Annex A · The Diagnostic Instrument
Seventeen questions to place an organization on the two axes
This book is diagnostic and not prescriptive by design: it says where an organization stands and what standing there means, and leaves prescription to the consulting work that builds on that diagnosis. This annex is the practical consequence of that decision — the instrument with which the diagnosis is taken.
It condenses into applicable form what the IRIS and MOTOR chapters develop with their full rationale: the diagnostic questions of every stage and every level, the rule for placing yourself, and the reading of the pair. It does not replace those chapters — the why of each stage lives there — but it allows them to be used without having been read end to end.
For whom? For an organization that wants to place itself, and for the consultant who needs to place a client. How long does it take? Between fifteen minutes, answered honestly and from memory, and a two-hour session with evidence on the table and several areas represented. The second produces a defensible diagnosis; the first, a useful suspicion.
On its use. This annex is published to be used — by an organization on itself, or by a consultant with their clients, paid work included. The only expectation is attribution: that the diagnosis say where it comes from.
How is it applied?
The two axes are answered separately. They are orthogonal: an organization can know a lot and do little, or the reverse. Measuring them together, or averaging them, erases precisely the finding the pair exists to produce.
Each axis is walked in order, from its first question. Levels cannot be skipped: each one rests on the foundations of the previous. An organization operating at a level without having settled the ones before it is not ahead — it is exposed. And when answers differ by area — or by the dimension being looked at: architecture, governance, culture — the lowest one rules: the chapters develop this, and the reason is that the lagging dimension is the real bottleneck.
Your position is the last question answered “yes” without hesitation — and the first question of each axis is read the other way around. Stage 1 of knowing and level 1 of doing do not describe an installed capability but a state one leaves: that is why their questions probe the exit, and a firm “no” to question 1 is the diagnosis — the walk of that axis ends there. From question 2 on, every “yes” with evidence advances the position; the first “more or less”, “in some areas” or “we’re working on it” marks the border, and that is where the diagnosis lies. Any yes that comes after that border is an island, not a stage: a capability without the foundations that hold it up. Islands are recorded — because they name exposure, not position — and they are the first thing the reading of the pair has to explain.
A “yes” demands evidence, not intention. If it cannot be shown working today, it is a no. Maturity models are not ruined by severity but by optimism, and optimism has a recognizable signature: answering for what the organization can do on its best day, or for what one area does and the others do not.
The KNOWING axis — how does information flow?
IRIS model: ten stages, from fragmented data to a self-managed intelligence ecosystem. Each stage is developed, with its observable characteristics, in the IRIS chapter.
| # | Stage | Diagnostic question |
|---|---|---|
| 1 | Fragmented | If I ask two people from different areas for the same indicator, will I get the same number? |
| 2 | Systematized | Is there a defined system for capturing the key business transactions, with consistent registration rules? |
| 3 | Centralized | Is there a centralized data warehouse with dashboards that business users consult regularly? |
| 4 | Modeled | Do the key business metrics have a single, governed definition that the whole organization shares? |
Here the foundations end. An organization at stage 4 has its data house in order — and its analytical exploitation still depends on technical intermediaries.
The border — the Quantum Leap. The step from stage 4 to 5 separates two worlds. Up to here, the online enterprise: the data is current, and someone has to look at it, interpret it, and decide. From here on, the real-time enterprise: the system detects, interprets, decides, and acts. It is also where the crossing of the Nadella Line begins — stages 1 through 4 are better tools for the same processes; from stage 5, the processes themselves begin to be transformed.
| # | Stage | Diagnostic question |
|---|---|---|
| 5 | Accessible | Can a manager get an analytical answer that was not pre-built into a dashboard, in minutes instead of weeks? |
| 6 | Contextual | Is the system capable of crossing sources and enriching an answer with relevant context the user did not explicitly ask for? |
| 7 | Proactive | Do business owners receive relevant information before asking for it, with enough context to act? |
| 8 | Bidirectional | Are there agents that execute operational actions autonomously with traceability, within rules that humans define and supervise? |
| 9 | Coordinated | Do agents from different areas of the business communicate with each other to coordinate actions with consistent information? |
| 10 | Orchestrated | Does the information ecosystem evolve autonomously — identifying gaps, improving models, and optimizing flows — with human supervision only at the strategic level? |
Notes for answering this axis. They head off the frequent errors; each one names what counts as evidence.
- Stage 1. Answer it thinking of the official business indicators. A full “yes” — any indicator, any pair of areas — anticipates stage 4; to leave stage 1 it is enough that the official ones agree. If not even those agree, the organization is at stage 1.
- Stage 3. If “data warehouse” means nothing to whoever is answering: is there a central place where the data of every area is consolidated, with dashboards that business people actually consult? The word does not matter — lakehouse, cloud semantic layer, or classic warehouse all count.
- Stage 5. This is the question that decides which side of the Quantum Leap the organization lands on, and so it carries its own guardrail: the “yes” also requires that two managers asking the same question get the same answer. Without that, what exists is a copilot over ungoverned data, not stage 5.
- Stage 6. The evidence is two concrete cases of answers that brought context nobody asked for — not the vendor’s declared capability.
- Stage 7. The evidence is a real alert with its why — “X went down because Y, and that affects Z” — not the automatic email a dashboard sends when a threshold is crossed.
The DOING axis — how much autonomy do the processes have?
