Before You Read

The complete system: identity, positioning, voice, visual design, and standards.

01 · Identity

Who setmode is

What setmode feels like. The values, posture, emotional territory and the arc it creates in every encounter.

Calm authority that opens into genuine excitement.

The design is calm. The effect is energising. A CEO dealing with AI noise needs relief first. Relief comes before excitement. The brand earns the excitement by delivering the relief first. Underneath the rigour: a genuine belief that AI changes what organisations can do.

The Five Values

The Anti-Values

How Values Govern Decisions

The Emotional Arc

Every first encounter with setmode moves the reader through the same sequence.

Does this create clarity, build trust and respect the intelligence of the person receiving it?

Yes: on-brand. No: rewrite.

02 · Positioning

What setmode stands for

The belief, the audience, the experience truths, and the associative field.

The Belief

The AI problem inside organisations is an operating model problem. The market treats it as a skills problem or a strategy problem. Both are wrong in the same structural way: they assume that improving the parts improves the system. It does not.

The Paradigm Shifts

The old reality and the new reality. Every positioning decision reinforces the right-hand column.

Old Reality	New Reality
AI as individual productivity tool	AI as organisational operating system
Scattered point solutions	Compounding intelligence architecture
Skills hidden and uneven	Capability distributed across leadership
Knowledge produced by training	Systems produced by doing
AI owned by a champion or an IT team	AI owned collectively: everyone builds, everyone compounds
Strategy separated from execution	Strategy and capability built simultaneously
Productivity gains stay with the individual	Throughput compounds for the organisation
AI initiatives are experiments	AI initiatives become institutional assets
Compliance inconsistent, risk unmanaged	Governance established, risk owned
Start from scratch with every new AI need	Build once, reuse, compound

The Audience

A CEO who already feels the gap between individual AI activity and organisational AI capability, suspects the gap is structural, and does not yet have the language or the architecture to act on that suspicion. Full audience definition, personas and disqualifiers in Section 03.

The Five Experience Truths

What must be true in the experience for the brand to be believed.

The Founding Arc

A CEO-signed declaration of institutional intent at the start, measured against delivery at the close.

The Two Pitches

Two pitches for two audiences. The CEO pitch answers the organisational question. The participant pitch answers the personal question.

The Associative Field

Six brands that share the visual and emotional territory setmode occupies. Each carries a specific quality.

The Seven Visual Principles

Seven principles emerge consistently across all six associative brands. These govern the design system.

setmode Visual Character

The seven principles translated into specific visual direction.

03 · Audience

Who setmode serves

The ideal organisation, the four personas inside every engagement, and the identity arc each travels. Boundaries included.

The Ideal Organisation

Mid-market organisations with a real, cohesive management team that is ready to move from AI experimentation to AI architecture.

Dimension	Criteria
Annual Revenue	$15M – $250M (sweet spot $30M – $150M)
Management Team	15 – 50 (CEO + C-suite + direct reports)
Structure	Cohesive team; CEO holds budget authority
AI Maturity	Past initial adoption, approaching operating model design
Digital Infrastructure	Basic SaaS stack in place; not AI-native
Geography	US, UK, Australia, Canada, Western Europe, Singapore, GCC / UAE

Buying Triggers

Disqualifiers

Organisations to walk away from. The wrong organisation should self-select out before a conversation starts.

Signal	Why It Disqualifies
Revenue under $15M	Management team too lean; insufficient functional depth to compound
No real management team	Programme requires distributed ownership; one person holding all authority cannot scale
Revenue over $300M	Committee buying cycles and procurement timelines prohibit the pace this programme requires
AI awareness at stage zero	Programme is designed for organisations past basic literacy
No CEO sponsorship	Without the economic buyer, the programme stalls at middle management
Seeking tool deployment	The organisation needs a systems integrator. setmode builds operating models.

Sector Prioritisation

Four sectors with active CEO's Playbook positioning. Each has a structural forcing function that makes the AI operating model question urgent now.

The Four Personas

Four distinct personas inhabit every engagement. Each has a different relationship to AI, a different stake in the outcome, and a different arc through the programme. A fifth dimension runs beneath all four: the identity arc from AI user to institutional architect.

	Sponsor	Champion	Functional Leader	Sceptic
Primary motivation	Competitive advantage	Internal recognition	Functional relevance	Avoiding wasted effort
Main fear	Falling behind competitors	Programme failing publicly	Irrelevance in an AI world	Another initiative that does not stick
Key moment	Anchor Workshop 1	Week 1: clean launch	Week 3–4: first use case	Week 9: first prototype
Win condition	Deliverables accumulating; prototype live	Positioned as transformation architect	Owns tools and workflows for their function	Becomes the governance champion

Detailed empathy maps (Think/Feel, Hear, See, Say/Do, Pains, Gains) and facilitation design implications for each persona are maintained internally.

What Good Looks Like

A qualified prospect will say some version of these things:

04 · Voice

How setmode speaks

The writing standards, the four qualities, the three registers, and the controlled vocabulary. These are the rules.

The Two Rules

The Voice: Calm Authority

Authority comes from clear reasoning and grounded evidence. Calm is the absence of effort. If a sentence feels like it is reaching for drama, the writing has become effortful. A CEO encountering this voice feels understood.

The Four Qualities

The Three Registers

Altitude

setmode writes at the level the reader operates. The reader is a CEO. Business outcomes, competitive position and organisational capability. A sentence that opens with a software product name before establishing a business reality has signalled the wrong register. Tools are evidence. They appear inside reasoning to ground a claim. They are never the opening observation.

The test: could this sentence appear in a board paper?

