The ACCCA™ Theory

About this text

This text sets out the theory of the ACCCA™ framework in a form that can be taught, studied and applied. It brings two sources together. The first is the published ACCCA™ whitepaper (Volume 1, Interactive Edition), which presents the framework as five interdependent organisational capabilities, a capability architecture, a value equation and a five-level maturity model. The second is the ACCCA™ Body of Knowledge development series (LAB-ACCCA-THEORY-001 to LAB-ACCCA-RESEARCH-004 and LAB-ACCCA-METHOD-005), which supplies the philosophy, formal construct definitions, theoretical propositions and practitioner standards.

The structure follows the whitepaper's chapter sequence, so that a reader who has seen the public framework will recognise the route. Each chapter then goes considerably deeper, adding the definitions, evidence, reasoning, examples and exercises needed for formal teaching.

Who this text is for

The primary audience is professionals who must lead, advise on or govern organisational AI adoption: senior managers, public-sector leaders, consultants, transformation and change professionals, risk and governance staff, and educators. It assumes no technical background in machine learning. Where technical concepts are needed, they are explained in plain language.

The text also serves three secondary audiences:

• Learners on LABOIA™ ACCCA™ Foundation and Practitioner programmes, for whom it is the core reading.
• Trainers, who can use each chapter as the basis for a teaching session. Learning outcomes, suggested timings and activities are provided.
• Researchers, who will find the formal constructs, propositions and research model in Chapter 13 and in the appendices.

How each chapter is organised

Each chapter follows a consistent teaching pattern, so that learners know what to expect and trainers can plan sessions predictably.

A note on evidence and claims

ACCCA™ is a proprietary framework at an early stage of formal validation. This text therefore observes a strict discipline about what it claims. Three categories are kept distinct throughout.

• Source-grounded elements are those established in LABOIA™'s existing ACCCA™ materials: the five capabilities, their order, the no-shortcuts principle, the link between Conviction and evidence, the cross-functional nature of Collaboration, the Implement → Track → Test → Refine cycle and the feedback from Action to renewed Awareness.
• Proposed elements are developments introduced in the Body of Knowledge series or in this text: detailed indicators, maturity descriptors, the self-assessment scoring rules, the balance rule and the eleven theoretical propositions. They are coherent and useful for teaching but are not yet validated.
• External evidence is research from outside LABOIA™. It is cited where it supports or qualifies the argument. It shows that the problems ACCCA™ addresses are real. It does not show that ACCCA™ itself solves them; that is a question for the research programme described in Chapter 13.

Learners should adopt the same language. ACCCA™ proposes, hypothesises and is designed to; it does not yet prove or guarantee.

Suggested programme structure

The text supports roughly 14 to 16 hours of structured learning, plus independent reading. One possible allocation is shown below; trainers should adjust it to the cohort.

Why ACCCA™

Most AI initiatives don't fail for technical reasons

Suggested teaching time · 60 minutes

1.1 The paradox of access

Organisations have never had easier access to artificial intelligence. Capable models are available through subscriptions, cloud platforms and features embedded in software that organisations already use. A small professional-services firm can now use tools that, a few years ago, only the largest technology companies could build.

Yet access has not translated evenly into value. Two organisations of similar size, in the same sector, with licences for the same tools, can produce very different results. One integrates AI into a handful of well-chosen workflows, measures the effect and steadily improves. The other runs a series of enthusiastic pilots, none of which survives contact with day-to-day operations.

This is the paradox from which ACCCA™ begins. If technology were the decisive factor, organisations with similar technology would achieve similar outcomes. They do not. Something else is doing the work.

ACCCA™ proposes that the "something else" is, to a significant degree, organisational capability: the ability of an organisation to recognise where AI matters, understand what it requires, build justified confidence, coordinate the right people and act in a governed, measurable way.

1.2 What the evidence suggests

The gap between adoption and value is not merely anecdotal. Several large surveys published in 2024 and 2025 point in the same direction, although each uses different definitions and samples and none should be treated as definitive.

Read together, these studies support a cautious but important conclusion: the binding constraint on AI value in most organisations is not the availability of technology. It lies in how organisations decide, coordinate, redesign work, govern risk and learn.

1.3 Access is not readiness

The ACCCA™ Body of Knowledge draws a sharp distinction between access to AI and readiness to use AI well. Access is a matter of procurement. Readiness is a matter of organisational condition. Before meaningful adoption can occur, an organisation typically needs credible answers to questions such as:

• Is our data sufficiently reliable for the use we have in mind?
• Is leadership aligned on what AI is for in this organisation?
• Are our governance and cybersecurity arrangements adequate?
• Do employees have the capabilities to use, question and supervise AI outputs?
• Are the proposed use cases commercially or operationally worthwhile?
• If a pilot succeeds, can it be governed, integrated and scaled?

None of these questions is answered by buying a licence. Each requires organisational work, and each sits outside the narrow boundary of "the technology project".

1.4 Why AI initiatives stall: the non-technical failure modes

The whitepaper's opening claim is that most AI initiatives do not fail for technical reasons. In practice, failure tends to arise from a small number of recurring organisational patterns. Each corresponds to a weakness in one of the five ACCCA™ capabilities, which is why the framework is structured as it is.

Two further patterns cut across all five capabilities:

• Uneven adoption. Different departments move at different speeds. One function becomes a heavy user while another is barely aware of the opportunity. Without coordination, this creates inconsistent risk exposure and makes enterprise value difficult to realise.
• Misplaced trust. Some staff over-rely on AI outputs they do not understand; others reject useful tools out of fear or unfamiliarity. Both are failures of calibrated trust, discussed in Chapter 6.

1.5 Why existing approaches are necessary but insufficient

Organisations rarely approach AI with no framework at all. Most rely on one or more of three established lenses. Each is valuable, and ACCCA™ does not replace any of them. Each, however, leaves an important gap.

The technology lens focuses on platforms, models, architecture and integration. It answers the question can we build or buy this? It does not answer should we, and will the organisation use it well?

The data lens focuses on data quality, availability and infrastructure. Good data is a prerequisite for most AI value. But an organisation with excellent data can still pursue the wrong use case, fail to secure commitment or neglect the people whose work will change.

The compliance lens focuses on legal and regulatory obligations, such as the EU AI Act, data-protection law and sector rules. Compliance sets boundaries that must be respected. Yet a fully compliant organisation can still create little value, and compliance frameworks are not designed to tell an organisation where to invest.

ACCCA™ adds a fourth lens: organisational capability. It asks whether the organisation can move, as a whole, from recognising an opportunity to realising measurable value, and whether it can do so repeatedly.

1.6 The question ACCCA™ addresses

The ACCCA™ Theory states the central organisational challenge as follows:

The framework's answer, developed through the rest of this text, is that the organisation must build five interdependent capabilities, broadly in sequence, and keep renewing them through learning:

Awareness → Comprehension → Conviction → Collaboration → Action → renewed Awareness

Rather than beginning with a technology purchase, an isolated pilot or a compliance checklist, ACCCA™ begins with the organisation: what it needs, what it understands, what it can justify, who must be involved and how it will act and learn.

1.7 Where ACCCA™ sits in the LABOIA™ knowledge architecture

ACCCA™ is designed to be more than a consulting process. LABOIA™ has defined a six-level knowledge architecture so that the framework can be researched, taught and applied, with each level kept distinct.

A compact way to remember the division of labour is:

• ACCCA™ Theory explains.
• ACCCA™ Methodology guides.
• L-ARM™ diagnoses.
• Implementation applies.
• Certification develops and recognises competence.

This text concentrates on the first two levels, with enough of the evaluation level to make the theory usable.

Common misconceptions

• "Our AI problem is a technology problem." Sometimes it is. More often, the evidence suggests, the technology works and the organisation is not yet able to use it well.
• "More pilots will get us there." Pilots without baselines, ownership or a route to operations tend to multiply rather than mature. The MIT and McKinsey findings describe exactly this pattern.
• "Compliance equals readiness." Compliance is a floor, not a strategy.
• "ACCCA™ is anti-technology." It is not. It places organisational reasoning before technology selection so that technology investment is better directed.

The Central Proposition

Technology enables change. People and organisations create value.

