Glossary

One home for every idea.

This glossary uses six categories: Core construct; Tool / practice; Diagnostic / heuristic; Software component; Established external concept; and Output. The categories clarify hierarchy rather than importance. Each entry points to the principal chapter or chapters where the concept is introduced or developed. Entries are arranged alphabetically.

Being in the glossary does not make a term a core framework concept. The framework is built from about a dozen enduring conceptual families; most entries here are tools, heuristics, or established ideas from outside the book that those families draw on.

76 of 76 terms

Adversarial Engine Software component · Ch 13; Ch 15: The AI-powered Red Team. A system that continuously generates the strongest available challenge to the organisation's current strategic commitments, drawing on external sources the organisation's own analysis might not surface. The Engine addresses the social problem the human Red Team faces: It has no status, no time pressure, no ego, and no career risk.
After-Action Review (AAR) Established external concept · Ch 5; Ch 8; Ch 16: The tool that closes the OTA loop after every significant action. Four questions in order: What did we intend to happen? What really happened? Why was there a gap? What will we do differently? Structural pair to the Pre-Mortem. In the AI-augmented organisation, the AAR feeds a searchable archive that makes every completed action substrate for the next team's planning.
Agency Core construct · Ch 1: The capacity to stay aligned with reality, make sense of what reality implies, and intervene with intent ahead of events. Observe, Think, Act is the operating system of agency.
Alert fatigue Diagnostic / heuristic · Ch 14: The failure mode in which high-volume, time-pressured monitoring trains humans to treat unflagged items as safe by default, so the model's blind spots become the analyst's blind spots too. It's most severe in the combination of high volume, time pressure, a strong track record, and an opaque interface – precisely the standard configuration of most deployed AI monitoring systems.
Anomaly Diary Tool / practice · Ch 2; Ch 6: A running log of signals that do not fit the current model – observations that cannot yet be explained but are too persistent or sharp to dismiss. The Diary is not a worry list; it is a structured prompt to revisit unexplained data before it has been forgotten.
Assumption register Tool / practice · Ch 13; Ch 15: A living list of the working assumptions on which current strategy, decisions, and scenario branches rest — what the organisation believes to be true but has not yet verified. In the AI era, the same register is monitored continuously against incoming evidence.
Asymmetry between failure and success Diagnostic / heuristic · Ch 12: The empirical observation that failure and success are not mirror images. Some modalities are disproportionately dangerous in failure while others contribute more strongly to success. The implication is descriptive: preventing catastrophe and building durable advantage require overlapping but different emphases.
Autonomy Ladder Tool / practice · Ch 14; Ch 16: The principle that AI earns expanded decision-making authority through demonstrated, measured performance in a defined context – never through a blanket decision to "trust AI." Each threshold is explicit, tied to a specific class of decisions, and the human authority to pull autonomy back is retained. A mechanism within Decision Architecture – see that entry.
Calibrated trust Tool / practice · Ch 13: The governing design principle behind every AI-native tool: enough confidence to use the system, enough independence to challenge it, and a track record of overrides and outcomes that shows where trust is actually earned. It guards against two failure modes – bypass (quietly routing around a system nobody trusts) and abdication (treating its outputs as conclusions rather than hypotheses).
Calibration Tool / practice · Ch 7; Ch 15: A discipline that runs through probability-bearing tools in the Toolkit: Pre-Mortem failure-mode scores, Scenario Mapping branch likelihoods, Red Team confidences, assumption-register confidence values, and AAR / double-loop predictions. Express predictions as numbers rather than verbal hedges, anchor on base rates, update with evidence, and track accuracy over time. In the AI era, the Calibration Tracker automates the scoring; the practice is unchanged.
Calibration Tracker Software component · Ch 13; Ch 15: A software component that scores probability-bearing outputs produced by both humans and AI against eventual outcomes. It produces calibration profiles for human decision-makers, AI systems, teams, topics, and time horizons. Individual human profiles can be confidential and visible only to the individual and their chosen coach; aggregate profiles are visible to leadership and feed the Sentinel at aggregate level.
Complacency Diagnostic / heuristic · Ch 10: The passive comfort – the absence of paranoid optimism – that stops the organisation scanning because nothing has recently gone wrong. One of the six named pathogens that corrupt the OTA cycle from the inside.
Core competence Established external concept · Ch 7: The capability at the heart of an organisation's competitive advantage – defined by three tests: it delivers distinctive value to customers, it is difficult for competitors to imitate, and it can be leveraged across multiple products or markets. Capabilities that fail any of these tests are competencies, not core competencies.