Posture

The reader already knows. A CEO who has initiated a conversation about an AI operating model has already diagnosed the gap. The writing confirms their diagnosis and presents the path through it. The tone is: you see this clearly; here is how we move through it.

Frame gaps as situations. Frame the programme's outcome as accumulation. What each person designs and builds, the organisation retains and compounds. Individuals and institution both grow.

The brand never describes its own rigour, craft, or care. The work demonstrates these. A sentence that asks the reader to notice how well something was built is a sentence that doubts whether they will. The brand does not narrate its own structure, claim its effects on the reader, or name what it aspires to become. It presents. The reader draws conclusions.

Sentence Structure

Two devices govern how setmode sentences land. Both are structural, not decorative. The reader feels the shape before they parse the meaning, and the shape is the argument.

British English

Controlled Vocabulary

Language is structural. This vocabulary is the standard.

The Banned List

Language in Context

05 · Visual System

How setmode looks

The complete design system: colour architecture, typography, spacing, components, and interaction patterns.

Two-Tier Colour Architecture

Canvas

Text

Identity Colour

No chrome accent. Setmode runs three colour systems already (Phase cool slate, Category warm badges, semantic states). A fourth chrome accent would compete with all three for attention. Active states, sidebar highlights, focus rings, breadcrumb current-page indicators and similar interactive states use neutral white at opacity on dark, or neutral warm charcoal at opacity on cream. Mulberry is no longer chrome. Mulberry is the wordmark colour and nothing else.

Phase Accent Colours

Phase Colour Registers

Standard (fills, borders) · Deep (text-safe) · Tint (12% wash backgrounds).

Phase Palette · Cool Slate (Dark Rebuild)

The new Phase palette is cool slate, used in page chrome, breadcrumbs, progress indicators, footer Phase markers. Never as content fills. Cool against the warm canvas creates instrument-vs-room contrast. Three steps per register; Execute is the most settled tone on dark and the most pronounced on cream. The perceptual flip across modes is correct, not a bug.

Legacy warm Phase tokens (--sage, --clay, --slate, --pewter) stay in this section above for backward compatibility while ~943 component selectors migrate. Retirement queues as a separate kill.

Category Palette · Warm Badges (Dark Rebuild)

Seven hues for the 7 Categories. Used as small filled chips on deliverables, filter pills, and content classification. Warm because category names label content; cool was reserved for system signals (Phase indicators, info). Calibrated for badge use; all pairings pass WCAG AA at 14px+ against the matched register canvas.

Badge text colour is --text-high-dark on dark badges and --text-high-cream on cream badges. All 14 pairings verified WCAG AA at 14px+.

Semantic States

Warm-shifted state palette (Dark Rebuild)

Five states, each with a dark- and cream-register variant. Warm-shifted so they sit naturally on the warm canvas. Info shares hexes with the Execute Phase by design, both are system signals. Existing --state-complete, --danger, --amber tokens above stay in place; component-level migration to the new tokens queues separately.

Dark Register

setmode operates in the dark register: marketing site, brand book, workspace app, academy. The dark register holds the chrome of every environment setmode runs in. Deliverables setmode produces and hands over (slides, proposals, briefs, reports) stay in the cream register. Tones here are cool neutral, near-black and charcoal, not warm-brown. The contrast across registers is intentional: cool instrument inside a warm artefact world.

Dark canvas, surface, elevated

Dark borders

Dark text

Cream canvas, surface, elevated (paired with dark register)

The cream tokens below are siblings of the dark tokens above. Each pair binds to the same semantic alias under the appropriate [data-theme] selector. --canvas-cream matches the existing --canvas; the explicit suffixed name exists so the mode-mirror reads cleanly in the SSOT.

Cream borders

Cream text

Mode binding

The system law. Dark register is where setmode operates. Cream register is what setmode produces and hands over. Environments declare data-theme="dark" on the html element. Deliverables declare data-theme="cream" or omit the attribute, since cream is the default. Each row below binds one --mode-* alias to its register-specific source token. The build script reads these rows and emits the matching [data-theme] blocks in brand.css.

Theme	Alias	Source token
cream	--canvas	--canvas-cream
cream	--surface	--surface-cream
cream	--surface-hi	--elevated-cream
cream	--border	--border-default-cream
cream	--divider	--border-subtle-cream
cream	--text	--text-high-cream
cream	--text-2	--text-mid-cream
cream	--text-3	--text-low-cream
cream	--mode-canvas	--canvas-cream
cream	--mode-surface	--surface-cream
cream	--mode-elevated	--elevated-cream
cream	--mode-border-subtle	--border-subtle-cream
cream	--mode-border	--border-default-cream
cream	--mode-border-strong	--border-strong-cream
cream	--mode-text-high	--text-high-cream
cream	--mode-text-mid	--text-mid-cream
cream	--mode-text-low	--text-low-cream
cream	--mode-text-faint	--text-faint-cream
cream	--mode-phase-align	--phase-align-cream
cream	--mode-phase-build	--phase-build-cream
cream	--mode-phase-execute	--phase-execute-cream
cream	--mode-cat-strategy	--cat-strategy-cream
cream	--mode-cat-workflow	--cat-workflow-cream
cream	--mode-cat-tooling	--cat-tooling-cream
cream	--mode-cat-governance	--cat-governance-cream
cream	--mode-cat-measurement	--cat-measurement-cream
cream	--mode-cat-rhythm	--cat-rhythm-cream
cream	--mode-cat-change	--cat-change-cream
cream	--mode-success	--success-cream
cream	--mode-warning	--warning-cream
cream	--mode-critical	--critical-cream
cream	--mode-terminated	--terminated-cream
cream	--mode-info	--info-cream
dark	--canvas	--canvas-dark
dark	--surface	--surface-dark
dark	--surface-hi	--elevated-dark
dark	--border	--border-default-dark
dark	--divider	--border-subtle-dark
dark	--text	--text-high-dark
dark	--text-2	--text-mid-dark
dark	--text-3	--text-low-dark
dark	--mode-canvas	--canvas-dark
dark	--mode-surface	--surface-dark
dark	--mode-elevated	--elevated-dark
dark	--mode-border-subtle	--border-subtle-dark
dark	--mode-border	--border-default-dark
dark	--mode-border-strong	--border-strong-dark
dark	--mode-text-high	--text-high-dark
dark	--mode-text-mid	--text-mid-dark
dark	--mode-text-low	--text-low-dark
dark	--mode-text-faint	--text-faint-dark
dark	--mode-phase-align	--phase-align-dark
dark	--mode-phase-build	--phase-build-dark
dark	--mode-phase-execute	--phase-execute-dark
dark	--mode-cat-strategy	--cat-strategy-dark
dark	--mode-cat-workflow	--cat-workflow-dark
dark	--mode-cat-tooling	--cat-tooling-dark
dark	--mode-cat-governance	--cat-governance-dark
dark	--mode-cat-measurement	--cat-measurement-dark
dark	--mode-cat-rhythm	--cat-rhythm-dark
dark	--mode-cat-change	--cat-change-dark
dark	--mode-success	--success-dark
dark	--mode-warning	--warning-dark
dark	--mode-critical	--critical-dark
dark	--mode-terminated	--terminated-dark
dark	--mode-info	--info-dark

