--- chapter: 02 part: Map title: The 9Q Grid subtitle: The Map Before the Journey page: page-0042 readingMinutes: 14 published: 2026-05-20 status: draft-ready version: racecraft-v1 --- ## The Map Before the Journey Every race begins with a map. Not a destination. Not a plan. A map. Before Red Bull designs a car, they study the track layout. Bahrain has long straights and heavy braking zones. Monaco is tight corners and no overtaking. Singapore is street circuits under lights. Spa is elevation changes and unpredictable weather. The same car won't win everywhere. The same setup won't work on every track. **Strategy begins with knowing the terrain.** Most organisations approach AI like a driver who shows up without studying the circuit. They have capability. They have budget. They have ambition. But they don't know where they are actually racing. "We need an AI strategy" isn't a strategy. It's a wish. A strategy requires a map. A map that shows you not just where you are, but where you could be. Not just what AI can do, but what kind of work it is doing, what degree of agency it holds, and what that implies for trust, governance, and value creation. In Helix language, this is **portfolio positioning**: where are you now, where should you be next, and what sequence of moves gets you there without pretending every AI bet carries the same risk or return? **The 9Q Grid is that map.** In the last chapter we met the Co-Intelligent Org and the question "Who does the thinking?". The 9Q Grid is how we turn that question into a map you can actually draw. --- ## One Grammar, Two Dimensions Here's what took me years to understand. There is a pattern that appears everywhere in human-AI collaboration. It shows up in how customers experience AI. It shows up in how teams do work. It shows up in how organisations mature. **The same pattern. Three stages. One grammar.** **ASSIST** means the AI helps while the human leads. The machine accelerates the work, but the work keeps the same shape. **AUGMENT** means human and AI iterate together. The machine narrows options, surfaces patterns, drafts possibilities. The human judges, refines, and decides. **ADAPT** means the AI acts while the human governs. The human sets boundaries, purpose, and escalation rules; the machine executes inside them. That grammar is the vertical axis of the 9Q Grid. But the second insight is what makes the grid really useful: AI does not just vary by capability. It also varies by the **kind of work** it is doing. Some work is structured. Some is dynamic. Some is genuinely complex. When you cross those two dimensions, nine distinct archetypes emerge. --- ## The Two Axes ### Axis 1: Agency The vertical axis asks: **How much agency does the AI hold in the work?** | Level | Name | Meaning | Human role | | :--- | :--- | :--- | :--- | | 1 | **Assist** | AI helps, human leads | **At the helm** | | 2 | **Augment** | Human and AI iterate together | **In the loop** | | 3 | **Adapt** | AI acts, human governs | **On call** | This is not a ladder of "better." It is a ladder of agency. The higher you go, the more trust the organisation must spend and the more governance it must have already earned. In Part 4, you will see what this grammar looks like running at 300 kilometres an hour. ### Axis 2: Work The horizontal axis asks: **What kind of work is this?** | Level | Name | Meaning | | :--- | :--- | :--- | | 1 | **Structured** | Defined, repeatable, rules-heavy work | | 2 | **Dynamic** | Iterative, changing, feedback-rich work | | 3 | **Complex** | Emergent, ambiguous, hard-to-specify work | This matters because a brilliant AI deployment in structured work is not the same as a brilliant AI deployment in complex work. The governance challenge is different. The value mechanism is different. The failure mode is different. **Agency tells you who acts. Work tells you what kind of terrain they are acting in.** --- ## The Grid The 9Q Grid maps nine archetypes across those two axes.
The Grid Itself Agency × Work · 9 cells
Agency
Structured Defined, repeatable work Dynamic Iterative, feedback-rich work Complex Emergent, ambiguous terrain
Adapt AI acts; human governs Augment Human and AI iterate together Assist AI helps; human leads
Q3 Automator Adapt × Structured Delegate structured work to the machine. Q6 Optimiser Adapt × Dynamic Let AI iterate dynamically toward a target. Q9 · Diagonal Conductor Adapt × Complex Govern adaptive complexity across a field. Q2 Tactician Augment × Structured Make bounded calls inside a fixed frame. Q5 · Diagonal Partner Augment × Dynamic Think together in mutual iteration. Q8 Explorer Augment × Complex Co-search ambiguous futures and novel space. Q1 · Diagonal Operator Assist × Structured Execute a known line of work faster. Q4 Weaver Assist × Dynamic Pull dynamic threads into coherence. Q7 Navigator Assist × Complex Orient in uncertain, emergent terrain.