Cultural suppression Diagnostic / heuristic · Ch 2: Failure mode in which the data exists but the culture punishes its delivery. It can be soft – feedback is heard, even applauded, then quietly filed away – or hard, where a toxic culture actively punishes truth-telling and the sensing system becomes the first casualty.
Curve dashboard Tool / practice · Ch 2: A conventional dashboard reports status at a single moment – what the number is right now. A curve dashboard reports trajectory instead: slope, curvature, inflection points, whether the number is accelerating or decelerating. A conventional dashboard can show healthy, on-target growth in the same period a curve dashboard reveals that the growth rate itself has been steadily bending downward.
Data estate Tool / practice · Ch 13: The organisation's data in its entirety: infrastructure plus the high-quality, structured, decision-time-captured, attributable data stored in it. Every AI-native tool draws on it, and it compounds into competitive advantage – two years of structured decision history is an advantage a competitor can't buy, license, or import, because it can only be built by living through the decisions.
Decision Architecture Tool / practice · Ch 11; Ch 16: The design of decision rights as the organisation's primary architecture: the map of which decisions live at which level under which conditions for upward referral. In the pre-AI organisation, every level on that map is human; in the AI-era organisation the level set expands to include AI as one of the echelons authority can live with. The Autonomy Ladder is the mechanism within this architecture that governs how AI earns a given level – see that entry.
Decision latency Diagnostic / heuristic · Ch 4: The time from insight to commitment. It is a system property shaped by process design, incentives, decision rights, and external leverage rather than simply by an individual's decisiveness.
Decision Ledger Software component · Ch 13; Ch 15: A structured log of every significant decision: what was decided, the load-bearing assumptions it rests on, the Pre-Mortem findings, and an open section to be completed at the After-Action Review. The Ledger makes the before/after learning bracket auditable, searchable, and protected from post-hoc revision.
Decision Record Tool / practice · Ch 4: A single document opened before a commitment is made and closed once execution is complete. It records, up front, the hypothesis, the load-bearing assumptions, how learning will be measured, the decision gate, and labels for reversibility and impact. At close, it records what happened versus what was expected and what would be done differently.
Designability test Diagnostic / heuristic · Ch 9: The test used to decide whether a dimension belongs among the five modalities: Can it be deliberately designed, built, or shaped? Direction, Structure, Processes, Capability, and Culture pass; outputs such as adaptation do not.
Direction alignment Tool / practice · Ch 8: The operational discipline that keeps an organisation's daily work connected to strategic direction, most often through OKRs or an equivalent system of measurable commitments.
Dissent log Tool / practice · Ch 11: The Decision Record field that captures any override of a tool finding, together with the name of the person responsible and the written reasoning. Over time it reveals which warnings are repeatedly dismissed, by whom, and with what results.
Domain Expert Sign-Off Tool / practice · Ch 8; Ch 16: A validation practice requiring an independent expert — someone with no material stake in the outcome — to produce a written opinion on a major commitment that rests on a technical claim. In the AI era, invocation can be automatic, and in narrowly defined domains where AI has earned authority it may perform the sign-off; the underlying practice keeps the same name.
Double-loop learning Established external concept · Ch 5; Ch 16: Chris Argyris's distinction between correcting action within existing assumptions and questioning the assumptions themselves. In Chapter 16, searchable AAR history makes machine-scale cross-cycle pattern detection possible.
Five Modalities Core construct · Ch 9; Ch 12: The five dimensions along which the OTA loop can break or excel: Direction (where the organisation is pointed), Structure (how it is governed and wired), Processes (how it operates day to day), Capability (what it can actually do), and Culture (how people behave when no one is watching).
Groupthink Established external concept · Ch 10: The social pressure toward alignment that narrows the models a group will seriously consider. Identified as a social dynamic by Irving Janis; operating in every group with hierarchy and belonging as organising principles. One of the six named pathogens that corrupt the OTA cycle from the inside.
HiPPO effect Established external concept · Ch 3: The Highest Paid Person's Opinion. The social dynamic in which the most senior person's perceived preference bends group analysis toward their position before anyone has spoken a word. The HiPPO effect does not require a domineering leader or a spineless team; it operates through the ordinary social physics of hierarchy and belonging.
Homogenisation risk Diagnostic / heuristic · Ch 13: The concern that if every organisation runs the same AI systems on similar data, their analysis converges toward the same conclusions, erasing the competitive edge the tools were meant to provide. AI raises the analytical floor for everyone but the ceiling is set by the distinctiveness of the human framing – what's asked, what's monitored, what's challenged.