Tonal Scales (Foundation)

Underneath every role token in the dark and cream registers is a tonal scale. Each palette family has nine steps from lightest tint (step 1) to deepest shade (step 9). The role tokens above (canvas, surface, text-high, the Phase/Category/State variants) are positions within these scales. The scales are the foundation; the roles are the consumer-facing API. When a surface needs a tonal step that the existing roles do not name, pick from the scale directly.

Neutral scale (cool zinc, the dark register foundation)

Hand-anchored to the existing dark register values so canvas, surface, text remain identical to today.

Identity scale

Identity purple. The wordmark mark colour.

Phase Align scale

Phase Align cool slate.

Phase Build scale

Phase Build deeper slate.

Phase Execute scale

Phase Execute deepest slate.

Category: Strategy scale

Strategy & Alignment muted slate blue.

Category: Workflow scale

Workflow & Process restrained teal.

Category: Tooling scale

Tooling & Prototyping deep brass amber.

Category: Governance scale

Governance & Risk aged brick red.

Category: Measurement scale

Measurement & Value ledger sage.

Category: Rhythm scale

Operating Rhythm & Scaling dusty mauve.

Category: Change scale

Change Enablement warm terracotta.

Success scale

Success state.

Warning scale

Warning state.

Critical scale

Critical state.

Terminated scale

Terminated state.

Info scale

Info state.

Legacy dark-callout tokens (retiring)

Used by the .dark-ground callout block and the framework.dark single-question layout. Surgical dark-on-cream usage that predated the full dark register. Slated for retirement once those callouts migrate to --canvas-dark and the theme-scoped aliases.

Print-Register Tokens

Used by generated document templates. setmode is 99% screen, these exist as compatibility tokens, not a separate brand context.

Accent Rules

Context	Active Accent	Rationale
Marketing site hero, nav, CTA	Mulberry	Brand-level. Not inside the programme yet.
Marketing site curriculum	Triad	Programme structure. Phase colours carry meaning.
Workspace module (Build)	Clay only	Single active phase. Others recede.
Workspace CTA	Mulberry	Action buttons always Mulberry.
Proposal / document	+ triad	Brand frames; phase inside programme content.
Email / social	Mulberry only	Pure brand context.

Typography

Gelasio sets the tone. Inter does the work. The H3 handoff: Gelasio for Display through H2 (22px+). Inter 600 for H3 and below.

Maximum weight: 600. No Bold (700). The wordmark is Gelasio 500. Headings match. Print uses Century Gothic.

Font-family tokens

Type-size scale (Dark Rebuild)

Locked 11-step size scale. Weights only 400 and 500. Italics: Gelasio italic at H1 / H2 for editorial flourish only. Display size used in the brand book cover and similar one-off display moments.

Spacing

Breakpoints

The five named viewport tiers used across the site. Authored as machine-readable tokens so any future change happens in one place. Note that CSS @media queries cannot interpolate variables; the literal values are written into media-query rules but must always match these tokens. The build script (scripts/build-brand-css.mjs) emits the tokens into brand.css's :root block alongside spacing and colour.

Responsive primitives

The site runs on a foundation-first responsive system. Display sizes scale fluidly via clamp(), narrow viewports get a phone-tier media-query block, touch devices get bumped tap targets, and five utility classes cover the patterns that come up most. All of this lives in brand.css; pages that link brand.css inherit it. The full programme, decisions, and acceptance bar are in workspace/mobile-responsiveness-plan.md.

Fluid display sizes

Display headings and stat numbers shrink smoothly as the viewport narrows. Desktop appearance is unchanged (max value matches the previous static size). Resize this window to see the cover heading and stat numbers below scale.

Touch targets

On coarse-pointer devices (phones, tablets), every <button>, [role="button"], [role="tab"], and .btn* class is bumped to a minimum height of 44px. Desktop visuals are unaffected. Inline text anchors are deliberately excluded; nav anchors are sized per-component as each surface migrates.

Utility classes

Five opt-in classes for the patterns that come up over and over. Use these instead of writing one-off media queries in component CSS.