Suggested teaching time · 60 minutes

2.1 The proposition

The whitepaper states the central proposition in a single line:

The proposition makes two claims. The first is modest and widely accepted: technology is an enabler. It creates possibilities that did not exist before. The second is the distinctive one: those possibilities become value only through organisational action. Value is created when people redesign work, make better decisions, serve customers differently and govern the results. The technology does not do this by itself.

This has a practical consequence for where organisations invest their attention. If value is created by organisations, then the development of organisational capability is not a "soft" add-on to a technology programme. It is the main event.

2.2 From proposition to philosophy

The ACCCA™ Philosophy (LAB-ACCCA-PHIL-002) translates the central proposition into a set of commitments about how AI adoption should be approached. It begins with a simple principle: AI transformation should begin with understanding before technology.

Technology can enable transformation. But technology alone does not define:

• the problem worth solving;
• the organisational need;
• the value sought;
• the level of readiness;
• the evidence required;
• the people who must be involved;
• the risks that must be governed; or
• the outcomes by which success should be judged.

Each of these must be established by the organisation. ACCCA™ therefore places organisational reasoning before technological intervention.

AI is a means, not the objective

The purpose of AI adoption should not be to deploy AI because AI is available. It should be to improve something that matters, such as decision-making, productivity, service delivery, customer experience, resilience or measurable business value. LABOIA™ explicitly distinguishes its approach from "AI for the sake of AI". The value of an AI initiative should be judged by the organisational value it creates, not by the novelty of the technology it uses.

Transformation begins with reality, not aspiration

Ambition matters, but ambition without an accurate understanding of the present leads to poorly targeted intervention. The first capability, Awareness, therefore reflects a commitment to organisational honesty. It asks what the organisation does now, what works, what does not, and whether AI is genuinely relevant.

Diagnose before prescribing

Recognising a problem does not explain it. The question should not begin as "Which AI tool should we buy?" The earlier questions are: What problem are we addressing? Why does it exist? What evidence supports that diagnosis? What conditions must change? Only then should technology selection become central.

Evidence should precede conviction

An organisation can understand AI conceptually without being justified in committing resources to a particular initiative. ACCCA™ rejects both blind enthusiasm and unsupported resistance. Commitment should be proportionate to evidence. Conviction is not certainty; it is justified confidence sufficient for responsible action.

Trust should be earned

ACCCA™ does not assume that trust in AI should simply increase. Trust should be earned through evidence and appropriate controls, and the right level of reliance should vary with risk, task, consequence of error, available evidence and the presence of human oversight.

People remain central

AI transformation is an organisational and human system, not merely a technical project. Its success may depend on leaders, employees, domain experts, technology and data teams, risk, cybersecurity and legal functions, customers, suppliers and partners. Technology participates in a system of human relationships; it does not replace the need for them.

Responsibility cannot be outsourced to AI

Even where an AI system contributes significantly to a decision, accountability remains with the organisation. Someone must still have authorised the use, own the process, monitor performance, evaluate harm, be able to intervene and answer for outcomes. AI can support responsibility; it cannot eliminate organisational accountability.

2.3 The twelve philosophical principles

The philosophy is summarised in twelve principles. They are the reference point for every later chapter and should be learned.

Several supporting principles help apply the twelve in practice.

• No shortcuts. Each capability prepares the conditions for the next. This does not mean every organisation spends equal time on each. It means later-stage activity cannot compensate for missing foundations. Speed should not be confused with readiness.
• Proportionality. A low-risk productivity tool and a high-impact decision system should not be treated identically. The strength of controls should reflect the significance of the decision, risk and potential impact.
• Evidence over assumption. Organisations should distinguish fact, evidence, assumption, interpretation, hypothesis and aspiration. What is believed should not automatically be treated as what is known.
• Capability before dependency. Organisations should build enough internal capability to understand, challenge, govern and evaluate AI, including when working with vendors. External expertise is valuable; abdicating understanding to it is not.
• Failure as evidence. A well-designed experiment that fails can still create value if it shows why an approach is unsuitable or which investment should not be scaled. The distinction is between learning from controlled failure and repeating avoidable failure without measurement.
• Prove before scaling; reassess while scaling. A successful pilot does not automatically justify organisation-wide deployment, which brings new requirements for infrastructure, governance, security, workforce capability and support.

2.4 Human-centred value

ACCCA™ supports human-centred value creation. This phrase is easily misunderstood, so the philosophy defines it carefully. Human-centred does not mean human-only, and it does not mean that automation should always be minimised. It means the role of people should be deliberately considered rather than accidentally displaced by technology-first implementation.

The design question therefore becomes: What relationship between human capability and AI capability best serves the intended organisational outcome?

AI can augment judgement, automate routine work, accelerate analysis, generate content, assist decisions or restructure workflows. ACCCA™ does not prescribe one universal configuration. Its principle is: automate deliberately, augment deliberately, retain human judgement deliberately. Decisions about where to place the boundary should consider task complexity, risk, consequence of error, explainability, regulatory requirements, ethical concerns, available expertise and required accountability.

2.5 Six dimensions of value

LABOIA™ materials consistently emphasise measurable business value. The philosophy retains that commercial discipline while recognising that value has several dimensions. An AI initiative that delivers economic value by creating unacceptable harm has not delivered value in the ACCCA™ sense.

The guiding rule is: value should be measurable where possible and considered in context.

2.6 Responsibility and governance as enablers

Two of the most practically important positions in the philosophy concern responsibility and governance.

First, responsible AI is not an afterthought. It should shape the entire adoption pathway rather than being added after implementation.

Second, governance enables action. Governance is often perceived as a brake. ACCCA™ takes the opposite view. Without clarity about ownership, acceptable use, accountability, data, risk, monitoring, escalation, security and decision rights, organisations tend either to act recklessly or to avoid useful innovation because uncertainty is too high. Appropriate governance reduces that uncertainty and allows the organisation to act with confidence.

2.7 What ACCCA™ does not endorse

Clarity about what a framework rejects is as useful as clarity about what it supports. ACCCA™ does not philosophically endorse:

• AI adoption primarily for appearance or market signalling;
• technology-first diagnosis;
• unsupported claims of AI capability;
• scaling without evidence;
• implementation without accountability;
• unmeasured deployment;
• blind reliance on AI outputs;
• exclusion of relevant stakeholders; or
• treating AI transformation as a one-time technical installation.

2.8 The ACCCA™ philosophical statement

The following controlled statement is proposed for formal LABOIA™ use and is suitable for quotation in programme materials:

For public-facing use, the concise form is: Understand before adopting. Evidence before scaling. People, responsibility and measurable value throughout.

2.9 The philosophical test

Any proposed AI initiative, training course, assessment tool or implementation service can be tested against ten questions. If the answer to several is "no", the activity is probably inconsistent with ACCCA™.

1. Does it begin with a genuine organisational need?
2. Does it distinguish diagnosis from solution selection?
3. Is the proposed action supported by appropriate evidence?
4. Are the relevant people and organisational functions involved?
5. Are responsibilities and accountability clear?
6. Are risks and governance proportionate to the context?
7. Is the intended value defined?
8. Can outcomes be measured?
9. Can the organisation learn from implementation?
10. Does the activity preserve the logic of Awareness → Comprehension → Conviction → Collaboration → Action?

Common misconceptions

• "Human-centred means keeping humans in every loop." It means deciding deliberately where human judgement belongs, which may mean more automation in some tasks and less in others.
• "Governance slows us down." Poorly designed governance can. Absent governance usually slows organisations more, through uncertainty, rework and incidents.
• "Value means ROI." Return on investment matters, but ACCCA™ recognises six dimensions of value, including responsible value.

The Model: Five Interdependent Capabilities

One framework, five capabilities, one learning cycle

Suggested teaching time · 60 minutes

3.1 The five capabilities at a glance

ACCCA™ names five organisational capabilities. The order of the letters is deliberate: each capability prepares the conditions for the next.

3.2 Why "capabilities" and not "phases"

Earlier LABOIA™ materials describe ACCCA™ as a "five-stage transformation framework". The whitepaper describes "five interdependent capabilities". Both descriptions are correct, and this text uses them together. The word capability is preferred for three reasons.