Hubris Diagnostic / heuristic · Ch 10: Excessive confidence in a mental model that has been right for long enough to feel permanently right. The model stops being a map and becomes the territory. One of the six named pathogens that corrupt the OTA cycle from the inside.
Illusion of Action Diagnostic / heuristic · Ch 1: The mistake of treating visible activity as effective action. Meetings, decisions, and motion can all be real while remaining detached from an accurate reading of reality and sound thinking.
Incumbent's trap Diagnostic / heuristic · Ch 4: The constraint that arises when the assets and ecosystem relationships that made an incumbent strong become leverage that partners can use to prevent it from changing. Unlike the innovator's dilemma, the constraint sits in the external value chain as well as inside the firm.
Inertia Diagnostic / heuristic · Ch 10: The tendency of successful organisations, when disrupted, to accelerate existing activities rather than change direction. Named "active inertia" by Donald Sull: The machinery of past success makes acceleration feel like action. One of the six named pathogens that corrupt the OTA cycle from the inside.
Inflection point Diagnostic / heuristic · Ch 17: The point at which a changing trajectory crosses a threshold and the strategic meaning changes. In Chapter 17, it is used for the moment when AI capability surpasses what an organisation can otherwise achieve and does not look back.
Instrument Audit Tool / practice · Ch 6, referenced in Ch 8; Ch 12; Ch 13; Ch 14: A periodic review of the full sensing system, asking not just whether instruments are functioning but whether they are still pointed at the right variables. The Audit catches instrument drift: measures that once tracked important signals but have slowly decoupled from the thing the organisation genuinely needs to know.
Interrupted loop Diagnostic / heuristic · Ch 5: An OTA cycle that may complete operationally but fails to carry its context and learning into the next cycle. The formal record survives while observations, reasoning, failed experiments, and tacit knowledge are lost.
Last-Mile Question Diagnostic / heuristic · Ch 1: The question 'If I had to act tomorrow, what is still unclear?' The answer diagnoses whether the remaining gap lies in Observe, Think, or Act before a consequential commitment is made.
Learning rate Established external concept · Ch 6: The experience-curve relationship, associated with Theodore Wright, in which cost or effort falls by a relatively stable fraction as cumulative production doubles. In Reality Instrumentation it helps turn slope observations into trajectory estimates.
Meta-Loop Software component · Ch 13; Ch 16; Ch 17: An AI-based tool with a different cadence and remit from the Sentinel. Reads the Sentinel's anomaly outputs over time plus scans external sources – research papers, peer-organisation practice, doctrine shifts, new AI capabilities. Applies the OTA loop to the Toolkit itself: observes how the tools are being used; thinks about whether they are still the right tools, configured the right way, for the failures the organisation encounters; acts by proposing additions, retirements, recombinations, and reconfigurations of the Toolkit to human leadership for ratification.
Methodological blindness Diagnostic / heuristic · Ch 2: Failure mode in which data is gathered rigorously but measures the wrong trend. It's more dangerous than having no data at all, because excellent measurement of the wrong variables produces false confidence rather than visible uncertainty.
Misleading Incentives Diagnostic / heuristic · Ch 10: Reward structures that have drifted from the outcomes the organisation genuinely needs. People optimise for what is measured and rewarded, not for what the strategy requires. One of the six named pathogens that corrupt the OTA cycle from the inside.
Model drift Diagnostic / heuristic · Ch 3: The widening gap between an internal model and external reality. Under volatility or regime change, a model that once worked can remain internally coherent while becoming progressively less useful.
Normalisation of Deviance Established external concept · Ch 10: The gradual process by which exceptions become standards. Each small deviation from protocol that does not immediately produce failure makes the next deviation slightly more acceptable, until the exception is the rule. One of the six named pathogens that corrupt the OTA cycle from the inside.
Observation Review Tool / practice · Ch 6, embedded in Ch 11; Ch 14: The recurring structured review at which the organisation's current signals are surfaced, assessed against prior expectations, and fed into the Thinking phase.
Observation Stack Tool / practice · Ch 2: A five-layer system that turns fragments from an anomaly diary into a coherent read on direction: raw signals accumulate into logged anomalies, which cluster into emerging patterns, which reveal slope, which supports a strategic interpretation. The Anomaly Diary asks "What feels off?" The stack asks "Is a pattern forming?"