Phone-tier media block

A single @media (max-width: 480px) block in brand.css applies three site-wide guarantees: inputs render at 16px minimum so iOS Safari does not auto-zoom on focus, body text uses overflow-wrap: break-word so long URLs and codes can never force horizontal page scroll, and common flex/grid children get min-width: 0 so they shrink with their parent.

Responsive components

Six reusable patterns published once in brand.css and used everywhere they apply. Each is opt-in (apply the class) and self-contained (no other CSS required to function). Use these instead of authoring bespoke navs, modals, drawers, or tab strips on each page.

1. Hamburger nav

Top nav with horizontal links above 768px and a collapsing hamburger panel below. Marker class .nav-shell on the wrapping element scopes all rules. JS toggles .open on the .nav-shell-mobile when the toggle button is clicked.

• Programme
• Sonar
• Perspectives
• Get in touch

Programme Sonar Perspectives Get in touch

Resize the window below 768px to see the hamburger appear. Click it to expand the panel. The toggle script is loaded once at the bottom of brand.html; copy it (or the inline onclick form used on the homepage) onto any page that uses the component.

2. Table wrap (.scroll-x-mobile)

Wrap a wide table or grid in a parent div with class .scroll-x-mobile. The horizontal scrollbar only appears when content exceeds container width. Same utility as Phase 1.6; treated here as the canonical pattern for any wide tabular surface (cohort tables, roster tables, contacts, deal-flow grids, etc.). Apply to the wrapper, never the table.

3. Responsive modal

Centered dialog box on desktop with a maximum width of 480px. At ≤480px the modal goes full-screen so cramped phones get the full surface. Wrap in .modal-backdrop for the overlay; apply .modal-responsive to the dialog itself.

4. Responsive drawer (side panel)

Fixed-width 380px drawer pinned to the right edge on desktop, full-screen on phone. Use for person panels, detail panels, contextual editors. Toggle visibility via JS by adding or removing the .open class. The drawer slides in from the right with a transform-only transition.

Markup: <aside class="drawer-responsive open">...</aside>. The drawer is fixed-positioned so it lives at the document level, not nested in cards.

5. Tab strip with horizontal scroll

Tabs flow left-to-right and scroll horizontally when they overflow. Native momentum on iOS, scroll-snap so taps still land neatly. The strip shows a thin scrollbar on desktop, none on mobile.

6. Auto-fit grid

Cards reflow into columns automatically based on a minimum card width. The min(N, 100%) trick prevents children from forcing horizontal page scroll when the viewport is narrower than the floor. Override the floor by setting --gaf-min on the parent (default 280px).

Buttons

Cards

Tags and Badges

Links

Form Inputs

Motion

Motion validates action. It does nothing else. Easing: cubic-bezier(0.16, 1, 0.3, 1)

Carousels

Two carousel types. Both use the system easing curve. Both have arrow + dot navigation with consistent styling. Neither auto-advances without user intent.

Shared rules: System easing curve on all transitions. Never bounce, spring, or snap. Arrow buttons: 40px circle, --divider border. Dots: 8px, --mulberry when active. Touch swipe supported on both. Wrap-around on Type B only.

06 · Applications

The brand applied

How the identity, voice, and visual system come together across every surface.

Marketing Site

Workspace

Academy

Four-layer facilitator certification programme. Document viewer, gate assessments, and progress tracking. Deployed as a single-page application alongside the workspace.

Social and Email

Slide Content Rules

Universal rules governing all setmode presentation formats: CEO presentation, proposal deck and operating principles. These rules exist because a slide is read in seconds, presented in a boardroom, and judged on clarity. Every word earns its place.

Presentation

CEO-level presentations delivered live. 16:9 dark canvas (#202124), projected in boardrooms and conferences. Century Gothic for PPTX output (Inter shown here as web proxy). Every element legible at 4 metres in a dimly lit room.

Property	Value
Aspect ratio	16:9 (1920 × 1080px)
Canvas	Dark #202124
Output	PPTX (primary), HTML single-file, or PDF via print
Typeface	Century Gothic (Regular, Bold). Courier New for metrics/codes.
Projection test	Every element legible at 4 metres in a dimly lit room
Accent rule	One accent colour per slide. Two at absolute maximum.
Amber reserve	Burnished Amber #8A6810 reserved for synthesis slide only

The deck uses a paired slide structure. Each principle gets two slides: the statement, then the evidence.

Proposal Deck

27-slide landscape A4 HTML presentation generated per organisation. Follows the elevator pitch arc: I/I gap, diagnosis, failure modes, client situation, evidence (8 principles), programme, sponsorship, ownership, compounding, investment. White print palette. Century Gothic typography. Mulberry/sage/clay/slate accent system. Travels without a presenter. Strategic register throughout.

AI Operating Model

The programme's crown deliverable. A landscape A4 Word document synthesised from all 28 deliverables at programme close. White paper palette. Century Gothic typography in production (Inter shown here as web proxy).

Reading System

The design system for every long-form document setmode produces: the AI Operating Model, proposals, reports, long-form perspectives, and this page. Any document where someone sits down to read, not scan, not navigate, not operate, uses this system.

Three principles: calm authority (warm canvas, restrained colour, generous whitespace), readable at every level (scanning, section reading, and deep reading all supported), and nothing decorative (every visual element carries meaning).