First, capabilities persist. A phase is completed and left behind. A capability is something an organisation has to a greater or lesser degree, and can lose. An organisation that was once strongly aware of its AI opportunities can become complacent; one that once had strong cross-functional collaboration can see it erode after a reorganisation.

Second, capabilities apply at several levels at once. A large organisation may be at Action on one use case (a mature document-processing system), at Conviction on another (a pilot of AI-assisted forecasting) and at Awareness on a third (whether AI is relevant to procurement at all). A phase model struggles to describe this. A capability model handles it naturally.

Third, capabilities are what the evidence points to. Chapter 1 showed that the main constraints on AI value are organisational. Framing ACCCA™ as capability development directs attention to what the organisation must become able to do, not merely to the activities it must complete.

3.3 Sequence and interdependence: reconciling the two views

The most important conceptual point in this chapter is how ACCCA™ can be both sequential and interdependent. Learners frequently assume these are contradictory. They are not.

The sequential view comes from the source principle "No shortcuts: each stage prepares the conditions for the next." Within any single initiative, there is a logical dependency. You cannot coherently justify an investment (Conviction) in a problem you have not understood (Comprehension), and you cannot understand a problem you have not recognised (Awareness). This is why the methodology uses decision gates (Chapter 9).

The interdependent view comes from the recognition that the five capabilities operate together as a system. Weakness anywhere limits value everywhere, which is the logic of the multiplicative value equation in Chapter 10. Learning in Action feeds back into Awareness. During Action, an organisation may discover that its Comprehension was incomplete, or that its Conviction rested on weak evidence. Movement between capabilities is therefore iterative, not one-way.

ACCCA™ reconciles the two as follows:

A useful analogy is building construction. Foundations must precede walls, and walls must precede the roof; that is the sequence. But a building is only as sound as its weakest structural element, and the inspection of one element can reveal problems in another; that is the interdependence. Nobody concludes from the second fact that foundations can be poured after the roof.

3.4 Construct boundaries

Because the five capabilities are related, they are easily blurred. Precise boundaries are essential for diagnosis, assessment and research. The simplest distinctions are:

• Awareness asks: What matters?
• Comprehension asks: Why is it happening, and what could address it?
• Conviction asks: What evidence justifies commitment?
• Collaboration asks: Who and what must work together?
• Action asks: How do we implement, measure and improve?

The following contrasts are the ones most often confused in practice.

3.5 Two dimensions of Collaboration

The whitepaper defines Collaboration as combining human expertise and AI in coordinated workflows. The Body of Knowledge defines it as coordinating the people, expertise, functions, partners, governance structures and resources required to achieve a shared objective. These are not competing definitions. They describe two dimensions of the same capability, and this text treats both as essential:

• Human–human collaboration: the cross-functional coordination of leadership, business owners, data, technology, risk, legal, security, the workforce and external partners.
• Human–AI collaboration: the deliberate design of workflows in which human judgement and AI capability are combined, with clear allocation of tasks, oversight and accountability.

Chapter 7 develops both.

3.6 The learning cycle

ACCCA™ does not end at Action. Implementation produces new evidence: about what works, what does not, new risks, new opportunities and assumptions that proved wrong. That evidence changes what the organisation recognises, so it feeds back into Awareness.

Awareness → Comprehension → Conviction → Collaboration → Action → Learning → renewed Awareness

The aim is for the organisation to become more capable of learning through adoption, not merely more capable of deploying technology. Each cycle should begin from a higher base of understanding than the last.

3.7 Construct integrity rules

To keep the framework coherent across teaching, assessment and consulting, LABOIA™ has adopted eight integrity rules. Learners should apply them in all written work.

1. Do not use Awareness and Comprehension as synonyms.
2. Do not describe interest or enthusiasm as Conviction unless evidence supports commitment.
3. Do not define Collaboration merely as communication.
4. Do not define Action as launch alone.
5. Do not treat strength in one capability as proof that earlier capabilities are mature.
6. Do not represent proposed indicators as empirically validated until validation has occurred.
7. Preserve the sequence: Awareness → Comprehension → Conviction → Collaboration → Action.
8. Preserve the feedback principle: Action → Awareness.

Common misconceptions

• "ACCCA™ is a waterfall." It is a dependency structure with feedback, not a one-way sequence.
• "We are an Action-stage organisation." Organisations rarely sit in a single stage. Profiles differ by capability and by initiative.
• "Strong Action proves we did the earlier work." Integrity rule 5 forbids this inference. Organisations can implement quickly and still have weak Comprehension, which is how the wrong problem gets automated efficiently.

Awareness

Recognise the gap and where AI is genuinely relevant

Suggested teaching time · 75 minutes

4.1 Definition

The three formulations are complementary. The source definition locates Awareness in the performance gap. The formal definition adds materiality (the gap must matter) and relevance (AI may or may not help). The whitepaper formulation reminds us that Awareness includes limits and risks, not only opportunities.

4.2 Purpose and central question

Awareness establishes whether there is a meaningful organisational reason to investigate AI-related change. Its most important function is protective: it prevents technology selection from becoming the starting point.

The central question is: What should the organisation recognise? In practitioner form: Where is AI genuinely relevant, and where is it not?

The second half of that question matters. Genuine Awareness includes the ability to say "AI is not the answer here". An organisation that sees AI opportunities everywhere is not highly aware; it is uncritical.

4.3 What Awareness includes

Awareness involves recognising:

• current performance and current workflows;
• pain points, constraints and decision bottlenecks;
• missed opportunities and capability gaps;
• strategic pressures from competitors, customers, regulators or funders;
• emerging opportunities and relevant AI possibilities;
• known limitations of AI and of the organisation;
• risk exposure, including the risk of not acting; and
• plausible future states.

Note the balance. Roughly half the list concerns the organisation's present reality; the other half concerns possibility. Awareness is the capability that connects the two.

4.4 What Awareness does not mean

Awareness should not be equated with:

• detailed technical knowledge of AI;
• selecting an AI solution;
• understanding root causes (that is Comprehension);
• approving investment or proving business value (that is Conviction);
• implementing a pilot (that belongs to later capabilities); or
• declaring the organisation "AI-ready" (that is a matter for L-ARM™ diagnosis).

A common pattern is the executive briefing that leaves leaders excited about AI's potential. This may contribute to Awareness, but on its own it usually produces general interest rather than recognition of a specific, material gap in this organisation.

4.5 Observable indicators and evidence

How can an assessor tell whether Awareness is present? An organisation demonstrating Awareness is typically able to:

• describe its current-state challenges concretely;
• identify relevant opportunities and articulate a desired future state;
• distinguish priority needs from general interest in AI;
• identify where evidence is missing;
• recognise that AI may not suit every issue; and
• describe its organisational constraints.

These are developmental indicators, not validated measures. Assessors should look for convergence across several of them rather than relying on any one.

4.6 The Awareness failure condition

LABOIA™ materials present this as a reasoned implication rather than an empirically measured failure rate. It is nonetheless a familiar pattern. It is visible whenever an organisation's AI strategy consists mainly of a list of tools, or when a team is asked to "find uses for" a platform that has already been licensed.

4.7 Entry, exit and decision gate

Entry condition. The organisation has interest in AI, pressure to change, an identified problem or a potential opportunity. The bar for entry is deliberately low; Awareness is where organisations begin.

Exit condition. The organisation can describe the current state, the potential state, why the gap matters and why further investigation is justified.

Decision gate. Progress to Comprehension only when the problem or opportunity is sufficiently clear, its organisational significance is understood and further analysis is justified. Output: a defined transformation question.

4.8 Practitioner test and interventions

The practitioner test for Awareness asks: Is the organisation sufficiently clear about what matters? If it cannot yet establish the relevant problem, the current–potential gap, why AI might be relevant, where it might not be, and the strategic significance of the issue, it should remain primarily at Awareness.

Interventions appropriate to Awareness aim to clarify what matters:

• executive awareness sessions grounded in the organisation's own strategy;
• strategic discovery and opportunity workshops;
• current-state analysis and baseline measurement;
• AI literacy for leaders and key staff; and
• an initial L-ARM™ readiness assessment.

4.9 Worked example

Common misconceptions

• "We are aware; everyone is talking about AI." General interest is not recognition of a specific, material gap.
• "Awareness is just for beginners." Mature organisations need renewed Awareness constantly, fed by learning from Action.
• "Awareness should conclude that AI is relevant." A finding that AI is not the right lever is a valid and valuable Awareness outcome.

