AI and Leadership Decision-Making: 5 Decisions Leaders Must Own
Updated: Aug 27
AI and leadership decision-making are becoming increasingly connected. AI can analyse more information than any individual could reasonably process, identify patterns hidden within large datasets, challenge assumptions, model scenarios and produce recommendations in seconds. Used well, this should improve both the quality and speed of management decision-making.
But there is an important distinction between using AI to improve a decision and asking AI to make the decision for you. As the technology becomes more capable, the temptation to cross that line will increase. For routine, reversible and rules-based decisions, that may be entirely appropriate. There are thousands of decisions inside organisations that probably do not require senior human judgement at all.
The difficulty comes when decisions involve values, people, uncertainty, accountability and consequences that may unfold over years. In these situations, AI should inform the judgement, but leadership should continue to own it.
I think there are five types of decisions in particular that leaders should never fully delegate to AI.
1. Decisions that define purpose and direction
AI can help a leadership team understand markets, competitors, customers and possible future scenarios. It can analyse where growth is occurring, identify threats, test assumptions and model different strategic options. What it cannot ultimately decide is what the organisation should become.
Strategy involves choices about where to compete, what not to pursue, where to allocate limited resources and what the organisation is prepared to sacrifice in pursuit of its objectives. Those choices may be informed by data, but they are not determined by data alone.
Imagine AI analysis recommends discontinuing a low-margin product category because it consumes disproportionate resources and depresses profitability. Look at the product in isolation and the recommendation may be entirely rational. But perhaps that product provides access to strategically important customers who subsequently purchase higher-margin products elsewhere in the portfolio. Removing it improves one measure while damaging another.
This is where Purpose and Reality, the first two dimensions of my PRISM Decision-Making Framework, can help. Purpose asks what the organisation is actually trying to achieve. Reality asks whether the apparent problem reflects the whole commercial picture or only the part visible in the data.
AI can calculate the economics of the product and model the alternatives. Leadership still has to decide how that product fits within the wider strategy and which outcome ultimately matters most.
2. Decisions with significant consequences for people
AI will increasingly influence recruitment, performance management, workforce planning and organisational design, and much of that influence could be positive. It can expose inconsistencies, identify patterns humans miss and provide leaders with evidence that previously would have taken weeks to assemble.
However, decisions that fundamentally affect someone's career or livelihood should retain meaningful human judgement.
Imagine AI and automation have reduced the workload within a customer service function. Analysis suggests the team now has 25% excess capacity and recommends reducing headcount accordingly. Financially, the business case may appear straightforward.
But there are questions the calculation alone cannot answer. Is the reduction sustainable during seasonal peaks? What knowledge would leave the organisation? What would happen to service levels if demand increased? How would losing experienced people affect the remaining team, customers or the organisation's ability to grow?
PRISM encourages a leader to widen the frame. Reality tests whether the apparent excess capacity represents normal operating conditions. Stakeholders considers the consequences for customers, employees and the remaining team. Maturity asks whether today's cost reduction could create tomorrow's capability problem.
The answer may still be to reduce headcount. Good judgement does not mean avoiding difficult decisions. PRISM isn't designed to make difficult decisions softer; it is designed to make them better considered.
AI can provide the analysis. Leadership has to own the consequences.
3. Decisions involving ethical boundaries
Some leadership decisions are not about identifying the most efficient or financially attractive answer. They are about deciding what the organisation is prepared to do.
Should customer data be used in a particular way? Should technology be deployed to monitor employee behaviour? Should an organisation pursue a profitable opportunity if leadership believes the way that profit is generated conflicts with its values?
Consider personalised pricing. AI could potentially analyse customer behaviour, purchasing history and willingness to pay with extraordinary precision, allowing a business to optimise pricing at an individual level. The commercial case might be convincing.
But the fact that something can improve margin does not automatically mean an organisation should do it.
This is where Stakeholders and Maturity become particularly important. How would customers view the practice if they fully understood it? Could short-term margin improvement damage long-term trust? What happens if the approach becomes public? Would leadership be comfortable explaining the decision openly to customers, employees or shareholders?