Work
Each cell is both a human role and a legitimate category of AI solution. The diagonal matters: - **Q1**: coherent assistance in structured work - **Q5**: genuine partnership in dynamic work - **Q9**: governed adaptation in complex systems That diagonal is the clean path of increasing co-intelligence. But most real organisations do not live on the diagonal. They live in the off-diagonal cells: productive mismatches, partial maturity, and bets that reveal what they really value. ---
9Q Grid Assets Spread · Canvas
Diagnostic Canvas
Use the grid while you read.
Open the two-page spread for the full reference view, or download the one-page canvas PDF for portfolio mapping.
Open Spread Download Canvas PDF
--- ## One-Line Guide to the Nine Cells Before we go deeper, here is the fastest way to read the grid: - **Q1 Operator**: Execute a known line of work faster. - **Q2 Tactician**: Make bounded calls inside a fixed frame. - **Q3 Automator**: Delegate structured work to the machine. - **Q4 Weaver**: Pull dynamic threads into coherence. - **Q5 Partner**: Think together in mutual iteration. - **Q6 Optimiser**: Let AI iterate dynamically toward a target. - **Q7 Navigator**: Orient in uncertain, emergent terrain. - **Q8 Explorer**: Co-search ambiguous futures and novel space. - **Q9 Conductor**: Govern adaptive complexity across a field. If that is all a reader remembers from this chapter, they can still use the grid. --- ## The Full Reference Table Use this as the chapter's working reference sheet. The table gives you the vocabulary. The next section gives you the feel of each cell. | Q | Name | Core meaning | Why the word works | Human-AI relationship | Typical AI pattern | Example class | Failure mode | | --- | --- | --- | --- | --- | --- | --- | --- | | Q1 | **Operator** | Execute a known line of work faster | Procedural, grounded, inside an existing system | Human runs the workflow; AI assists | Copilot, summarisation, completion, retrieval | Accountant with Copilot, developer with code completion, clinician with note drafting | Tool dependence | | Q2 | **Tactician** | Make bounded calls inside a fixed frame | Judgment under rules, not top-level direction | Human decides; AI surfaces options, flags, scores | Decision support, anomaly detection, explainable AI | Underwriter with risk model, dispatcher with routing AI, compliance analyst | Recommendation capture | | Q3 | **Automator** | Delegate structured work to the machine | Clear, unsentimental, rule-based delegation | Human sets rules and governs; AI runs | RPA, IDP, rules engines, workflow automation | Claims processing, invoice automation, compliance monitoring | Drift on autopilot | | Q4 | **Weaver** | Pull dynamic threads into coherence | Integrative, compositional, iterative | Human authors and connects; AI generates threads | Drafting, synthesis, ideation, research support | Writer, consultant, policy adviser using GenAI | Loss of thread or voice | | Q5 | **Partner** | Think together in mutual iteration | Plain, strong, reciprocal | Human and AI refine together | Paired analysis, interactive reasoning, iterative drafting | Strategist with model, analyst with simulation partner | Capability atrophy | | Q6 | **Optimiser** | Let AI iterate dynamically toward a target | Says both power and risk | Human governs the objective; AI adjusts continuously | Dynamic pricing, recommendation engines, adaptive routing | Pricing engines, logistics optimisation, algorithmic allocation | Optimising the wrong thing | | Q7 | **Navigator** | Orient in uncertain, emergent terrain | Implies wayfinding, not certainty | Human still finds the way; AI improves the map | Signal detection, scenario support, live modelling | Emergency coordination, policy navigation, outbreak response | Signal overload | | Q8 | **Explorer** | Co-search ambiguous futures and novel space | Broad enough for discovery with a cognitive partner | Human and AI explore possibility space together | Strategic foresight, hypothesis generation, scenario planning | Foresight teams, novel-domain research, scenario design | Endless exploration | | Q9 | **Conductor** | Govern adaptive complexity across a field | Orchestration without pretending to control every move | AI acts across system complexity; human governs atmosphere, thresholds, legitimacy | Multi-agent systems, adaptive orchestration, ecosystem coordination | Autonomous supply chains, adaptive policy systems | Rubber-stamping outputs | *We’ll keep returning to this grid in later parts of the book—when we talk about licensing (governance), machine design (teams and workflows), and the race (operational loops). Keep it in mind as your mental map.* --- ## The Nine Archetypes
Click a cell Agency × Work
Q3 Automator Adapt × Structured AI runs defined work. Q6 Optimiser Adapt × Dynamic AI iterates toward a target. Q9 Conductor Adapt × Complex AI governs adaptive complexity. Q2 Tactician Augment × Structured AI narrows the call. Q5 Partner Augment × Dynamic Human and AI think together. Q8 Explorer Augment × Complex AI expands possible futures. Q1 Operator Assist × Structured AI speeds up known work. Q4 Weaver Assist × Dynamic AI supplies threads. Q7 Navigator Assist × Complex AI makes terrain legible.
