Organisations rarely fail because they lack information. They fail because the story they tell themselves about that information no longer holds, and no one is able, or willing, to say so clearly enough.
Every organisation operates with a narrative. It is rarely written down in a single place, but it is widely understood. It explains why the strategy is sound, why the business model is resilient, why the risks are manageable, and why current performance is a reasonable guide to future outcomes. This narrative is not a deception in any deliberate sense. It is a stabilising device. It allows people to act with confidence in the face of uncertainty.
The difficulty is that narratives do not degrade abruptly. Instead, they shift gradually, often in ways that are difficult to detect at the point at which they begin.
Small inconsistencies appear. A result that does not quite fit. A workaround that becomes normal. A concern that is raised but not pursued. None of these, in isolation, is sufficient to challenge the prevailing story. Each can be explained, absorbed, or deferred. Over time, however, the gap between the official narrative and lived reality widens. The organisation continues to act as if the story is intact, even as the conditions that supported it have shifted.
Traditional risk management struggles to detect this form of drift. It is designed to identify discrete risks, not to question the coherence of the narrative within which those risks are interpreted. Registers catalogue issues that fit existing categories. Dashboards track indicators that have already been agreed. Neither is well suited to recognising when the underlying story itself is no longer reliable.
Risk Intelligence treats narrative as an object of attention. It asks not only what risks have been identified, but whether the framing through which those risks are understood still holds. It pays particular attention to points of tension. Where does behaviour diverge from stated intent? Where does confidence remain high despite accumulating ambiguity? Where are explanations becoming more elaborate in order to preserve a simple story?
These are not failures of control. They are early signs of self-deception.
Self-deception in organisations is rarely malicious. It is adaptive. It allows activity to continue in the presence of uncertainty. The danger arises when the mechanisms that might challenge the narrative are weakened. When dissent is framed as negativity, when anomalies are treated as noise, and when success reinforces the belief that the story must be correct, drift accelerates.
AI does not remove this dynamic. In many cases, it intensifies it. Models trained on historical patterns will tend to reinforce the narrative that those patterns imply. When outputs align with expectation, confidence increases. When they do not, the tendency is to adjust the model or the data rather than to question the narrative. The system becomes more precise in describing a reality that no longer exists.
Building for Risk Intelligence therefore requires creating conditions in which narrative can be questioned without destabilising action, and in which divergence between story and reality can be surfaced early enough to matter. This is not straightforward. It involves making space for alternative interpretations, preserving minority views, and revisiting assumptions even when there is no immediate trigger to do so. It requires treating coherence as something to be tested, not assumed.
If narrative is treated as an object of risk, most current infrastructures are blind to it. They capture what fits the story, not where the story is beginning to fail. Addressing this requires environments that preserve tension between competing interpretations and make divergence visible over time. Without that, drift remains undetected until it becomes collapse. RiskMetrica is designed to hold that tension rather than resolve it prematurely, allowing divergence to be seen before it becomes failure.
Within such an environment, narrative is not replaced. It is made visible, contestable, and revisitable. Signals that challenge prevailing assumptions are not suppressed or deferred, but connected and tracked over time. Judgement remains exposed, and therefore open to challenge. This does not eliminate self-deception, but it makes it harder to sustain unnoticed.
For boards, the implication is direct. Oversight cannot rely solely on the consistency of reporting. It must engage with the consistency of the story itself. Where that story begins to require increasing effort to maintain, attention is warranted, even if no formal risk has yet crystallised.
The question is not whether the narrative is coherent, but whether the organisation can still recognise when it no longer is.
Richard Anderson
Founder, AndersonRisk
This essay was first published on 12 May 2026 on my Substack.