Comprehension

Understand the causes and the levers before choosing a solution

Suggested teaching time · 75 minutes

5.1 Definition

The whitepaper formulation emphasises understanding of AI; the Body of Knowledge emphasises understanding of the problem. Comprehension requires both. An organisation that understands its problem but not the technology may choose an unsuitable tool. An organisation that understands the technology but not its problem will automate the wrong thing.

5.2 Purpose and central question

Comprehension exists to prevent premature prescription. It converts recognition into structured understanding. The source framework is explicit: Comprehension guards against adopting a solution before the diagnosis is right.

The central question is: What do we need to understand before deciding what to do?

A useful summary of the logic is: Symptom → Causes → Levers → Direction.

5.3 The domains of Comprehension

Comprehension spans several domains. Not all will be equally important for every initiative; proportionality applies.

The last row deserves emphasis. Comprehension must include the possibility that AI is not the best lever. Many problems first framed as AI opportunities are better solved by fixing a form, a hand-off or a policy.

5.4 What decision-makers need to understand about AI

The whitepaper stresses understanding how AI works, where it fits, and how to govern and interpret it. Decision-makers do not need to be engineers, but Comprehension requires a working grasp of several characteristics of contemporary AI systems.

• Probabilistic outputs. Most modern AI systems produce outputs that are likely rather than guaranteed to be correct. Generative systems can produce fluent, confident and wrong content (often called "hallucination").
• Dependence on data. Systems reflect the data on which they were trained or to which they have access, including its gaps and biases.
• Performance varies by context. A tool that performs well on general tasks may perform poorly on specialist, local or unusual cases.
• Limited explainability. It can be difficult to explain why a system produced a particular output, which matters greatly where decisions affect people's rights or welfare.
• Drift. Performance can change over time as data, users or the underlying model change.
• New security and privacy risks. Examples include confidential information entered into external tools and manipulation of system behaviour through crafted inputs (prompt injection).

These characteristics are why interpretation and governance belong within Comprehension. An organisation that does not understand them cannot set appropriate levels of trust (Conviction) or design sensible human oversight (Collaboration).

5.5 What Comprehension does not mean

Comprehension should not be equated with:

• general AI awareness or enthusiasm;
• executive sponsorship;
• purchasing technology;
• confidence without evidence;
• implementation; or
• organisational alignment.

5.6 Observable indicators and evidence

An organisation demonstrating Comprehension can typically:

• distinguish symptoms from root causes;
• explain relevant AI capabilities and limitations for the specific use;
• identify data dependencies and articulate the major risks;
• compare alternative interventions, including non-AI options;
• understand the governance requirements that apply;
• identify capability requirements; and
• explain why a particular direction may or may not be appropriate.

5.7 The Comprehension failure condition

This failure is particularly costly because it can look like success. A project can be delivered on time and on budget and still automate a broken process. The planning department in Chapter 4 could have bought a tool to read incomplete applications faster. Comprehension would have revealed that the root cause was the submission form.

5.8 Entry, exit and decision gate

Entry condition. Awareness has established a meaningful gap or opportunity worthy of investigation.

Exit condition. The organisation can explain why the issue exists, which conditions influence it, what options are available, what risks matter and what direction appears defensible.

Decision gate. Progress to Conviction only when the major causes are understood, material constraints are visible, relevant risks have been considered and plausible intervention directions exist. Output: a defensible direction.

The practitioner test asks: Does the organisation understand why the issue exists and what conditions influence it? If understanding is weak, the instruction is simple: do not prescribe prematurely.

Interventions appropriate to Comprehension aim to develop defensible understanding: process and root-cause analysis, use-case analysis, feasibility review, data assessment, governance assessment and risk assessment.

5.9 Worked example

Common misconceptions

• "We understand AI; we've all used ChatGPT." Personal familiarity with a general tool is not organisational understanding of a specific use, its data, its risks and its governance.
• "Comprehension is the IT team's job." Technical understanding is one domain among ten. Process, people, risk and governance understanding usually sit elsewhere.
• "Analysis delays value." Proportionate analysis prevents the much larger delay of implementing the wrong thing.

Conviction

Justified confidence, built on evidence

Suggested teaching time · 75 minutes

6.1 Definition

6.2 Purpose and central question

Conviction bridges understanding and organisational commitment. It exists to ensure that commitment is based on evidence rather than excitement, fear or unsupported assumptions.

The central question is: What evidence justifies moving forward?

The logic is: Understanding → Evidence → Commitment.

Conviction is the capability most often counterfeited. Senior enthusiasm, vendor demonstrations, competitor anxiety and media coverage can all generate the feeling of conviction without its substance. ACCCA™ insists on the qualifier justified.

6.3 Conviction is not certainty

Conviction should not be equated with blind optimism, certainty, senior enthusiasm alone, vendor claims, general belief in AI, formal implementation or organisation-wide adoption.

The philosophy rejects two extremes: blind enthusiasm and unsupported resistance. Neither is an evidence-led position. An organisation that refuses a well-evidenced, well-governed use case out of generalised fear has failed at Conviction just as surely as one that approves a poorly evidenced one out of excitement.

6.4 Trust, calibration and the human side of Conviction

The whitepaper links Conviction to trust. This connects ACCCA™ to an established body of research on how people rely on automated systems.

For ACCCA™, this means Conviction operates at two levels:

• Organisational conviction: the leadership's justified decision to commit resources to an initiative.
• User-level calibrated trust: the extent to which the people who will work with the system rely on it appropriately, neither blindly nor dismissively.

Both are built the same way: through transparency about what the system does and does not do, governance that sets limits and oversight, evidence of performance in this context, and ethical consideration of who could be affected. These are precisely the four foundations named in the whitepaper's definition.

6.5 Forms of evidence

Not all evidence is equal. Learners should be able to judge the relative strength of evidence offered in support of an AI investment.

The general principle is that evidence generated in the organisation's own context, against a recorded baseline, with defined measures, is stronger than evidence generated elsewhere. This is why the controlled pilot is central to Conviction.

6.6 Designing a pilot that produces evidence

Many pilots fail to generate Conviction because they were designed to demonstrate a tool, not to test a hypothesis. A pilot designed for Conviction has the following features.

1. A benefit hypothesis. A clear statement of what is expected to improve, by how much and for whom. For example: "AI-assisted drafting will reduce average time to produce a client summary by at least 30% without increasing the error rate."
2. A baseline. Measurement of current performance before the pilot begins.
3. Defined measures. KPIs for value (time, cost, quality), adoption (usage, satisfaction) and risk (errors, incidents, complaints).
4. Decision criteria agreed in advance. What result would justify proceeding, redesigning or stopping? Agreeing these before seeing the results guards against post-hoc rationalisation.
5. Governance controls. Approved data, acceptable-use rules, human review requirements and a named owner.
6. A representative setting. Real users, real data (lawfully used) and real workflow, not only enthusiasts in ideal conditions.
7. Bounded scope and duration. Long enough to see past novelty, short enough to limit exposure.

6.7 Indicators and evidence

An organisation demonstrating Conviction may have agreed priority use cases, defined expected outcomes, decision criteria, committed sponsors, pilot approval, agreed risk tolerances, evidence supporting investment and defined measurement methods.

6.8 The Conviction failure condition

Weak Conviction shows up in two opposite ways. In one, initiatives are approved without sufficient evidence and are later abandoned when early difficulties expose the absence of real commitment. In the other, good initiatives never secure commitment because nobody has assembled evidence that decision-makers trust. ACCCA™ addresses both.

6.9 Entry, exit and decision gate

Entry condition. The organisation has sufficient Comprehension to identify one or more plausible intervention directions.

Exit condition. The organisation possesses sufficient evidence and commitment to justify mobilisation.

Decision gate. Progress to Collaboration only when evidence supports the use case, expected outcomes are defined, success can be measured, principal risks are acceptable or manageable and sufficient sponsorship exists. Output: evidence-based commitment.

The practitioner test asks: Is there sufficient evidence to justify commitment?

Interventions appropriate to Conviction aim to create evidence sufficient for responsible commitment: business-case development, controlled pilots, proofs of concept, KPI design, benefit measurement, governance controls and agreed decision criteria.