OKRs Established external concept · Ch 8; Ch 13; Ch 16: Objectives and Key Results – the most widely used method for direction alignment. The Objective answers what the organisation wants to achieve; Key Results answer how it will know it got there.
Operator test Diagnostic / heuristic · Ch 9: A boundary test for Processes versus Capability: If more skilled people were substituted while the process stayed the same, would the outcome change? If yes, the problem is Capability; if no, it is Processes.
OTA 200 Core construct · Introduction: The empirical foundation of the book's framework: 200 companies, 100 failures and 100 successes, scored against the 15-cell grid. Cases were selected on two criteria: diversity across industries, geographies, and eras; and data quality sufficient to support rigorous root-cause (failure) or differentiation (success) analysis. The full methodology – selection criteria, scoring rubrics, known limitations – is documented in the book’s companion website.
OTA Drift Diagnostic / heuristic · Ch 1: Failure after success caused when a strength that once produced advantage is applied for too long after conditions have changed. The right capability for yesterday becomes the blind spot for tomorrow.
OTA Failure Triangle Diagnostic / heuristic · Ch 1: A self-diagnostic heuristic for three recurring OTA failure spaces: blind (Observe failure), clever but wrong (Think failure), and paralysis or performative action (Act failure).
OTA loop Core construct · Ch 1; Ch 5: The book's central organising framework: Observe, Think, Act. Three phases that any individual, team, or organisation must execute continuously to maintain an accurate picture of reality, reason about it rigorously, and act before the window closes. Chapter 5 uses the flywheel as the metaphor for how repeated loops compound learning and performance.
OTA × modalities grid / 15-cell grid Core construct · Ch 9; Ch 12: The 15-cell diagnostic map created by crossing the three OTA phases with the five modalities. It makes visible where failure concentrates, where success concentrates, and where the two patterns differ.
Paranoid Optimism Core construct · Introduction: The disposition this book argues for. Paranoid enough to look for everything that can go wrong – because reality does not reward wishful thinking and weak signals precede strong consequences. Optimistic enough to believe that knowing what can go wrong puts you ahead of those who refuse to look. Vigilance is not the opposite of confidence; it is its precondition.
Practice inventory Diagnostic / heuristic · Ch 11: A behavioural inventory of which pre-AI Toolkit practices are actually embedded in an organisation. It asks what teams repeatedly do, not what they say they value.
Pre-Mortem Established external concept · Ch 7: A structured thinking tool run before a commitment is made. The team imagines the project has already failed and asks: What went wrong? The Pre-Mortem does not prevent commitment – it stress-tests assumptions before the momentum of execution makes revision socially costly. Structural pair to the After-Action Review.
Psychological Safety Established external concept · Ch 6; Ch 10; Ch 13: The condition under which people can surface what they know, doubt, or believe is wrong without humiliation or punishment. Amy Edmondson's work and later team research established the concept. In this book it is a prerequisite for truthful use of tools such as Pre-Mortem, AAR, Red Team, and the Sentinel; it is an established external concept, not an OTA pillar.
Reality Instrumentation Tool / practice · Ch 2; Ch 6: The deliberate design and maintenance of what an organisation measures so that its instruments are pointed at strategically relevant reality: slopes, KPIs, anomalies, leading indicators, owners, and review cadences. It is the human-era lineage that the Reality Lattice extends.
Reality Lattice Software component · Ch 13; Ch 14: The AI-augmented observation layer. A six-tier stack that extends the organisation's perceptual range beyond what human observers can process – from raw data ingestion through signal extraction, pattern recognition, anomaly detection, and prioritised delivery to decision-makers.
Red Team Established external concept · Ch 7, evolved to Adversarial Engine in Ch 15: A designated challenger tasked with constructing the strongest possible version of the counter-case. Not a devil's advocate performing objections, but a person or group that takes the opposing argument seriously. The test of whether it is working: The people in the room are slightly worried the Red Team might be right.
Residual risk after applying the Toolkit Diagnostic / heuristic · Ch 12: The residual risk that remains, in the OTA-200 model, after the five-tool prevention set is applied. The residual changes shape as it shrinks, concentrating more heavily in Processes and Culture. The phrase describes the result rather than naming a separate framework architecture.
Reversibility / culture of reversibility Tool / practice · Ch 4; Ch 8: The discipline of distinguishing decisions that can be undone or modified at acceptable cost from those that cannot, and matching decision speed and governance accordingly. Chapter 8 describes the culture of reversibility: push reversible decisions toward the people closest to the work and reserve heavier deliberation for commitments that cannot easily be undone.
