By Martin Colyer, Director of Digital and AI, LACE Partners,
Most senior leaders I speak to are using AI to prepare reports, briefings and recommendations to save time. But when I ask what happens when something in that output is wrong and a decision has already been made based on it, there is usually a pause.
The recent resignation of West Midlands Police Commissioner Craig Guilford made that question very public. AI-generated content containing inaccuracies was used in a sensitive report connected to discussions around Maccabi Tel Aviv supporters attending a football match in Birmingham. Many stated that it wasn’t sufficiently challenged before it influenced decision-making. Political and reputational chaos ensued in the media.
The specific circumstances are a matter of public record. What interests me more is what the incident reveals about where most organisations actually are with AI right now and the gap between how confidently they are adopting these tools and how prepared their leaders are to use them responsibly.
Accountability doesn’t transfer to the technology
In regulated environments or, frankly, in any setting where decisions carry social, legal, or reputational consequences, accountability lies with the person. The leader who chooses to rely on AI to form judgement and signs off on the decision remains responsible for it. Just because the AI output looks authoritative, it should never be perceived as a holy grail.
This is where a lot of organisations are exposed right now. AI tools can produce content that looks very structured and confident. But we don’t know whether the key context is missing or the details are wrong. If no one is clearly responsible for questioning the output, it tends to move through review processes on the assumption that someone else has already checked it. However, that assumption can present a serious risk.
Adoption is running ahead of education
Almost 70% of UK businesses are either already using AI or actively exploring it. While the pace of adoption is accelerating, education and governance rarely keep up. And that is a big problem.
In most organisations, AI use starts informally. Most employees would use a tool to draft a document, summarise a meeting or pull together research. It saves time, it becomes a habit, and before long, AI-generated content is embedded in workflows. Quite often, this happens before anyone has agreed on where the boundaries sit or who is responsible for the decisions those outputs inform.
By the time a policy lands, the habits are already in place. And written policies alone don’t change how people behave when they’re under pressure, and the output in front of them looks credible. That’s the reality in most of the organisations – the guardrails come after the fact.
Writing a policy is a reasonable starting point, but it’s not where the work ends. The bigger gap is in education – helping leaders genuinely understand how these systems produce outputs, what they can and cannot do, and when a decision still requires the ‘human-in-the-loop’ step.
As humans, we are tempted to trust something that looks legitimate. Familiarity with a tool breeds ease, and ease can tip into over-reliance, particularly when time is short. The Guilford case is a reminder that the stakes of relying too heavily on AI are very high. When it comes to making decisions that involve public trust, data privacy, or employee relations, it carries weightier consequences that no time-saving can offset. In those situations, judgment matters far more than efficiency.
Leaders need to feel genuinely equipped to pause an AI output, ask the right questions, and decide when to set the tool aside. While most organisations will be focusing on mandating AI as a technical skill across the board, the bigger focus should be on judgment skills, which need to be deliberately developed.
What good governance actually looks like
Effective AI governance comes down to three things: clarity on when AI can be used, a clear process for how outputs are validated before they influence a decision, and named accountability for sign-off. Without those in place, built into how work actually gets done, organisations are leaving themselves exposed.
This matters just as much in HR functions, where AI is increasingly used for recruitment screening, performance reviews and employee relations. These are decisions with a direct impact on people’s working lives. Technology has no awareness of organisational history, team dynamics or emotional context. It cannot read the room. Those considerations remain human responsibilities.
Why leaders set the standard
Senior leaders may not hold the deepest technical knowledge in the room, but they set the tone for how AI is used and how their teams exercise accountability. When leaders visibly question decisions and perform initial checks to ensure they were made using reliable sources, teams learn by example. Otherwise, responsibility and a single source of truth become hard to locate.
The West Midlands incident shows how fast trust can unravel when leadership judgment and knowledge is found wanting. While public institutions might feel it through media coverage, private organisations feel it through disengagement and attrition. In both cases, recovery takes time, and the reputational cost is real.
AI will keep changing how information is produced, summarised and presented. That is not going to slow down. What does not change is the expectation of leaders. When a decision carries real consequences, for an organisation, the public or for an individual employee, responsibility belongs to the person who chose to act on it, whatever information they relied on.