### Q1 — The Operator **Assist × Structured** The human performs defined work; AI accelerates known steps. The work is the same shape before and after AI. It just gets faster. Examples: - tab-complete coding - Excel Copilot - OCR and extraction - ambient clinical documentation - suggested responses inside a fixed script **The tell:** task definition existed before AI. The AI changed speed, not shape. **Failure mode:** atrophy. The tool becomes the skill. ### Q2 — The Tactician **Augment × Structured** The human iterates inside a defined frame. AI narrows the option space; the human makes the call. Examples: - fraud detection with human adjudication - loan underwriting AI with sign-off - triage scoring - audit flagging systems - compliance review with judgment gates **The tell:** AI ranks, flags, or narrows. The human clears decisions one by one. **Failure mode:** recommendation capture. The human starts rubber-stamping the machine. ### Q3 — The Automator **Adapt × Structured** AI runs structured work autonomously. The human governs outcomes and exceptions, not every input. Examples: - RPA - workflow automation - invoice processing - onboarding workflows - auto-routing and closure of routine tickets UiPath's My Plan Manager case is a clean Q3 example: OCR, RPA, and a rules engine automate structured claims and accounts-payable work inside a governed process. **The tell:** work runs without human attention most of the time. **Failure mode:** silent drift. The automation keeps doing yesterday's right thing in today's changed world. ### Q4 — The Weaver **Assist × Dynamic** The human weaves together a dynamic output; AI provides threads, fragments, synthesis, draft variants, and research support. Examples: - conversational drafting with Claude or ChatGPT - research synthesis - design exploration - PM draft specs - writing and editing with AI support **The tell:** the final output is unmistakably the human's voice. AI contributed threads; the weave is still human. **Failure mode:** voice erosion. The human lets the AI dictate sequence, tone, and shape. ### Q5 — The Partner **Augment × Dynamic** This is the true co-intelligence cell for most serious knowledge work. Human and AI iterate together. Context deepens. The work accumulates. Neither party is the sole author of the result. Examples: - persistent AI workspaces - long-form strategic thinking with AI - legal research partnerships - pair programming with sustained context - collaborative policy drafting Morgan Stanley is a strong public Q5 case: advisors use AI daily not to replace judgment, but to deepen retrieval, synthesis, and client-facing iteration. **The tell:** the conversation has memory, weight, and return visits. **Failure mode:** dyad dependency. The human stops operating without the AI partner. ### Q6 — The Optimiser **Adapt × Dynamic** AI autonomously iterates on dynamic work. Humans set objectives, thresholds, and what must not be broken. Examples: - dynamic pricing - recommendation engines - real-time logistics optimisation - ad bidding - inventory optimisation - yield management Airline-style dynamic pricing and JB Hi-Fi recommendation systems are strong Q6 patterns: machine-led adjustment toward a live commercial target. **The tell:** the AI makes thousands to millions of decisions at a speed no human could touch. **Failure mode:** optimising the wrong thing at scale. ### Q7 — The Navigator **Assist × Complex** The territory is ambiguous. AI makes the field more legible. The human still makes the meaningful decisions. Examples: - executive briefings from live signal streams - public-health intelligence during outbreaks - geopolitical or disaster dashboards - investigative document triage - horizon scanning NASA's hurricane-response work is a clean Q7 example: AI improves signal visibility and decision speed, but humans still navigate the storm. **The tell:** AI cannot produce the answer. It produces legibility. **Failure mode:** signal drowning. More telemetry than judgment. ### Q8 — The Explorer **Augment × Complex** Human and AI iterate together in genuinely emergent territory. Nobody knows the answer in advance, including the AI. Examples: - scenario planning - war-gaming - deep research - systems modelling - ethnographic pattern exploration - speculative design - novel-domain hypothesis generation BCG's strategic-foresight work fits Q8 well: AI expands the search space of plausible futures before leaders commit. **The tell:** the work is exploratory, not merely convergent. **Failure mode:** simulation capture. The model becomes the work and reality waits. ### Q9 — The Conductor **Adapt × Complex** AI acts autonomously in a complex field. The human governs atmosphere, constraints, purpose, and escalation rules, not individual decisions. Examples: - multi-agent orchestration - adaptive supply chains - self-healing infrastructure - autonomous cyber response - city-scale orchestration systems - research agents coordinating at system scale Microsoft's multi-agent orchestration work in Copilot Studio is one of the clearest public Q9 signals: agents coordinating across systems while humans govern aims and escalation. **The tell:** the human governs conditions, not operations. **Failure mode:** orchestra drift. The system evolves goals nobody consciously signed off on. --- ## Four Patterns Worth Teaching ### 1. Cell volume is inverse to cell maturity Q1 is the most common and most mature. Q9 is the rarest and least mature. Most organisations talk about Q8 and Q9. Most of their real footprint is still Q1, Q3, and Q4. ### 2. The diagonal is a capability path, not a deployment strategy Organisations do not "move to Q9." They operate across multiple cells simultaneously. The question is whether they are honest about which cell a given deployment is actually in. ### 3. Mislabelling is the central failure Q3 sold as Q9.