Reading Surface

Property	Specification
Ground	--canvas (#F8F5F0). Warm paper, not white. --surface (#FFFFFF) only for elevated elements (sticky header, data cards)
Content column	720px max-width, centred. Produces 65–80 characters per line at 18px, the proven readable range
Prose column	640px max-width for pull quotes, governing thoughts, declarations. Emphasis through compression
Body text	Inter 400, 1.125rem (18px), line-height 1.65. Paragraph spacing 1.5rem (24px)
Headings	Gelasio 500. Document title 1.75rem. Chapter title 1.4rem. No weight exceeds 600 anywhere in the system

Colour Restraint

Element	Colour	Signal
Content H3 headers	--mulberry	Structural section break within a deliverable
Pull quote left border	--mulberry 3px	"This is the key finding"
Content H4 left border	--mulberry 2px	Sub-section within structured content
Strong text	--mulberry	Emphasis, creates scannable anchor points
Em-dash list bullets	--mulberry	List items (functional)
Chapter divider rules	--sage / --clay / --slate	Phase transition (Align / Build / Execute). The only colour in the body
Active navigation	--mulberry	Current position in sidebar and sticky header
Governing thought	--mulberry	The single most important sentence in the document

Navigation & Structure

Summary view. Full navigation specification (retreat/reveal, scroll-spy, responsive breakpoints, motion) in Section 09.

Component	Specification
Cover	Centred title moment. Logo, org name (Gelasio 1.75rem uppercase), document type, date. Metrics visible on first screen. No sidebar, no header, just the content
Sticky header	48px, fixed top, appears on scroll past cover (opacity 200ms). Back arrow, logo, name, chapter label, download icon
Progress bar	2px --mulberry fill below sticky header. Updates via requestAnimationFrame
Sidebar	240px fixed left, floats over margin without displacing content. Fades in after cover (400ms). Retreats after 120px downward scroll. Never retreats while cursor is over it. Reappears on scroll-up or mouse near left edge. Hidden below 1024px
Chapter dividers	Top/bottom 2px coloured rules. Number + title on same line (flex row), subtitle below. 4rem between chapters
Deliverables	Flat (no cards). Number + mid-dot + name + status dot. Pull quote (first paragraph as Gelasio italic). Full content below
Declarations	Label (uppercase 0.65rem) + Gelasio italic text + attribution (em-dash prefix)
Print	Sidebar, header, progress bar hidden. Content fills page. Chapter dividers trigger page-break-before. Exact colour printing enforced

Application

Document	Uses
AI Operating Model	Full system: cover, sidebar, sticky header, 7 phase-coloured chapters, 28 deliverables with pull quotes, declarations, contributors, closing
Proposals	Reading surface + typography + content elements. No sidebar (short enough for linear reading). No phase colours. Pull quotes placed manually
Perspectives (long-form)	Reading surface + typography. No sidebar. Section breaks via horizontal rules. Pull quotes for key findings
Reports	Full system including sidebar. Chapter dividers with --divider rules (no phase colours unless programme-related)
Brand Truth	Typography system (18px body, 1.65 line-height, 96px section spacing). Retains 1120px container and horizontal nav for reference-document layout

Module Guide

Each of the 12 modules follows a fixed seven-component schema. Rendered in the workspace workbook and academy document viewer. The schema ensures every module is structurally complete before delivery.

LinkedIn Banners

Two distinct banners: the personal profile banner for Pierre and the company page banner for setmode.io. Both use canvas ground with mulberry accent, replacing the previous black-background treatment.

Personal Profile Banner

1584 × 396. Content shifted right to clear the profile photo overlay zone (lower-left ~20%).

Company Page Banner

1128 × 191. Narrower aspect ratio. Horizontal composition. Logo overlaps lower-left, so content centred or right-biased.

07 · Logo & Assets

The Logo Mark & Asset Library

Circle and capsule/pill shape. Mulberry mark, Oyster wordmark. Both unchanged across light and dark grounds. Every setmode brand asset, mark, wordmark, lockup, favicons, OG cards, LinkedIn banners, photography, is canonical here and served at setmode.io/assets/{filename}. Every other surface links by URL. Client logos live in Supabase (client-logos bucket, referenced via clients.logo_url).

Mark Only

Wordmark

Social Card Previews

Asset Library

Every brand asset in one place. Click to download. SVG is the source format, use it for all digital applications. PNG exports are provided for contexts that require raster files.

Logomarks

Favicons

Social & OG

Photography

08 · Standards

Do and Don't

Side-by-side reference for correct brand application across every surface.

Colour

Typography

Voice

Layout

Identity

Dark Ground

Near-black #242628, cool neutral. The only inverted surface in the system. Used for the website footer.

Role	Light Ground	Dark Ground
Ground	#F8F5F0 (canvas)	#242628 (near black)
Primary text	#2C2824	#F8F5F0 (canvas white)
Secondary text	#6B6560	#9A9490 (oyster)
Links	var(--text) (mulberry)	#F8F5F0 + underline
Border	#D9D5CD	#3A3C42
Phase chips	Solid or tinted	Solid or outline (no tints)

09 · Page Navigation

How readers move through pages

A reusable navigation system governing every multi-section page setmode produces. Three components, header bar, progress bar, and sidebar, combine to give persistent orientation without visual weight. The reference implementation is the AI Operating Model reader at workspace.setmode.io.

System Overview

Not every surface uses all three components. The header bar is always present. The sidebar appears on pages with defined section structure. The progress bar appears on long-form reading surfaces.

Header Bar

The header bar provides persistent identity and orientation once the reader scrolls past the cover. It is solid white, never frosted, never blurred, to maintain the institutional character of the reading surface.

