AI as an organisational truth serum describes the way AI adoption forces an organisation to state, in explicit terms, how its decisions are actually made: who really decides, which priority wins when goals conflict, and which workarounds keep the work moving. To automate or support a decision, someone has to write down its logic. That is usually the moment leaders discover the logic was never agreed.
The effect goes by several names: AI as a forcing function, the organisational X-ray, decision-rights exposure and the gap between the formal and the real operating model. Each describes the same thing. The organisation chart and the process manual say one thing, and the way work is actually done says another. AI cannot run on the second version until someone makes it visible.
This matters at board level because the frameworks boards are now expected to follow assume decision authority is already clear. ISO/IEC 42001:2023 requires top management to assign and communicate responsibilities and authorities for relevant roles in the AI management system. The NIST AI Risk Management Framework (AI RMF 1.0) places accountability structures, with documented roles and responsibilities, inside its Govern function. For high-risk systems, the EU AI Act (Regulation (EU) 2024/1689) requires human oversight by people with the competence and authority to exercise it. None of these can be satisfied by an organisation that does not know who really makes the call.
The gap between official and actual practice is already wide. 78% of AI users are bringing their own AI tools to work, and 60% of leaders worry their organisation’s leadership lacks a plan and vision to implement AI (Source: Microsoft and LinkedIn, 2024 Work Trend Index Annual Report, 2024).
This analysis builds on a recent conversation with Bud Caddell, founder of the organisational design consultancy NOBL.
Why does AI expose how decisions really get made?
AI exposes how decisions are made because it cannot work on unwritten rules. People navigate ambiguity through relationships and seniority. A system needs its inputs, thresholds and escalation points specified. The act of specifying them turns tacit habits into visible choices.
Caddell’s point is that AI is not creating new organisational problems. It is accelerating the exposure of old ones. As he puts it, “We don’t even know how our organizations are aligned.” He does not treat this as failure: “Organizations run on politics,” he notes. The difficulty is that a system cannot read the room.
This is also why so many AI pilots stall. A pilot can succeed inside one team’s informal rules. Scaling it means crossing into other teams whose rules differ, and nobody has decided which set applies. We examine that pattern in AI Pilot Purgatory.
What does AI typically reveal, and why is that board-level intelligence?
AI typically reveals three things: unclear decision rights, unresolved trade-offs and undocumented workarounds. Each is board-level intelligence because it describes where the organisation’s real risk and real value sit.
| What AI surfaces | What it tells the board | Executive decision required |
|---|---|---|
| Nobody can say who breaks a tie between functions | Decision rights exist in practice but not on paper | Name the decision owner for each contested decision |
| Teams pursue growth and efficiency, or speed and control, at the same time | Trade-offs are being settled by escalation and delay | State which priority wins, and the threshold at which that changes |
| Critical steps depend on spreadsheets, personal knowledge or unofficial tools | Controls and processes differ from what is reported | Decide which workarounds to formalise and which to retire |
| Staff use unapproved AI tools to get work done | Official processes are failing the people who run them | Provide sanctioned tools and a clear use policy |
Caddell’s remedy for the second row is simple: choose one priority over another until a defined threshold is reached. That does not remove the trade-off. It makes it operational, so teams stop escalating the same question.
Is this exposure a threat to leadership?
It is not a threat, although it can feel like one. What AI exposes was already true. The difference is that the board can now see it, and a known weakness can be governed while an unknown one cannot. The real risk is the opposite reaction: encoding today’s unclear rules into automated systems, which makes the confusion faster and harder to challenge.
Caddell’s practical advice supports this. Start where work hurts most, with the manual data entry, repetitive reporting and broken handoffs people dread. Relief builds trust faster than sophistication, and those painful processes are usually where the undocumented workarounds are hiding. For how that trust turns into lasting use, see From Hype to Habit.
The AIPEX view: treat every AI project as a decision-rights audit
At AIPEX we think the most valuable output of early AI adoption is often not the automation. It is the map of how the organisation actually decides. AI produces it as a by-product, and most organisations discard it.
We recommend that the board make the map a formal governance instrument, a decision rights register, maintained alongside the AI portfolio:
- Capture: every AI use case records the decisions it touches, who holds each one today in practice, and any workaround found along the way.
- Resolve: contested decisions and conflicting priorities go to the executive team for an explicit ruling, not back to the project team to absorb.
- Report: the register comes to the board periodically, as evidence of how the organisation really operates and where accountability is thin.
The register also answers a question the standards above assume has been settled: who is accountable when an automated decision goes wrong. We address that directly in Who Owns the AI Agent When It Breaks?
What should the board ask about what AI is revealing?
The board should ask what AI adoption has taught the organisation about itself, not only what it has delivered:
- Which decisions have our AI projects found to have no clear owner, and who now owns them?
- Where have we discovered that actual practice differs from documented process?
- Which trade-offs have we explicitly settled, and which are we still resolving through escalation?
- How widely are staff using unapproved AI tools, and what does that say about our official processes?
- Are we encoding unresolved disagreements into automated systems?
If nobody at executive level is collecting these answers, the organisation is paying for the insight and not keeping it. A Fractional Chief AI Officer can own the decision rights register and bring its findings to the board. The Executive Strategy Day surfaces the unwritten rules and unresolved trade-offs across your leadership team before they are built into your AI systems.
Frequently asked questions
Why do AI projects surface hidden decision-making problems?
To automate or support a decision, its logic, thresholds and escalation points must be written down. That exercise reveals where decision rights, priorities and processes were never actually agreed. For a full explanation, see AI as an Organisational Truth Serum: What It Reveals About How Decisions Really Get Made →
What is a decision rights register for AI?
It is a board-level record, kept alongside the AI portfolio, of the decisions each AI use case touches, who holds them in practice and any workarounds discovered. Contested decisions go to the executive team for an explicit ruling. For a full explanation, see AI as an Organisational Truth Serum: What It Reveals About How Decisions Really Get Made →
How do unclear decision rights keep AI pilots from scaling?
A pilot can succeed within one team's informal rules, but scaling it crosses into teams whose rules differ and nobody has decided which apply. Without a named owner and agreed decision logic, proven pilots stall. For a full explanation, see AI Pilot Purgatory: Why Successful Pilots Never Reach Production →