6.10 Worked example

Common misconceptions

• "The CEO is committed, so we have Conviction." Sponsorship is necessary but not sufficient. Without evidence it is fragile.
• "A successful proof of concept proves value." A proof of concept usually proves technical feasibility, not organisational value.
• "Conviction means removing all doubt." It means doubt proportionate to the risk, and a commitment structured to learn.

Collaboration

No transformation succeeds in isolation

Suggested teaching time · 75 minutes

7.1 Definition

7.2 Purpose and central question

Collaboration converts justified commitment into coordinated organisational capability.

The central question is: Who and what must work together for this to succeed?

The word what is important. It includes people and functions. It also includes the AI system itself, as a participant in redesigned workflows. This is why Collaboration has two dimensions.

7.3 Dimension one: human–human collaboration

AI initiatives cross organisational boundaries. A typical initiative touches strategy, operations, data, technology, risk, legal, security, human resources and, often, external suppliers. When any of these is missing, predictable problems follow.

Collaboration is not simply holding meetings, communicating a project, asking for feedback once, assigning the whole initiative to IT, using an external vendor or obtaining executive approval. The distinguishing concept is coordinated contribution toward a shared objective, which requires named owners, defined responsibilities and agreed decision rights.

7.4 Dimension two: human–AI collaboration

The whitepaper's emphasis on combining human expertise and AI in coordinated workflows reflects the fact that AI adoption usually changes how work is done, not only which tool is used. The McKinsey finding in Chapter 1, that high performers were far more likely to have fundamentally redesigned workflows, points in the same direction.

Good human–AI workflow design addresses five questions.

1. Allocation. Which tasks will the AI perform, which will people perform, and which will be shared? The allocation should follow the philosophy's principle: automate deliberately, augment deliberately, retain human judgement deliberately.
2. Oversight. Where must a human review, approve or be able to override an AI output? Oversight should be proportionate to the consequence of error.
3. Competence. Do the people exercising oversight have the knowledge, time and authority to do so meaningfully? Oversight by someone who cannot realistically detect an error is nominal, not real.
4. Feedback. How will people report errors, edge cases and improvements, and how will those reports change the system or the workflow?
5. Accountability. Who answers for the outcome of the combined human–AI process?

7.5 Indicators and evidence

An organisation demonstrating Collaboration may have named owners, defined responsibilities, cross-functional participation, agreed decision rights, communication and escalation structures, shared objectives, implementation governance, resource commitments and coordinated partner involvement.

7.6 The Collaboration failure condition

The hypotheses in Chapter 13 illustrate how Collaboration interacts with other capabilities. Strong Conviction with weak Collaboration may produce commitment without delivery. Strong Collaboration with weak Comprehension may produce well-coordinated implementation of the wrong intervention.

7.7 Entry, exit and decision gate

Entry condition. The organisation possesses sufficient Conviction to mobilise people and resources.

Exit condition. The required people, functions and capabilities are sufficiently coordinated to support responsible implementation.

Decision gate. Progress to Action only when ownership is clear, required functions are involved, resources are available, governance is defined and roles and decision rights are understood. Output: mobilised organisational capability.

The practitioner test asks: Are the right people, capabilities and decision structures aligned?

Interventions appropriate to Collaboration aim to create coordinated organisational capability: governance model design, steering groups, role design, RACI, stakeholder mobilisation, workforce capability development, vendor alignment and operating-model design.

7.8 Worked example

Common misconceptions

• "Collaboration means consultation." Consultation is input. Collaboration is coordinated contribution with ownership.
• "IT owns AI." IT owns technology delivery. Business owners own outcomes.
• "A human in the loop guarantees safety." Only if that human is competent, has time, has authority and is actively checking.

Action

Implement, track, test, refine, and learn

Suggested teaching time · 75 minutes

8.1 Definition

8.2 Purpose and central question

Action converts organisational capability into measurable operational change. The central question is: How will the organisation implement, govern, measure and improve the intervention?

Action should not be equated with buying software, launching a pilot, switching a system on, completing a project milestone or declaring transformation complete. Action is explicitly a managed cycle.

8.3 The Action cycle

Implement. Embed the intervention in real operations: integrate it with systems and workflows, train users, put governance controls into operation and confirm ownership. Implementation should proceed in proportion to evidence. A cautious, staged roll-out is often wiser than a "big bang".

Track. Monitor the agreed measures continuously or at defined intervals: usage, performance, value, incidents and user feedback. Tracking turns operation into evidence.

Test. Examine whether the assumptions behind the initiative still hold. Is the benefit hypothesis being realised? Has performance drifted? Are there unintended consequences for staff, customers or particular groups? Are the controls working in practice?

Refine. Act on what tracking and testing reveal. Improve prompts, configurations, workflows, training, controls or the scope of use. Then decide: continue, adjust, scale or stop.

8.4 Designing the measurement framework

Philosophical principle 8 states: measure what you implement. Action without measurement provides weak evidence for future decisions. A good measurement framework covers four families of measure.

Measures should connect to the benefit hypothesis agreed at Conviction and be compared with the baseline recorded then. Without that continuity, the organisation cannot tell whether the initiative worked.

8.5 Responsible scaling

Philosophical principle 11 states: scale should follow evidence. Scaling an AI initiative is not simply doing more of the same. It creates new requirements for infrastructure, governance, cybersecurity, workforce capability, change management, integration, support, monitoring and financial sustainability. A pilot that worked with 20 motivated users may behave very differently with 2,000 ordinary ones.

The Action decision gate therefore reads: continue or scale only when implementation is operating, measurement exists, governance is functioning, learning is being captured and evidence supports continued investment. Output: measured operational capability.

Scaling may also require a return to earlier capabilities. A new department may need its own Awareness and Comprehension work. New functions may need to join the Collaboration. This is not failure. It is the interdependence described in Chapter 3.

8.6 Stopping well

The stop rule applies throughout Action. A practitioner may recommend continuation, redesign, pause, contraction or discontinuation. Stopping an initiative that is not creating value is a sign of capability, not weakness. The philosophy's principle of failure as evidence applies. A well-measured initiative that is stopped leaves behind knowledge that informs future decisions. A poorly measured one leaves behind only cost.

8.7 Indicators and evidence

An organisation demonstrating Action may operate the AI intervention in practice, monitor defined metrics, track exceptions and failures, evaluate outcomes, maintain governance controls, capture user feedback, refine processes, update controls, review business value, and scale or stop based on evidence.

8.8 The Action failure condition

This failure is common because organisational project structures reward delivery. Projects close when systems go live. The team disbands. Nobody owns the question of whether the system is still creating value a year later. ACCCA™ treats this as a capability gap, and the remedy is to assign continuing ownership of measurement and improvement.

8.9 From Action to renewed Awareness

Action has no permanent exit condition. Instead, it generates:

• performance evidence;
• implementation learning;
• new risks;
• new opportunities; and
• renewed organisational knowledge.

These feed back into renewed Awareness. The organisation now recognises things it could not see before: an adjacent process that could benefit, a risk not previously understood, an assumption that proved wrong. Each cycle should begin from a richer base than the last.

8.10 Worked example

Common misconceptions

• "Go-live is the finish line." Go-live is the start of the Action cycle.
• "If nobody complains, it's working." Absence of complaint is not evidence of value. Measure against the baseline.
• "Stopping an initiative is failure." Stopping on evidence is part of the capability.

Capability Architecture

How the system works

Suggested teaching time · 60 minutes

9.1 The architecture

The whitepaper presents ACCCA™ as a system with inputs, a capability core, an output and a feedback loop.

Reading from top to bottom:

1. The external environment creates pressure and possibility: technological change, competitors, customers, regulation, funding conditions and labour markets.
2. Organisational leadership interprets that environment, sets direction, allocates resources and establishes the conditions in which the capabilities can develop.
3. The five capabilities convert direction into results: Awareness, Comprehension, Conviction, Collaboration and Action.
4. Measurable organisational value is the output, across the six dimensions of value described in Chapter 2.
5. Continuous learning carries evidence from Action back into renewed Awareness.

