Part 2 of a 3-part series on governing AI at machine speed
In Part 1, I outlined a growing challenge facing organisations adopting AI.
Not a lack of governance…
…but a mismatch between how governance operates and the nature of what it is trying to control.
Traditional approaches rely on static controls applied periodically.
AI operates continuously, evolves rapidly, and introduces new scenarios that cannot always be anticipated in advance.
If governance is to keep pace, it needs to function differently.
Rather than focusing purely on structures, policies and documentation, organisations need to ensure they have the underlying capabilities required to govern effectively in a dynamic environment.
From what I’m seeing in practice, five capabilities are becoming increasingly important.
1. Principle-Based Governance
Most governance models rely heavily on prescriptive rules.
While this works in predictable environments, it becomes difficult to sustain when systems evolve quickly and new scenarios emerge daily.
AI introduces exactly that challenge.
Rather than attempting to define every possible use case, organisations need to establish clear principles, boundaries and acceptable risk thresholds that guide decision-making.
This creates a more flexible foundation, allowing people to act with confidence even when explicit rules do not exist.
The question is not whether rules are in place…
but whether decisions can still be made when those rules don’t yet exist.
2. Human-in-the-Loop Capability
“Human-in-the-loop” is often referenced as a control.
In reality, it is frequently passive.
As AI becomes more embedded in operations, individuals are expected to interpret outputs, assess risk and make decisions in real time.
That requires more than oversight.
It requires capability.
People need:
- clarity of accountability
- confidence to act
- and an understanding of when something doesn’t look right
Without this, accountability remains theoretical.
Governance ultimately depends on people who can make decisions under pressure — not just follow process.
3. Continuous Validation
Governance is often tested on paper.
Policies are reviewed. Controls are checked. Compliance is assessed.
But real breakdowns rarely happen in controlled conditions.
They happen when:
- information is incomplete
- time is limited
- and decisions need to be made quickly
AI amplifies this.
Systems can drift, outputs can degrade, and edge cases can emerge without warning.
Periodic review is no longer enough.
Organisations need mechanisms to continuously test how governance performs in practice.
This is where simulation and scenario-based approaches become particularly valuable.
The real test of governance is not whether it exists…
but whether it holds up when conditions change.
4. Cultural Operating System
Governance is often defined through documentation.
In reality, it is expressed through behaviour.
Culture determines:
- how people respond to ambiguity
- whether issues are escalated
- and how decisions are made when guidance is unclear
In an AI-driven environment, this becomes even more important.
Many risks are not immediately visible.
They rely on individuals recognising when something feels wrong and taking action.
Organisations with strong governance cultures tend to respond faster and more effectively.
Those relying purely on formal controls often struggle to move beyond discussion.
Policies don’t act — people do.
5. AI-Augmented Governance
The final capability is the one most organisations are only beginning to explore.
If AI is operating at scale and speed, governance needs support to keep up.
This is where AI itself becomes part of the solution.
Rather than relying solely on human observation, organisations can use AI to:
- monitor behaviour and outputs in real time
- identify patterns and anomalies
- and surface signals that require attention
This does not replace human judgement.
It enhances it.
It allows governance to operate with greater awareness, ensuring that decision-makers are focused on the moments that matter.
Humans provide judgement.
AI provides speed.
Governance needs both.
Bringing It Together
Individually, these capabilities are not new.
Many organisations already have elements of them in place.
What is changing is the need for them to operate together, consistently and in real time.
Without this, governance remains fragmented.
With it, governance becomes something far more effective:
a system that can adapt as the environment changes, while maintaining clarity, control and accountability.
What Comes Next
These capabilities define what effective governance requires.
The next question is how they are applied in practice.
In Part 3, I’ll explore how governance can operate at machine speed — and the role AI itself can play in enabling real-time monitoring, interpretation and escalation.
