Treating every system as equally dangerous is not caution. It is a failure of discernment, and it produces exactly the recklessness it was meant to prevent

There is a widely held assumption that ethics and speed are opposed. On this view, governance is a brake: necessary perhaps, regrettable certainly, and always paid for in delay. Organisations that move fast are understood to have accepted some risk. Organisations that govern carefully are understood to have accepted some slowness.

The assumption is wrong, and its wrongness is demonstrable rather than merely hopeful.

What actually slows organisations is not governance but bad governance: undifferentiated review that treats a scheduling tool and a medical triage system as equivalent, approval processes with no defined criteria, committees that meet monthly regardless of urgency, and requirements so burdensome that the reliable way to ship anything is to avoid the process entirely.

That last outcome is the important one. Governance that is too heavy does not produce caution. It produces evasion. Teams find workarounds, adopt tools without registering them, and describe systems in whatever language avoids the highest tier of scrutiny. The organisation ends up with less oversight than if it had asked for less, and the oversight it does have is applied to the projects honest enough to volunteer for it.

Proportionality Is a Moral Principle

The remedy is not less governance. It is governance that distinguishes.

Most decisions carry little risk. A tool that formats documents, drafts an internal summary, or suggests meeting times does not require ethics review. Subjecting it to one wastes the reviewer’s attention, which is the scarcest resource in the whole system, and teaches everyone that the process is theatre.

Some decisions carry real risk. A system that screens job applicants, allocates clinical attention, assesses creditworthiness, flags a family for investigation, or influences access to a public service will change lives, and often the lives of people with the least capacity to contest it.

A governance regime that cannot tell these apart is not rigorous. It is undiscerning, and the distinction matters because discernment is the whole task.

The Catholic moral tradition has long held that the gravity of scrutiny should match the gravity of what is at stake. Prudence is not maximum caution applied uniformly. It is the capacity to judge what a particular situation requires. Applied to technology governance, this yields a practical shape:

Low risk: proceed. Pre-approved tools, clear guidance on acceptable use, no case-by-case review. Speed here is not a compromise, it is correct.

Moderate risk: review. Validation before deployment, privacy and security assessment, defined human oversight in operation. Days, not months.

High risk: controlled. Sensitive data or high-impact decisions about persons. Enhanced review, external audit, named accountability, tested procedures for suspension and remedy.

The tiers do their work by protecting the top from the bottom. When trivial cases do not consume review capacity, the serious cases get the attention they actually need.

Why Early Governance Is Faster

The deeper efficiency argument concerns timing rather than volume.

Governance placed at the end of development is friction by construction. The system is built, the budget is spent, the launch is announced, and the reviewer is asked to bless it. Any objection now costs a rebuild. The pressure to approve is enormous, and it does not require anyone to behave badly for the outcome to be predictable. Governance in that position can only veto, which means in practice it can only approve.

Governance placed at the beginning does something different. It shapes the design. Questions about what data is necessary, who is affected, how a decision will be explained, and what happens when the system is wrong are cheap to answer in a design document and enormously expensive to answer after deployment.

Engineers understand this intuitively from every other domain. Nobody thinks structural requirements should be considered after a building is finished. Nobody regards fire safety as a constraint on architectural ambition. These requirements shape the design from the outset, and the result is a building that is both safe and buildable.

Ethical requirements behave identically. Identified early, they shape a design elegantly. Identified late, they require demolition. The organisation that governs at the beginning is not trading speed for safety. It is avoiding the rework that late discovery guarantees.

The Trust Compound

There is a further effect, slower and larger.

Systems people do not trust are not used, or are used defensively, with workarounds that undo their benefit. Clinicians ignore alerts they have learned are unreliable. Caseworkers override recommendations they cannot explain to the person in front of them. Citizens disengage from services that made a decision about them without explanation.

Trust is the condition on which technology delivers value at all. And trust is not produced by marketing. It is produced by a track record of systems that behave as described, fail visibly rather than silently, and can be contested by the people they affect.

An organisation with that record can deploy new systems quickly because the presumption is in its favour. An organisation without it faces resistance at every deployment, regardless of the merits of the particular case. Governance is therefore an investment in future speed, and the return arrives late, which is why it is chronically underfunded.

Responsible AI Should Be the Easier Path

The design goal that follows from all this is a specific one: within an organisation, the responsible way to use AI should be easier than the risky way.

This is achievable, and it is mostly a matter of making the safe path convenient rather than making the unsafe path forbidden.

Pre-approved tools available immediately, so nobody needs to reach for something unvetted. Clear written guidance a person can consult in five minutes rather than a policy nobody has read. Templates that build in the requirements, so compliance is a by-product of using the standard approach. Review turnaround measured in days, so escalating a genuine question does not endanger a deadline.

When the compliant path is the path of least resistance, compliance stops depending on individual virtue. When it is not, an organisation is relying on every employee choosing the harder option every time, which is not a governance strategy but a hope.

Neither Reckless nor Paralysed

Two failures sit on either side of this, and both are real.

Innovation without governance produces harm that is discovered by the people it lands on. Governance without proportion produces paralysis, evasion, and eventually a culture that regards ethics as an obstacle to be managed rather than a discipline to be practised.

The organisations that will handle this well are not the ones that care most or move fastest. They are the ones that can tell the difference between a system that needs three months of scrutiny and one that needs none, and that ask the hard questions early enough for the answers to matter.

That is not a compromise between ethics and innovation. It is what both look like when they are done properly.