9.2 The external environment

The capabilities do not develop in a vacuum. The external environment shapes what the organisation must recognise and how quickly. Relevant forces include:

• technology: the pace and direction of AI development, and the availability of tools;
• markets: what competitors and customers are doing and expecting;
• regulation: for example the EU AI Act's risk-based obligations, data-protection law and sector-specific rules;
• labour markets: the availability of skills, and workforce expectations;
• society: public trust in AI and expectations of responsible use.

The environment also moderates the capabilities. In a highly regulated sector such as healthcare or financial services, Conviction and Action require stronger governance evidence than in a low-risk internal productivity use. This is the proportionality principle at work.

9.3 Organisational leadership

Leadership sits between the environment and the capabilities because it performs three functions that no other part of the organisation can.

• Sense-making. Leaders interpret external signals and decide which matter. This shapes Awareness.
• Resource allocation. Leaders commit budget, people and attention. This shapes Conviction and Collaboration.
• Climate setting. Leaders establish whether it is safe to question, experiment, report failure and stop initiatives. This shapes every capability, and especially the feedback loop.

Leadership is not itself one of the five capabilities, but every capability depends on it. An organisation whose leaders treat AI as a technology procurement matter will struggle to develop Collaboration. One whose leaders punish failed experiments will struggle to learn from Action.

9.4 Transition logic

Each transition between capabilities has a distinct logic:

9.5 The five decision gates

The gates translate the no-shortcuts principle into practice. Each gate has explicit conditions and a defined output.

9.6 Four practitioner rules

The Methodology (LAB-ACCCA-METHOD-005) adds four rules that govern how the gates are applied.

The transition rule. Progression should be based on evidence, not on activity. Completing a workshop does not by itself justify moving to the next capability. Progress when sufficient evidence exists that the central question of the current capability has been addressed to a level appropriate to the organisational context and risk.

The no-shortcuts rule.

• Do not skip Awareness if the problem is poorly defined.
• Do not skip Comprehension if causes and requirements are unclear.
• Do not skip Conviction if value remains unsupported.
• Do not skip Collaboration if ownership and capability remain fragmented.
• Do not treat Action as completion if measurement and learning are absent.

The stop rule. ACCCA™ explicitly permits a recommendation of no AI intervention, further investigation, pause, redesign or discontinuation. This protects the framework from becoming biased towards implementation.

The escalation rule. Practitioners should escalate to appropriate specialists where evidence is materially contradictory, regulatory exposure is uncertain, cybersecurity risk is material, legal interpretation is required, high-impact automated decision-making is contemplated or serious ethical concerns exist. ACCCA™ provides a structure for judgement; it does not replace specialist expertise.

9.7 Proportionality in the gates

The gates do not require the same depth of evidence for every initiative. A team trialling an AI writing assistant for internal meeting notes may pass all five gates in a few weeks with light documentation. A bank deploying AI in credit decisions will require extensive evidence at every gate, specialist escalation and regulatory engagement. The gates are constant; the evidence threshold scales with risk and impact.

9.8 ACCCA™ and L-ARM™

LABOIA™ maintains two complementary frameworks, and learners must keep them distinct.

L-ARM™ is LABOIA™'s readiness assessment model. It currently assesses twelve organisational AI-readiness dimensions and produces an evidence-led readiness profile. ACCCA™ interprets that evidence in capability terms and guides intervention. Their integration logic is:

L-ARM™ evidence → ACCCA™ capability interpretation → appropriate intervention → implementation evidence → reassessment

L-ARM™ should not be renamed or presented as an ACCCA™ assessment. The two perform different roles. Neither should be used to justify predetermined technology sales.

9.9 Intervention fit

A practical implication of the architecture is that interventions should fit the organisation's current capability profile. An organisation whose primary gap is Awareness should not receive the same engagement as one whose gap is Collaboration.

Common misconceptions

• "Gates mean bureaucracy." Gates scale with risk. For low-risk uses they can be light and fast.
• "L-ARM™ and ACCCA™ are the same thing." L-ARM™ diagnoses readiness; ACCCA™ structures progression.
• "Leadership sponsorship is enough." Leadership creates the conditions; the capabilities still have to be built.

The ACCCA™ Value Equation

One equation, one insight

Suggested teaching time · 45 minutes

10.1 The equation

The whitepaper summarises the ACCCA™ argument in a single expression:

10.2 What the terms mean

The mapping is approximate rather than exact. Each capability contributes to more than one term. Collaboration, for instance, also builds trust among the people whose work changes. The mapping is nonetheless useful for teaching because it shows how the five capabilities combine into value.

10.3 Why multiplication matters

The equation's insight lies in its operator. If the terms were added, strength in one could compensate for weakness in another: excellent technology could make up for poor execution. Because they are multiplied, no term can compensate for the absence of another. A zero anywhere produces zero overall.

This expresses, in compact form, three ideas developed elsewhere in the text:

• the no-shortcuts principle: later strength cannot make up for missing foundations (Chapter 3);
• the complementarity hypothesis (H8): balanced capability profiles should outperform highly uneven ones (Chapter 13);
• the productivity J-curve: technology produces value only alongside complementary organisational investment (Chapter 2).

It also explains a strategic pattern that is otherwise puzzling. Organisations often respond to disappointing AI results by upgrading the technology. If the binding constraint is trust or execution, this produces little improvement, because the weakest term is limiting the product.

10.4 A worked illustration

Suppose, purely for illustration, that each term is scored on a scale from 0 to 1.

Three lessons follow. Organisation B, with less advanced technology, produces almost four times the value of Organisation A. Upgrading Organisation A's technology to perfection barely changes its result. Doubling its trust, however, doubles its value. The highest-return investment is usually in the weakest term.

10.5 Limits of the equation

The value equation is a conceptual heuristic: a compact way of expressing the theory's central insight. It is not a validated measurement model. Learners and practitioners should observe the following limits.

• The terms are not yet operationally defined in a way that allows reliable, comparable scoring.
• The multiplicative form is a hypothesis. Real relationships may involve thresholds, diminishing returns or partial compensation between terms. The research programme in Chapter 13 includes alternative models for this reason.
• Value is multi-dimensional (Chapter 2). A single "AI Value" figure simplifies what is in practice a profile of economic, operational, human, customer, strategic and responsible value.
• Other factors matter. Market conditions, regulation, timing and luck all affect outcomes and are not represented.

Used as a thinking tool, for example asking "which term is closest to zero for us?", the equation is powerful. Used as a calculator, it would overstate what is currently known.

Common misconceptions

• "We can calculate our AI value with the equation." Not yet. It is a heuristic, not a validated metric.
• "Better technology always increases value." Only if the other terms are non-trivial. The weakest term limits the result.
• "Trust is soft and unmeasurable." Trust can be evidenced, through calibrated reliance, governance confidence and user behaviour, even if it is not yet scored on a validated scale.

Organisational Maturity

Five levels, one trajectory

Suggested teaching time · 60 minutes

11.1 Why a maturity model

A maturity model describes a typical trajectory of development. It helps an organisation benchmark where it stands, see what the next level looks like and prioritise the work needed to get there. The ACCCA™ maturity model describes the journey from fragmented experimentation to AI as a core, value-creating capability.

11.2 The five levels

11.3 Capability-level descriptors

Overall maturity summarises a more detailed picture. Each of the five capabilities can be at a different level. The following descriptors, proposed for teaching and assessment, describe what each capability typically looks like at each level.

11.4 Reading maturity correctly: profile over score

A single overall level conceals as much as it reveals. Two organisations at "Integrated" overall can have very different profiles: one balanced, one strong in Action but weak in Comprehension. Following the value-equation logic in Chapter 10, the second is likely to be more fragile, because it may be efficiently executing poorly diagnosed initiatives.

For this reason, ACCCA™ maturity should always be reported as an overall level and a capability profile, never the overall level alone.

11.5 Uses and limits

Maturity assessment is useful for:

• benchmarking an organisation's starting position;
• prioritising the capability that most limits progress;
• communicating with boards and stakeholders in a common language; and
• tracking development over time.

It also has limits that learners must respect.

• Maturity is not a goal in itself. Not every organisation needs to be Transformational. A small firm may sensibly aim for Integrated in a few processes.
• Maturity can vary by unit and initiative. An organisation-wide level is an average across real variation.
• The levels and descriptors are proposed, not validated. They are a structured basis for professional judgement, not a psychometric instrument.
• Self-assessment overstates maturity. Organisations tend to rate themselves on intentions and policies rather than evidenced practice. Chapter 12 discusses how to counter this.