AI can identify regulatory considerations, analyse customer behaviour, model the financial upside and highlight potential risks. It cannot remove the need for somebody to determine where the organisation's ethical boundary sits.
The question changes from “Can AI optimise this?” to “Should we optimise it in this way?”
That judgement, and the accountability that accompanies it, should remain human.
4. Decisions made under genuine ambiguity
AI performs particularly well when there is information to analyse, but leadership often becomes most valuable when there isn't enough.
A competitor behaves unexpectedly. A major customer suddenly changes direction. A market shifts. A new technology creates an opportunity for which there is little historical evidence. In these situations, there may simply be no dataset containing the answer.
Imagine a major competitor suddenly reduces prices by 15%. AI analyses the available market information, models elasticity and concludes that matching the price reduction offers the best probability of protecting volume.
It may be right.
But what if the competitor is clearing excess inventory? What if it is experiencing financial difficulties? What if the price reduction is unsustainable? What if your customers value service, availability or quality sufficiently that matching the price is unnecessary?
The problem isn't that the AI analysis is wrong. The problem is that the information available to the AI may not contain the answer.
This is where Reality and Insight become valuable. What do we actually know? What are we assuming? What evidence contradicts the recommendation? Which perspectives have we not considered? What additional information could change our view?
AI can be extremely useful in answering those questions. It can generate alternative interpretations, challenge assumptions and identify possibilities a leadership team may have overlooked. In fact, I think this could become one of its most valuable roles in executive decision-making.
PRISM does not replace the AI analysis. It provides a structure for challenging it.
Uncertainty should encourage leaders to use AI to improve their thinking, rather than use AI to avoid making the judgement themselves.
5. Decisions whose consequences will outlive the data
Some of the hardest leadership decisions are those where the immediate answer looks attractive but the long-term consequences are uncertain.
Closing a facility may improve next year's EBITDA. Reducing investment might protect this year's margin. Cutting prices could restore lost volume. Removing capability may lower the current cost base. Each decision can appear rational within the timeframe being analysed while creating consequences that only become visible years later.
Imagine AI modelling shows that closing a manufacturing facility and consolidating production elsewhere could add £2 million to EBITDA over the next two years.
That deserves serious consideration.
But what happens in year three if demand increases? What specialist knowledge disappears when the facility closes? What happens to lead times, supply-chain resilience or customer relationships? If the decision turns out to be wrong, how expensive would it be to rebuild the capability?
This is where Maturity, the final dimension of PRISM, becomes particularly powerful.
Maturity asks leaders to borrow tomorrow's eyes and consider how a decision might age. Rather than only asking whether the decision makes sense today, it asks what the organisation might think about the same decision in three, five or ten years.
AI can help enormously. It can model scenarios, stress-test assumptions and estimate the financial consequences of different futures. But eventually leadership still has to decide which risks to accept and which future it is prepared to create.
PRISM and AI and Leadership Decision-Making: A Human Judgement Layer
None of this means leaders should avoid using AI for important decisions. I increasingly think the opposite is true. For consequential decisions, leaders should actively use AI to challenge their thinking.
Ask AI what assumptions you might be making. Ask what evidence contradicts your position. Ask which stakeholder perspectives are missing. Ask it to construct the strongest argument against your preferred option. Ask what second-order consequences you may have overlooked and what would need to be true for your decision to turn out badly.
This is where I increasingly see PRISM as a useful human judgement layer around AI-assisted decision-making.
Purpose asks: What are we ultimately trying to achieve?
Reality asks: What do we actually know, and what are we assuming?
Insight asks: What evidence, experience and perspectives might challenge our view?
Stakeholders asks: Who experiences the consequences of this decision?
Maturity asks: How might this decision look when we borrow tomorrow's eyes?
AI can contribute enormously to every dimension. It can interrogate Reality, generate Insight, identify overlooked Stakeholders and model possible futures. It can even challenge whether the stated Purpose is consistent with the decision being considered.
But it should not own the final answer.