Property	Specification
Height	48px
Position	fixed, top: 0, full width, z-index 100
Background	--surface (#FFFFFF), solid white, no backdrop-filter
Border	1px solid --divider (bottom only)
Padding	0 1.5rem
Content (left)	Back button · Logo · Page title (Gelasio 0.75rem / 500 / --text)
Content (right)	Chapter breadcrumb (Inter 0.72rem / 500 / --text-2) · Action icons
Visibility	opacity: 0 → 1, 200ms ease, triggered when cover leaves viewport

Progress Bar

Property	Specification
Position	fixed, top: 48px, full width, z-index 99
Height	2px
Track	--divider (#E8E5DE)
Fill	--mulberry (var(--text))
Transition	width 100ms ease-out
Visibility	Appears with header (same 200ms fade)
Facilitator variant	top: 72px when facilitator status bar is present
Calculation	scrollTop / (scrollHeight − clientHeight) × 100%

Sidebar Container

The sidebar floats over the left margin without displacing content. Its warm canvas background distinguishes it from the reading surface without creating a hard visual break.

Property	Specification
Width	240px
Position	fixed, left: 0, top: 48px
Height	calc(100vh − 48px)
Background	--canvas (#F8F5F0), explicit, not transparent
Border	None, no right border
Padding	1.5rem 0
Overflow	overflow-y: auto
Entrance	opacity: 0 → 1, 400ms ease
Facilitator variant	top: 72px, height: calc(100vh − 72px)

Item Hierarchy

Three levels of sidebar item exist: section (top-level navigation like "Summary", "Contributors"), chapter (collapsible groups like "What We Decided"), and sub-item (nested deliverables like "D-02 AI Vision"). Not every surface uses all three levels. The Brand Truth page uses only sections. The Operating Model reader uses all three.

Level	Font	Padding	Inactive Colour
Section	Inter 0.8rem (12.8px) / 500 / line-height 1.35	0.4rem 1.5rem	--text-2
Chapter	Inter 0.8rem (12.8px) / 500 / line-height 1.35	0.5rem 1.5rem	--text-2
Sub-item	Inter 0.72rem (11.5px) / 400	0.25rem 1.5rem 0.25rem 2.25rem	--text-3

Active State

Hover + Status Dots

Colour shifts to var(--text) #2C2824. No background change. No border change. Transition: color 150ms. The hover state is subtle, it confirms interactivity without competing with the active state.

Status	Colour	Meaning
Synthesised	--sage #4E6E4A (5px circle)	Content has been synthesised into the operating model
Awaiting	--border #D9D5CD (5px circle)	Content pending or not yet synthesised

Status dots are specific to the Operating Model reader. Other surfaces (Brand Truth, Perspectives) do not use status dots on sidebar items.

Collapsible Chapters, Dividers + Download

Property	Specification
Collapsed state	max-height: 0, overflow: hidden
Expanded state	max-height: 600px
Transition	max-height 250ms ease
Trigger	Active chapter auto-expands. Clicking a chapter heading toggles expansion
Dividers	1px solid --divider, margin: 0.75rem 1.5rem. Between logical groups
Download button	Inter 0.72rem / 600, --mulberry text + 1px border, 4px radius. Hover: --mulberry background, white text

Behaviour Modes

The sidebar operates in one of two modes depending on the surface.

Retreat / Reveal Specification

Behaviour	Specification
Retreat trigger	scrollDownDistance exceeds 120px of sustained downward scroll
Accumulation	scrollDownDistance increments on each scroll event where currentScrollTop > previousScrollTop
Reset	Any upward scroll (currentScrollTop < previousScrollTop) resets scrollDownDistance to 0
Reveal (scroll)	Immediate, sidebar appears as soon as scroll direction reverses
Reveal (mouse)	Mouse enters within 40px of left viewport edge (clientX < 40)
Hover lock	mouseenter sets lock; mouseleave releases. Sidebar never retreats while lock is active
Nav-click lockout	Smooth-scroll navigation clicks suppress retreat for 1500ms
Animation	opacity 0 ↔ 1, transition: opacity 400ms ease

Scroll-Spy

Active section tracking updates the sidebar highlight and header breadcrumb as the reader scrolls. Two observer strategies work together: an IntersectionObserver for deliverable-level tracking and a scroll handler for chapter-level tracking.

Observer	Target	Threshold	rootMargin	Purpose
Cover	Cover element	0	default	Toggle header/sidebar visibility
Deliverable	Each deliverable section	0.1	−20% 0px −60% 0px	Track active deliverable in sidebar
Chapter	Scroll handler	,	,	getBoundingClientRect().top vs HEADER_HEIGHT + 40px

Smooth scroll offset: 80px (48px header + 24px facilitator status allowance + 8px breathing room). Smooth-scroll navigation clicks temporarily suppress retreat behaviour (1500ms lockout).

Responsive + Print

Breakpoint	Sidebar	Header	Body
Above 1024px	Full 240px sidebar	48px, padding: 0 1.5rem	Standard
Below 1024px	Hidden entirely	48px, padding: 0 1.5rem	padding: 0 1.5rem 3rem
Below 768px	Hidden	48px, padding: 0 1rem	Cover and content font sizes reduced
Print	Hidden	Hidden	Content fills page width. Progress bar hidden

Application by Surface

Surface	Header	Sidebar	Progress	Mode
setmode.io (homepage)	48px, white, scroll-spy on section links	None	None	,
setmode.io/sonar	64px top nav	240px fixed filter	None	Persistent
setmode.io/brand	48px, white, breadcrumb	240px, 9 sections	2px mulberry	Hover-reveal
Operating Model reader	48px, white, breadcrumb	240px, chapters + deliverables	2px mulberry	Hover-reveal

Individual Perspectives articles, the workspace application, Academy, and print stylesheets are out of scope for this spec.