Common misconceptions

• "We are Level 3, so we are fine." The profile matters. A weak capability can undermine an apparently healthy average.
• "Everyone should aim for Level 5." Target maturity should follow strategy and context.
• "Having a policy means we are Integrated." Maturity is evidenced by practice, not by documents alone.

Self-Assessment and Diagnosis

Where does your organisation stand today?

Suggested teaching time · 60 minutes

12.1 Purpose

The ACCCA™ self-assessment provides a candid starting point, not a formal audit. It helps leaders form a working estimate of their maturity, start a structured conversation and identify where a deeper assessment would be valuable. It consists of ten statements, two for each capability.

12.2 The instrument

For each statement, choose the score that best reflects evidenced practice in your organisation, not intentions or policies alone.

12.3 Scoring

Step 1: Capability scores. Add the two scores for each capability. Each capability score ranges from 2 to 10.

Step 2: Overall score. Add all ten scores. The overall score ranges from 10 to 50.

Step 3: Provisional maturity level.

Step 4: Capability levels and balance check. Convert each capability score to a level: 2–3 Emerging, 4–5 Developing, 6–7 Integrated, 8–9 Optimised, 10 Transformational. Apply the balance rule from Chapter 11: overall maturity should be read as no more than one level above the weakest capability.

Step 5: Priority. The capability with the lowest score is normally the primary priority. Where two are tied, prioritise the earlier one in the sequence, following the no-shortcuts principle.

The interactive web edition of this text includes a version of the self-assessment that performs these calculations automatically.

12.4 Why self-assessment is not enough

Self-assessment has well-known weaknesses. Respondents tend to rate intentions rather than practice. Senior leaders often see a more favourable picture than front-line staff. Different respondents interpret the same statement differently. The ACCCA™ construct assessment rule therefore states that no capability should be assessed solely through a single self-reported question.

A full ACCCA™ assessment combines, where appropriate:

• stakeholder interviews at several levels of the organisation;
• documentary evidence (strategies, policies, business cases, governance records);
• operational evidence (usage data, KPIs, incident logs);
• observed practice;
• quantitative data; and
• qualitative professional judgement.

Asking several respondents from different levels and functions to complete the self-assessment independently, then comparing results, is a simple and revealing first step. Large differences between groups are themselves a finding.

12.5 The evidence discipline

ACCCA™ assessment distinguishes five things that are often run together.

This discipline protects both the organisation and the practitioner from unsupported scoring. It also mirrors the philosophical principle: what is believed should not automatically be treated as what is known.

12.6 From diagnosis to intervention

The purpose of assessment is to inform a decision. A practical sequence is:

1. Locate. Identify the dominant capability position, the strongest and weakest capabilities and the transition barriers.
2. Prioritise. Classify issues as critical foundation, high-value progression, developmental, monitor or not currently justified.
3. Design. Select interventions that fit the primary gap (Chapter 9, Section 9.9).
4. Agree measures. Define how progress will be recognised.
5. Reassess. Repeat the assessment after a defined period to track change.

Common misconceptions

• "A high self-assessment score means we are ready." Self-assessment measures perception. Evidence may tell a different story.
• "The lowest score is a failure." It is the most useful finding, because it identifies where investment will have most effect.
• "One respondent is enough." Differences between respondents are often the most informative result.

Theoretical Propositions and Research Model

From framework to evidence

Suggested teaching time · 75 minutes

13.1 Why a framework needs a theory

A framework tells practitioners what to do. A theory explains why doing it should work, and does so in a way that can be tested. ACCCA™ aspires to be both. The ACCCA™ Theoretical Propositions and Research Model (LAB-ACCCA-RESEARCH-004) sets out the theory in researchable form. This chapter teaches it.

13.2 The research problem and question

Ten supporting research questions follow. They ask, for example, whether each capability is associated with the next (RQ1–RQ5), whether weaknesses constrain later capabilities (RQ6), whether ACCCA™ explains outcomes after controlling for technology access (RQ7), whether relationships vary by sector, size or regulation (RQ8), whether the constructs are empirically distinct (RQ9), and whether ACCCA™ progression predicts better outcomes than technology-first implementation (RQ10).

13.3 The core proposition

P0 is a research proposition, not an established fact. The eleven propositions below specify it.

13.4 The eleven propositions

Proposition 7 can be illustrated by the predicted consequences of weakness at each point:

• Low Awareness → a poorly defined problem.
• Low Comprehension → an inappropriate intervention.
• Low Conviction → weak commitment.
• Low Collaboration → fragmented implementation.
• Weak Action discipline → insufficient organisational learning.

Proposition 8 can be illustrated by uneven profiles:

• Strong Conviction with weak Collaboration may produce commitment without delivery.
• Strong Collaboration with weak Comprehension may produce coordinated implementation of the wrong intervention.
• Strong Action with weak measurement may produce activity without learning.

13.5 Mediators and moderators

Two technical terms are essential for reading the research model.

Proposed mediators

Proposed moderators

13.6 The research model

The preliminary ACCCA™ Research Model v0.1 runs from organisational context, through L-ARM™ readiness, through the five capabilities (with evidence quality and resource commitment as mediators), to organisational AI outcomes (value, responsible adoption and sustainable capability), then to organisational learning and back to renewed Awareness.

The primary dependent variable is not "AI adopted: yes or no". It is Organisational AI Adoption Effectiveness, a multi-dimensional construct that includes operational integration, measurable value, adoption sustainability, governance quality, responsible use, learning capability and appropriate scale. Secondary outcomes include operational, strategic, human, governance, financial and learning outcomes.

The primary unit of analysis is the organisation, although future research may examine business units, project teams or ecosystems. Researchers must specify the unit and level of analysis and avoid mixing levels without justification.

13.7 Connections to established theory

ACCCA™ did not emerge in a theoretical vacuum. LABOIA™ source materials identify connections with change management, technology adoption, organisational learning, dynamic capabilities, sociotechnical systems, AI governance and digital transformation. The connections below are offered as starting points for a formal literature review, not as established equivalences.

The diffusion-of-innovations parallel deserves comment because learners often notice it. Rogers' model describes how an individual or decision unit comes to adopt an innovation. ACCCA™ describes how an organisation develops the capabilities to adopt AI repeatedly and well. It adds two elements that Rogers' individual model does not centre: cross-functional Collaboration, and governed, measured Action that feeds back into renewed Awareness. Whether these additions improve explanation is exactly what the research programme should test.

13.8 Falsifiability and alternative models

A useful theory must be capable of being shown wrong. ACCCA™ should be revised if evidence repeatedly shows that:

• the five constructs cannot be empirically distinguished;
• Awareness has no meaningful relationship with Comprehension;
• Comprehension does not influence Conviction;
• Collaboration does not influence implementation quality;
• sequence does not matter;
• Action does not create meaningful learning;
• ACCCA™ adds no explanatory value beyond existing readiness measures; or
• the model consistently performs worse than simpler alternatives.

A serious research programme should therefore test ACCCA™ against plausible alternatives: a fully sequential model; a partially sequential model in which some capabilities influence several later ones; a parallel-capability model in which all five operate concurrently; a higher-order model in which the five are dimensions of one overarching capability; and a configurational model in which different combinations produce similar outcomes. Testing alternatives reduces confirmation bias. Failure to support a proposition should be treated as evidence for refinement, not concealed.

Boundary conditions also need investigation. ACCCA™ may apply differently in very small organisations, in individual AI use without organisational integration, in emergency implementation, in highly centralised organisations, in highly regulated automated-decision settings and in technology companies with unusually mature AI capability.

13.9 The research programme

The proposed programme moves through eight research maturity stages: conceptual → exploratory → measurement → explanatory → predictive → longitudinal → replication → theory refinement. In practical terms:

1. Conceptual validation: expert panels, Delphi studies and academic review of definitions and relationships.
2. Content validity: expert rating of measurement items; cognitive interviews.
3. Exploratory study: exploratory factor analysis to test whether items cluster into five constructs.
4. Confirmatory study: confirmatory factor analysis and structural equation modelling on independent data.
5. Predictive testing: do ACCCA™ scores predict meaningful outcomes?
6. Longitudinal validation: do changes in capability precede changes in outcomes over time (baseline, post-intervention, operational, review)?