Q1 sold as Q5.
Q6 sold as "optimisation" when it is really an ungoverned strategic leap. > **"Most AI failure is diagnosis failure before it becomes technology failure."** ### 4. Q5 and Q8 are where co-intelligence is most literally true Q5 is bounded partnership.
Q8 is exploratory partnership. If the book has an aspirational centre of gravity, it lives here. --- ## Reading the Grid as Portfolio Positioning The 9Q Grid is not a ladder. Q9 is not automatically "better" than Q1. A nuclear safety system should probably stay closer to structured, tightly governed territory. A strategy team working through novel futures may need Q8 more than Q6. The point is not to label a company as one box. It is to see where different initiatives sit across the grid at the same time. The grid gives you a way to place bets. More than that, it gives an executive team a shared frame for trading those bets off. Which initiatives belong in low-agency territory?
Which deserve investment to move upward?
Which are too ambitious for your current trust budget?
Which have been sold internally as one thing but are actually another? This is why the 9Q Grid matters. It is not a taxonomy exercise. It is a decision discipline. --- ## A Tale of Two Banks Here’s how the Trust Gradient plays out in practice: same destination quadrant, two very different paths, two very different levels of pain. Let me make this concrete. **Bank A** deploys basic assistants everywhere. Chatbots, summaries, internal copilots, template generators. It becomes good at Q1 and good enough at Q2. It tells the board it is "doing AI." But the portfolio never deepens. The work stays structured. Humans stay overloaded. No part of the operating model changes. **Bank B** starts in similar territory, but treats the grid as a positioning exercise rather than a procurement exercise. It uses Q1 where speed is the point.
It builds Q2 where judgment still matters.
It deliberately builds toward Q5 in the highest-value decision environments.
Only when the trust, data quality, and governance are real does it move portions of the portfolio toward Q6. Both banks have AI. Only one has a portfolio. That is the difference between "doing AI" and using the 9Q Grid properly. --- ## The Trust Gradient The grid is also a trust map. The diagonal from Q1 to Q5 to Q9 is the clean path of increasing co-intelligence. It is also the clean path of increasing trust demand. At Q1, failure is irritating. At Q5, failure distorts decisions. At Q9, failure can destabilise the entire system. This is the new scarcity. You cannot buy the trust required for higher-agency positions. You earn it by proving reliability, building governance, and advancing honestly through the cells you are actually ready for. > **"You cannot buy the trust required for higher-agency positions. You earn it by proving reliability, building governance, and advancing honestly."** Most AI failures happen when organisations try to buy a Q9 outcome with a Q1 trust budget. They do not fail because the technology is bad. > **"They do not fail because the technology is bad. They fail because they are insolvent in the currency of trust."** --- ## Before We Move On Plot your top ten initiatives. Do not ask, "How much AI do we have?" Ask: - Which cell is each initiative actually in? - Which cells are overrepresented? - Where is our current state? - Where should we be next? - Which moves are coherent, and which are fantasy? By the time you finish this chapter, you should not just have a list of AI projects. You should have a portfolio position. *Next: Chapter 3 Strategic Shapes — Why organisations are not points on a map, but constellations that reveal their true strategic shapes.*

Chequered Flag03 takeaways

  1. The 9Q Grid is portfolio positioning. It shows where your initiatives sit, where they should move next, and what kind of work they are really doing.

  2. Mislabelling is the central failure. Most organisations' stories about their AI are more advanced than their actual cells.

  3. The grid is a decision discipline. It helps leadership trade bets off honestly instead of talking about AI as one thing.