Design Tokens

Token	Value	Role
--surface	#FFFFFF	Header background
--canvas	#F8F5F0	Sidebar background
--divider	#E8E5DE	Header border, progress track, sidebar dividers
--mulberry	var(--text)	Active border, active text, progress fill, download button
--text	#2C2824	Hover text
--text-2	#6B6560	Inactive section/chapter text, breadcrumb
--text-3	#6B635B	Inactive sub-item text, header subtitle
--border	#D9D5CD	Awaiting status dot
--sage	#4E6E4A	Synthesised status dot

Motion Summary

Nothing animates by default, motion only occurs on state change.

Element	Duration	Easing	Property
Header appear	200ms	ease	opacity
Progress bar appear	200ms	ease	opacity
Progress bar fill	100ms	ease-out	width
Sidebar appear	400ms	ease	opacity
Sidebar retreat / reveal	400ms	ease	opacity
Collapsible expand / contract	250ms	ease	max-height
Hover colour	150ms	ease	color
Active state	instant	,	color, font-weight, border-left-color

10 · Intellectual Foundation

What the research says

Eight research tenants drawn from McKinsey, BCG, Gartner, Bain, Stanford HAI, MIT, IBM, Kyndryl, RAND, S&P Global, Accenture and IDC. More than 30,000 executives surveyed across 100+ nations.

Tenant 1 · Adoption Is Universal. Value Is Rare.

The adoption question is settled. 88% of organisations use AI regularly in at least one function. Corporate AI investment hit $252.3 billion in 2024. The value question tells a different story: more than 80% see no tangible EBIT impact. Only 5-6% qualify as high performers. The gap between adoption and value is the defining structural condition of organisational AI.

Finding	Source	Year	Sample
88% regular AI use in at least one function	McKinsey, State of AI	Nov 2025	1,993 respondents, 105 nations
78% organisational adoption (up from 55% in 2023)	Stanford HAI AI Index	2024	Aggregated from McKinsey surveys
80%+ see no tangible EBIT impact from generative AI	McKinsey, State of AI	Mar 2025	1,491 respondents, 101 nations
74% have yet to show tangible value	BCG, Where's the Value in AI?	Oct 2024	1,000 CxOs, 59 countries
72% breaking even or losing money on AI	Gartner, CIO Survey	May 2025	506 CIOs
Only 5-6% qualify as AI high performers	BCG; McKinsey	2025	1,250+ executives

Tenant 2 · Value Comes from Design, Not Technology

The 5-6% generating value share structural characteristics about how the organisation is designed, not which technology it deployed. Workflow redesign is the single largest driver of EBIT impact. High performers are three times more likely to have redesigned workflows. 70% of AI challenges are people and process. Most organisations invert the ratio.

Finding	Source	Year
Workflow redesign: single largest effect on EBIT impact (of 25 attributes)	McKinsey	Mar 2025
55% of high performers redesigned workflows vs. 21% overall	McKinsey	Mar 2025
70% of AI challenges are people and process	BCG	Oct 2024
16% with modernised processes achieve 2.5x revenue growth	Accenture	Oct 2024
Workflow redesign yields 25-30% productivity vs. 10-15% from tools	Bain	2025

Tenant 3 · AI Programme Failure Is Institutional

42% of companies abandoned the majority of their AI initiatives in 2025. 88% of proofs of concept fail to reach production. More than 80% of AI projects fail, twice the non-AI IT rate. The consistent finding: failures are institutional, not technical. The AI component functions. The organisation rejects it, abandons it, or cannot operate it in production. Four of five root causes are organisational.

Finding	Source	Year
42% abandoned the majority of AI initiatives (up from 17% YoY)	S&P Global 451 Research	Early 2025
88% of AI PoCs fail to reach production	IDC / Lenovo CIO Playbook	Feb 2025
80%+ of AI projects fail (2x non-AI IT failure rate)	RAND Corporation	2024
4 of 5 root causes of AI failure are organisational	RAND Corporation	2024

Tenant 4 · Governance Is the Structural Prerequisite

Only 18% have an enterprise-wide AI governance council. Only 28% have a CEO overseeing AI governance. CEO governance oversight is the element most correlated with EBIT impact. 57% estimate their data is inadequate for AI. 97% of CDOs struggle to demonstrate AI business value. The governance gap and the value gap are causally connected.

Finding	Source	Year
Only 18% have enterprise-wide AI governance council	McKinsey	May 2024
Only 28% have CEO directly overseeing AI governance	McKinsey	Mar 2025
CEO governance oversight most correlated with EBIT impact	McKinsey	Mar 2025
57% estimate data is inadequate for AI	Gartner	Feb 2025
97% struggle to demonstrate AI value (data + org limitations)	Informatica	Jan 2025
Only 12% of Fortune 100 disclosed board-level AI training	EY / Harvard Law Forum	2025

Tenant 5 · Workflow Redesign Is the Largest Value Driver

Workflow redesign produces the single largest effect on EBIT. High performers are three times more likely to have fundamentally reworked processes. BCG's 10-20-70 ratio: 70% of effort should go to people and processes. AI productivity gains reverse at 4+ tools: 33% more decision fatigue, 39% more errors, 9-point increase in intent to quit. The remedy is the same as the value driver: redesign the workflow.

Cross-reference: workflow redesign evidence also appears in Tenant 2 (value comes from design). Tenant 2 establishes the principle; Tenant 5 isolates the mechanism.

Finding	Source	Year
High performers 3x more likely to have redesigned workflows	McKinsey	Mar 2025
AI productivity reverses at 4+ tools ("brain fry")	BCG / HBR	Mar 2026
33% more decision fatigue, 39% more errors beyond threshold	BCG / HBR	Mar 2026
9-point increase in intent to quit	BCG / HBR	Mar 2026

Tenant 6 · Investment and Alignment Are Different States

95% of senior executives report investing in AI. Only 14% have aligned workforce, technology and business goals. The 81-percentage-point gap represents organisations that have funded, deployed and trained without designing the conditions for coordinated institutional outcomes. 80% approach AI with efficiency as the primary objective. The organisations generating the greatest EBIT impact set growth and innovation alongside cost reduction.