MOTOR model: seven levels, from ad-hoc use of AI to orchestrated autonomous execution. It measures who executes the work and with what degree of autonomy, not which tools were bought. Each level is developed in the MOTOR chapter.
| # | Level | Diagnostic question |
|---|---|---|
| 1 | Invisible · Shadow AI | Do you know how many AI tools are in use across the organization, who uses them, and what data is shared with them? |
| 2 | Trusted | Do you have official AI tools with governance, audit trails, and protection of sensitive data in place? |
| 3 | Collaborator | Do employees delegate complete tasks to AI agents — defining the objective and receiving the result — instead of just asking for targeted assistance? |
| 4 | Analyst | Do you have agents operating 24/7, coordinated with one another, consulting a digital model of the system, with human intervention only on exceptions? |
| 5 | Specialist | Do you have specialist agents able to predict events and simulate scenarios using proprietary data that competitors cannot replicate? |
| 6 | Manager | Do you have agents with authority to assign work, manage budgets, or make management decisions backed by a digital model of the entire organization? |
| 7 | Authority | Do you have agents that participate in strategic decisions with a digital model of the complete ecosystem — competition, market, regulation — and long-term future simulation? |
The first question deserves a warning: almost no organization answers yes, and the reason is that AI already came in through the back door. Level 1 does not describe those who do not use AI — it describes those who use it without knowing.
Notes for answering this axis.
- Levels 1 and 2. A “yes” at level 2 presupposes the “yes” at level 1: official tools without an inventory of what is used outside them is governance by halves. The pair “no to 1, yes to 2” is not an island — it is the signal that the 1→2 transition is half crossed, and the diagnosis is level 1.
- Level 2. Audit trails: an auditable record of who used which tool, for what, and with what data.
- Level 4. Digital model of the system: a digital replica of the operation, self-synchronizing, that agents consult as the source of truth. And since the question chains four conditions — 24/7, coordination, digital model, humans only on exceptions — the “yes” requires all four; if one is missing, the level is the previous one.
- Level 5. The evidence for the “yes” is the prediction in operation — the event anticipated, the simulation used to decide — and the years of proprietary data feeding it; the advantage over competitors is not exhibited, it is inferred.
Every jump between levels has a name, and the name works as a diagnostic signal: if an organization says “we’re implementing AI governance”, it is in the 1→2 transition.
| Jump | Capability it demands | From … to |
|---|---|---|
| 1 → 2 | Trust Infrastructure | From risk to governance |
| 2 → 3 | Wingworking | From assistance to delegation |
| 3 → 4 | Agentic Infrastructure · descriptive twin | From delegation to autonomy |
| 4 → 5 | Agent marketplace · predictive twin | From generic to specialized |
| 5 → 6 | Autonomous organization · organizational twin | From capabilities to authority |
| 6 → 7 | Social agentic · ecosystem twin | From intra-organization to ecosystem |
How is the result read?
The diagnosis is not a stage or a level: it is the pair. That is where the finding lives that neither axis delivers alone, and it is what turns two numbers into a leadership conversation.
| The pair | What does it mean? | Where is the bottleneck? |
|---|---|---|
| High knowing · low doing | Impeccable dashboards and bottled-up decisions. The organization sees and does not act: the information arrives, and someone still has to look, interpret, and order. | On the doing axis — governance before delegation. Data has stopped being the excuse. |
| High doing · low knowing | Blind automation: fast processes over data nobody governs. It is the pair that produces damage fastest, because it scales errors efficiently. | On knowing: without stage 4 — single, governed definitions — every autonomous process scales errors instead of results. |
| Both low | The honest starting point of most. It is not bad news: it is the only position from which one can plan without self-deception. | In the foundations, in parallel — stages 1 to 4 of knowing, and on doing, the step from level 1 to 2: governance. No agent pilot before that. |
| Both high | A real-time enterprise in operation. The advantage is no longer technological: it is one of coordination and governance. | In architecture and trust — the terrain of AgencyDomains. |
And one reading worth saying out loud before the committee: a large gap between the two axes almost always names an organizational problem, not a technical one — an area that bought autonomy without governance, or an impeccable data team nobody bought the action from.
What gets recorded fits in three lines: the coordinate pair, the question where each axis stopped, and the islands — the loose yeses further up, which do not change the position but name where the organization is exposed.
The baseline — the four measures
The pair places you; these four measures quantify what is at stake. They come from the first book of the trilogy — The Real-Time Enterprise, where each corresponds to one of the four faces — and can start being measured on Monday, even by hand. Their value is twofold: they give the baseline against which progress will be measured, and the current numbers are usually the most persuasive argument available in front of a leadership team.
| Measure | How is it taken by hand? |
|---|---|
| How long a business question takes to become an answer | Over the last five real analytical questions that reached the data team: from when they were asked to when they were answered. |
| Hours of uninterrupted work protected, per person per day | A handful of people log one week of their deep-work blocks without interruption. The average tends to be uncomfortable. |
| The fraction of leadership time that is information transport | Over last week’s calendar of one executive: which meetings and reports existed only to move information from one place to another. |
| The hour the day truly ends | From the same people who log the second measure: the timestamp of their last work message of the day, not the time of leaving the office. |
What follows the diagnosis?
The result opens three paths depending on where the bottleneck landed, and each has its instrument in this same book:
- If the problem is data governance — who rules over the data when the consumers are agents and not people —: the Data Canon chapter.
- If it is not known how the organization really operates, because the real processes live in people’s heads and not in the documentation: Wingmap, which reconstructs them from the digital traces the operation already produces, without interviews.
- If the diagnosis must become a plan: the Use Cases chapter — from diagnosis to prioritized portfolio, with the beneficiary framework and the two canonical starting strategies, illustrated with the inventory in Annex B.
What none of the three delivers is the will to cross. That is built in the leadership conversation, and that is what the first book of the trilogy is for.