Because ACCCA™ makes claims about progression and learning, longitudinal and mixed-method designs are particularly important. Cross-sectional surveys can show association; only designs that follow organisations over time can show sequence and change. Qualitative case studies can explain how the mechanisms operate.

Research ethics and independence are integral. Studies should address consent, confidentiality, data minimisation, conflicts of interest and transparency about LABOIA™'s ownership of ACCCA™. Credibility will ultimately depend on research conducted or reviewed by parties who do not develop or sell the framework, including universities, doctoral researchers and peer-reviewed publication.

13.10 Current evidence status and claim discipline

Until validation is completed, the following language is appropriate: "ACCCA™ proposes…", "The framework hypothesises…", "The model is designed to…", "Research will test whether…". The following should be avoided: "ACCCA™ proves…", "Research shows…", "ACCCA™ guarantees…", "The five capabilities cause successful AI adoption…".

Common misconceptions

• "A framework used by many clients is validated." Adoption is not validation. Only systematic evidence validates.
• "Finding that a proposition is not supported damages ACCCA™." It improves the theory. Concealing it would damage credibility.
• "ACCCA™ is just Rogers' model renamed." There are clear parallels, but the level of analysis and emphases differ. The research programme should test whether the differences add explanatory value.

Integrated Case Study: Meridian Services Ltd

Applying all five capabilities to one organisation

Suggested teaching time · 90 minutes

14.1 The case

Meridian Services Ltd is a medium-sized professional-services company. Management believes AI could improve internal productivity. Employees spend significant time preparing client summaries, locating internal information and drafting routine communications.

The CEO has announced that the company should become "AI-enabled within twelve months".

The company currently has:

• Microsoft cloud systems;
• a large collection of internal documents;
• limited formal AI training;
• no approved AI-use policy; and
• inconsistent document-management practices.

A small innovation team tested a generative-AI assistant. The pilot team reported that it was useful. However:

• no formal baseline was recorded;
• confidential information was entered during testing;
• legal and cybersecurity teams were not involved;
• no agreed business KPIs were established; and
• several employees are concerned about the impact on their jobs.

The CEO wants to deploy the tool throughout the organisation immediately.

14.2 Case questions

Work through the following questions individually or in groups before reading the guidance.

1. Awareness. What organisational problems or opportunities can you identify? Separate known issues from assumptions.
2. Comprehension. What does Meridian still need to understand? Consider data, workflow, security, governance, employees and business value.
3. Conviction. Does Meridian currently have enough evidence to justify organisation-wide deployment? Identify evidence supporting progression and evidence still missing.
4. Collaboration. Which functions should participate before wider deployment, and why?
5. Action. Assume Meridian eventually proceeds. Identify three measures, two risks to monitor and one review point.
6. Overall judgement. Which ACCCA™ capability should receive primary attention now? Support your answer with evidence from the case.
7. Value equation. Which term of the value equation is currently closest to zero?
8. Maturity. Using the descriptors in Chapter 11, estimate Meridian's capability profile and apply the balance rule.

Applying ACCCA™ Responsibly

Turning AI ambition into organisational capability

Suggested teaching time · 30 minutes

15.1 Practitioner conduct

ACCCA™ will only be as credible as the people who apply it. The Methodology sets out several standards that every learner should adopt.

• Independence from technology sales. ACCCA™ and L-ARM™ must not be used as a justification for predetermined technology recommendations. A practitioner who always concludes "buy this product" is not applying ACCCA™.
• Evidence over assertion. Judgements should rest on evidence, with the evidence discipline applied: claim, evidence, interpretation, judgement and action.
• Professional judgement. ACCCA™ structures judgement; it does not replace it. Context, proportionality and experience remain essential.
• Specialist boundaries. ACCCA™ practitioners are not, by virtue of the framework, lawyers, security specialists or data scientists. The escalation rule applies.
• Honesty about the framework's status. Practitioners should describe ACCCA™ accurately: a structured, source-grounded framework whose detailed indicators, maturity levels and value equation are not yet validated.

15.2 What ACCCA™ is, and is not

15.3 ACCCA™ in one page

For executive audiences, the theory can be summarised as follows.

15.4 Closing reflection

Artificial intelligence creates new capabilities. It does not remove the need for sound organisational judgement. The ACCCA™ theory asks organisations not simply "What can AI do?" but "What should we do, why should we do it, what evidence supports it, who must be involved, and how will we know whether it created value?"

The technology will change, quickly and repeatedly. The organisational discipline of recognising, understanding, justifying, coordinating, acting and learning should remain.

Awareness · Comprehension · Conviction · Collaboration · Action

From AI ambition to responsible, measurable organisational value.

Formal Master Definitions

These are the controlled ACCCA™ definitions for use in teaching, assessment and research.

Glossary

Summary of Hypotheses

All eleven are proposed hypotheses and remain subject to empirical testing.

References

• Argyris, C. and Schön, D. A. (1978) Organizational Learning: A Theory of Action Perspective. Reading, MA: Addison-Wesley.
• Boston Consulting Group (2024) Where's the Value in AI? Press release: "AI Adoption in 2024: 74% of Companies Struggle to Achieve and Scale Value", 24 October 2024.
• Brynjolfsson, E., Rock, D. and Syverson, C. (2021) "The Productivity J-Curve: How Intangibles Complement General Purpose Technologies", American Economic Journal: Macroeconomics, 13(1), pp. 333–372.
• Davis, F. D. (1989) "Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology", MIS Quarterly, 13(3), pp. 319–340.
• Edmondson, A. (1999) "Psychological Safety and Learning Behavior in Work Teams", Administrative Science Quarterly, 44(2), pp. 350–383.
• European Union (2024) Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), Official Journal of the European Union.
• ISO/IEC (2023) ISO/IEC 42001:2023 Information technology — Artificial intelligence — Management system. Geneva: ISO.
• Kolb, D. A. (1984) Experiential Learning: Experience as the Source of Learning and Development. Englewood Cliffs, NJ: Prentice Hall.
• Kotter, J. P. (1996) Leading Change. Boston, MA: Harvard Business School Press.
• LABOIA™ (2026) The ACCCA™ Framework: A Practical Methodology for Responsible AI Adoption. Volume 1, Interactive Edition. Available at: https://accca.laboiaeurope.ie/
• LABOIA™ (2026) ACCCA™ Body of Knowledge development series: LAB-ACCCA-THEORY-001; LAB-ACCCA-PHIL-002; LAB-ACCCA-CONSTRUCT-003; LAB-ACCCA-RESEARCH-004; LAB-ACCCA-METHOD-005. Version 0.1, controlled development drafts.
• Lee, J. D. and See, K. A. (2004) "Trust in Automation: Designing for Appropriate Reliance", Human Factors, 46(1), pp. 50–80.
• McKinsey & Company (2025) The State of AI in 2025: Agents, Innovation, and Transformation. QuantumBlack, AI by McKinsey, November 2025.
• MIT NANDA (2025) The GenAI Divide: State of AI in Business 2025. As reported in Fortune, 18 August 2025.
• National Institute of Standards and Technology (2023) Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1. Gaithersburg, MD: NIST.
• OECD (2019, updated 2024) Recommendation of the Council on Artificial Intelligence (OECD AI Principles). Paris: OECD.
• Parasuraman, R. and Riley, V. (1997) "Humans and Automation: Use, Misuse, Disuse, Abuse", Human Factors, 39(2), pp. 230–253.
• Rogers, E. M. (2003) Diffusion of Innovations. 5th edn. New York: Free Press.
• Teece, D. J. (2007) "Explicating Dynamic Capabilities: The Nature and Microfoundations of (Sustainable) Enterprise Performance", Strategic Management Journal, 28(13), pp. 1319–1350.
• Trist, E. L. and Bamforth, K. W. (1951) "Some Social and Psychological Consequences of the Longwall Method of Coal-Getting", Human Relations, 4(1), pp. 3–38.
• Venkatesh, V., Morris, M. G., Davis, G. B. and Davis, F. D. (2003) "User Acceptance of Information Technology: Toward a Unified View", MIS Quarterly, 27(3), pp. 425–478.