deliberately, with human approval | Tested controls, audit trail, clear accountability |
| Autonomous agents acting on your behalf | Start at low autonomy and raise on evidence | Autonomy limits, monitoring and a kill switch |
How does the board avoid both paralysis and recklessness?
By separating the decisions. The board sets the risk appetite once: which kinds of AI use are welcome, which need approval and which are off limits. Management then moves at pace inside those limits, and brings anything outside them back. This removes the two common failures: the organisation that debates every tool at board level and does nothing, and the organisation that lets every team adopt whatever it likes and discovers the risks later.
The board also needs a way to know whether the pace is right. Two signals help. If pilots are reaching production and delivering measured value, the pace is sustainable. If pilots pile up, costs grow and nobody can show the return, the organisation is moving faster than its foundations allow.
What would make a business deliberately slow down?
Some conditions justify slowing a specific use, not the whole programme: a model provider changes terms or behaviour; a regulator issues new guidance that affects the use; an incident shows a control does not work; or the data behind a use proves unreliable. A mature programme has triggers like these written down in advance, so pausing is a routine decision rather than a crisis.
Pausing also needs to be cheap. If switching off a tool would halt a critical process, nobody will make that call quickly. That is why the board should expect fallbacks for anything important: a manual process, a second supplier or a simpler tool that can take over while a problem is fixed. An organisation that can pause safely can afford to move faster, because a mistake does not become a crisis.
What should the board ask management?
- Where are we ready to move faster, and where would speed expose gaps in our data or controls?
- Has the board approved a risk appetite for AI, and does management know where its limits are?
- How many pilots have reached production in the last year, with a measured result?
- What would cause us to pause a specific AI use, and who decides?
- What is the cost to us of waiting another year in our most important use case?
Where to start
Start by finding out where you are ready and where you are not, before deciding how fast to go. Our briefing on whether your business is ready for AI and what boards should check before funding it sets out the six readiness domains and eight board questions. The first step on the AIPEX roadmap is Step 1, Check: the free AIPEX Score check gives a first read in about three minutes. For the full picture, Step 2, Assess, the AI Readiness Assessment (£999 + VAT), scores 30 areas across six domains with a prioritised roadmap.
Frequently asked questions
Should businesses slow down AI adoption?
Not as a blanket rule. The debate about slowing AI concerns the most capable frontier models. For a business adopting AI, the right pace depends on readiness: clear priorities, sound data and controls, and a way to measure value. Move faster where ready and fix gaps where not.
How should a board set the pace of AI adoption?
By approving a risk appetite once, stating which AI uses are welcome, which need approval and which are off limits, then letting management move at pace inside those limits. Pilots reaching production with measured value show the pace is sustainable.
When should a business pause an AI use?
When a provider changes terms or behaviour, a regulator issues guidance that affects the use, an incident shows a control failing, or the data proves unreliable. Write the triggers down in advance. See AI readiness starts before AI →
Correct as of 28 September 2026. This content reflects the law, official guidance and deadlines as published on that date. The EU AI Act and its guidance, standards and deadlines are still being amended and clarified. If you are reading this after 28 September 2026, check for later amendments, European Commission guidance and harmonised standards before relying on it. It is general information, not legal advice, and AIPEX accepts no liability for decisions taken on it without specific professional advice.