Scenario Mapping Core construct · Ch 7; Ch 11; Ch 15: A thinking tool that keeps several plausible futures alive, identifies leading indicators for each branch, and supports reversible preparation on secondary branches. In the AI era, monitoring and probability updates become continuous.
Sentinel Software component · Ch 13; Ch 16: A software component that consumes structured traces from the other AI-era tools — including Reality Lattice signals, Decision Ledger entries and outcomes, Adversarial Engine challenges and responses, Scenario Mapping movements, double-loop patterns, aggregate Calibration Tracker scores, and assumption-register updates. It surfaces signatures of trouble before they compound, including the six pathogens and other anomaly classes.
Shadow committee Tool / practice · Ch 11: A rotating group of junior or mid-level employees that reports directly to senior leadership on operational reality. It creates a sensing channel that bypasses the editing effects of hierarchy and should rotate regularly to avoid learning what leaders want to hear.
Signal brief Output · Ch 13; Ch 14: A short synthesis of the most significant signals from observation, prepared before key meetings or reviews. In the human-era organisation a person composes it; in the AI-native form the system drafts it from instrumentation data and a human edits and approves. It is an output, not a framework construct.
Six Pathogens Core construct · Ch 10: The human and organisational diseases that infect the OTA cycle from within. They arrive with success and are sustained by the structure success builds. They are: Hubris, Complacency, Normalisation of Deviance, Groupthink, Misleading Incentives, and Inertia.
Slow loop Diagnostic / heuristic · Ch 5: An OTA loop whose cadence is tied too closely to the annual planning cycle, so signals are observed, interpreted, and acted on only after the competitive window has narrowed or closed.
Strategy Coherence Engine Software component · Ch 13; Ch 15: A software component that treats mission, vision, values, strategy, OKRs, governance, incentives, and budgets as a graph of artefacts whose claims can be checked against one another and against observed reality. It surfaces the pairs whose contradictions are largest and least explained; leadership decides which side should change. It works alongside the Decision Ledger, assumption register, and Adversarial Engine rather than forming a second framework.
Structural blindness Diagnostic / heuristic · Ch 2: Failure mode in which the data exists but can't reach decision-makers – it sits in the wrong dashboards, gets judged by the wrong metrics, or is blocked by incentives. Nobody is hiding anything; the organisation's structure simply prevents the signal from becoming strategically meaningful.
Synthetic certainty Diagnostic / heuristic · Ch 14: The failure mode in which the numerical format of an AI output suppresses the retrieval of contradictory information. A high-confidence score – presented as a point estimate without qualification – actively shortens the human's information search: The check becomes cursory because the number has already provided an answer. That shortcut compounds two ways: partly through output design (confidence intervals and reasoning shown alongside scores restore the check) and partly through volume (when confident alerts arrive in the thousands, the triage heuristic becomes the model's blind spot).
Thinking Pipeline Tool / practice · Ch 3; Ch 7: A six-step discipline for rigorous reasoning: separate observation from interpretation; form at least two competing hypotheses; identify the load-bearing assumption each rests on; anchor against base rates and reference classes; map first- and second-order consequences; define the breaking condition – the evidence that, if it appeared, would falsify the preferred hypothesis.
Toolkit Core construct · Ch 8; Ch 13: The leader's working repertoire of practices, frameworks, rituals, and conditions for training Observe, Think, and Act. The initial set of tools is introduced in Part II; in Part IV, those tools evolve into a set of AI-native tools built on the same underlying logic.
Triangulation Tool / practice · Ch 6; Ch 14: No single source of truth for a major strategic signal – aim for at least three independent sources, not three reports that trace back to the same data. In Part IV, the same discipline is applied to AI output: requiring an independent second source before high-stakes action counters synthetic certainty and alert fatigue.
Truth pipeline Tool / practice · Ch 11: The structural channels that carry frontline reality to people with authority to act. In Chapter 11 it is built from instrument owners, the Observation Review, the shadow committee, and standard whistleblowing channels, with psychological safety as the condition that lets truth enter the pipeline.
What Would Have to Be True (WWHTBT) Established external concept · Ch 7; Ch 15: A reframing technique that shifts the question from 'should we do this?' to 'what conditions would have to hold for this to be the right choice?' Each option generates testable conditions. WWHTBT populates the assumption register and also runs inside Pre-Mortems and Scenario Mapping branches.
Wiring test Diagnostic / heuristic · Ch 9: A boundary test for Structure versus Culture: If everyone had perfect psychological safety, would information still fail to flow? If yes, the problem is structural; if no, it is cultural.