The objective is not to choose between AI and human judgement. It is to combine the analytical capability of AI with the contextual understanding, perspective and accountability of leadership.
AI may make human judgement more valuable, not less
There is an understandable debate about which management activities AI will eventually replace. I think there is another side to that question.
As AI makes information easier to access, analysis faster and recommendations cheaper, the relative value of good human judgement may actually increase.
Imagine two leaders receiving exactly the same AI-generated analysis. One accepts the recommendation because the numbers appear convincing. The other challenges the assumptions, considers what the data cannot see, seeks different perspectives, examines who will experience the consequences and considers how the decision might age.
They have access to the same technology. The difference in the quality of the eventual decision comes from judgement.
That may become one of the defining leadership capabilities of the AI era.
The best leaders will not be those who resist AI or insist that human instinct is somehow superior to technology. Nor will they be those who automatically accept whatever recommendation the most sophisticated model produces. They will know when to trust the analysis, when to challenge it, what questions to ask and when the technically optimal answer is not necessarily the right decision.
Most importantly, they will retain accountability.
A leader cannot reasonably stand in front of employees, customers, shareholders or a board after a consequential decision and say, “The AI recommended it.”
AI can provide the analysis, challenge assumptions, identify alternatives and even recommend a course of action. But for decisions that define organisations, affect people's lives, establish ethical boundaries, navigate genuine uncertainty and shape the future, leadership still has to own the judgement.
As AI becomes more capable, that responsibility becomes more important, not less.
Read More
FAQs:
Should leaders use AI to make business decisions?
Yes, AI can be extremely useful in business decision-making. It can analyse large amounts of information, identify patterns, test assumptions and generate alternatives. However, leaders should distinguish between using AI to improve a decision and delegating responsibility for the decision. The leader remains accountable for important decisions and their consequences.
What decisions should leaders never fully delegate to AI?
Leaders should be particularly cautious about fully delegating decisions involving people, ethics and values, strategic direction, stakeholder trade-offs and accountability. AI can provide analysis and recommendations in each of these areas, but the final judgement should remain with a human leader.
Why is human judgement still important when using AI?
AI can process information and generate recommendations, but leadership decisions often involve ambiguity, competing priorities, context, values and consequences that cannot be reduced to data alone. Human judgement allows leaders to interpret the information, consider what matters and take responsibility for the eventual decision.
Can AI replace leadership decision-making?
AI is more likely to augment leadership decision-making than replace it. It can make leaders better informed and expose assumptions they might otherwise miss, but important decisions still require judgement, context and accountability. As AI capabilities improve, deciding when to rely on AI and when human judgement must take precedence becomes an increasingly important leadership skill.
How should leaders use AI when making important decisions?
Leaders should use AI as an input into the decision rather than the owner of the decision. AI can help gather evidence, analyse scenarios, identify risks, challenge assumptions and present alternative perspectives. Leaders can then combine those insights with experience, context, stakeholder considerations and human judgement before deciding what to do.
Who is accountable for a decision made using AI?
Accountability should remain with the individual or organisation responsible for the decision. Using an AI system to analyse information or recommend an action does not remove leadership responsibility for evaluating that recommendation and deciding whether it should be followed.
What is the role of PRISM in AI-assisted decision-making?
The PRISM Framework, developed by Paul Bensley, provides a structured human judgement layer around AI-assisted decisions. It encourages leaders to examine five dimensions: Purpose, Reality, Insight, Stakeholders and Maturity, helping them consider not only what AI recommends, but whether the recommendation makes sense in the wider context of the decision.
How can leaders prevent over-reliance on AI?
Leaders should avoid treating AI outputs as inherently correct. Important recommendations should be challenged, assumptions tested and relevant evidence verified. Leaders should also consider perspectives and contextual factors that may not be adequately represented in the information available to the AI.
Will AI make human judgement more or less important?
As AI makes information, analysis and recommendations increasingly accessible, the relative importance of judgement may actually increase. The leadership advantage becomes less about possessing information and more about interpreting it, understanding context, making appropriate trade-offs and taking responsibility for what happens next.
Comments