Finding	Source	Year
95% investing in AI, only 14% achieved alignment	Kyndryl	May 2025
14% ("AI Pacesetters") are 3x more likely to have change management	Kyndryl	May 2025
80% cite efficiency as primary AI objective	McKinsey	Mar 2025
Only 36% of employees believe AI training is sufficient	BCG, AI at Work	Jun 2025

Tenant 7 · The Performance Gap Is Widening and Compounds

AI leaders show 1.7x revenue growth, 3.6x three-year total shareholder return and 1.6x EBIT margin versus laggards. The gap widened on every measured dimension from 2024 to 2025. Only 7% have fully scaled AI. The organisations that moved from pilot to scale built governance and workflow redesign into the pilot phase rather than deferring it.

Finding	Source	Year
AI leaders: 1.7x revenue growth, 3.6x TSR, 1.6x EBIT margin	BCG, Widening AI Value Gap	Sep 2025
Gap widened across every dimension from 2024 to 2025	BCG	Sep 2025
Only 7% have fully scaled AI	McKinsey	Nov 2025
Leaders plan to spend 2x+ on AI vs. laggards	BCG	Sep 2025

Tenant 8 · The Post-Programme Stall Is Predictable and Preventable

Organisations that build genuine capability during a structured programme frequently stall in the months immediately following programme close. The capability does not degrade. What disappears is the operating rhythm: scheduled reviews, accountability cadences, facilitated decision points. Four cadences are the minimum viable structure: weekly operational, monthly leadership, quarterly governance, annual strategic. If the rhythm is functioning by month three, capability compounds independently. If absent, the stall has begun.

Evidence Governance

Every claim published under the setmode name is cross-referenced against primary sources. An internal discrepancy register tracks any attribution corrections, figure updates or partially verified findings. Corrections are applied to published content within one review cycle.

Full source bibliography (32 primary sources with URLs, sample sizes and publication dates) maintained internally.

13 · Frameworks

Brand frameworks

The visual IP the programme teaches. Each framework is authored once here; a build script extracts to structured data that module deck renderers consume. Change a framework in this page and the decks pick it up on the next build.

The Opportunity Space

What counts as an opportunity under the AI Vision?

Idea

Unvalidated. Aspirational. Not yet scoped or sized. Valuable as signal; not a commitment.

Opportunity

Vision-aligned. Scoped. Sized. A real option the room can evaluate and rank. Three example shapes: a decision that took a human hour gets drafted in a minute. A document the firm produces every week gets generated against the house style. A pattern buried in client data gets surfaced for every leader on Monday.

Initiative

Committed. Resourced. On the path. Owned by a named person with a named deadline.

What this is. What it is not.

Read it once. Decide if you are in.

This is

This is not

A working programme that produces twenty-eight deliverables your organisation owns at the end

A training course or a series of workshops

A working build of the structures your organisation will run AI through

An audit of where you sit against your peers

A twelve-week build with a signed founding statement and a measured commissioning

Something that runs without your time, your evidence and your signature

Return On Operating Model

This is the one you build. 12 weeks.

Deliverables.

28 named outputs across 7 categories. Each carries the name of the leader who built it.

Prototypes.

2 working AI workflows in live operations. The first demonstrates the pattern. The second proves it repeats.

Rhythm.

An operating cadence {{cohort.client_name}} runs from week 13. It compounds.

How the room holds itself.

Three commitments. They hold from week 1 through week 12.

The room builds the model.

setmode brings the framework. The room produces the content. What the room builds, {{cohort.client_name}} runs.

The signature is the commitment.

The CEO signs the Founding Statement. Functional leaders sign their sections. A signature converts a draft into a deliverable.

The work compounds.

Every module produces a deliverable. Every deliverable enters the operating model. The operating model compounds across 12 weeks.

The most senior voice in this room is the one with the most specific observation.

Individual vs Institutional Intelligence

The operating model gap is the distance between them.

Individual

Knowledge in heads.
Walks out the door.

→

Institutional

Knowledge in the system.
Compounds forever.

Individual Intelligence

The capacity of each person to work faster with AI tools. It grows with the person, improves with the person and belongs to the person.

Institutional Intelligence

The collective compounding capability of the organisation, encoded in shared systems, owned across functions, persistent over time.

Where most organisations are

After the programme

AI as individual tools

AI as operating infrastructure

Knowledge stays with the individual

Knowledge compounds in shared systems

Scattered adoption, informal governance

Governed capability across functions

Strategy and execution separated

Strategy and capability built simultaneously

Consultant dependency, fragile capability

Institutional ownership, permanent capability

Starting again after every departure

28 named deliverables compounding from session one

The organisations building an AI-enabled operating model are accumulating structural advantage with each passing quarter.

The AI Operating Model

What does an AI operating model actually contain?

D02 · AI Vision

What the firm is trying to become, not just buy.

D05 · Value stream

The one process the firm redesigns with AI first.

D06 · Canvas

Ownership, decision rights, rhythms, escalations written down.

D07 · Workflow library

Each workflow mapped with its human checkpoints.

D08 · Blueprints

Two workflows documented to production-ready standard.

D09 · Use cases

Quantified portfolio across time, money, risk, quality.

D10 · Prototypes

Two live deployments in production.

D12 · Scaling roadmap

From prototype to programme, sequenced.

Intent Before Opportunity

Why does intent come before opportunity?

Without intent

Opportunities multiply. Choices are random. Value leaks across scattered bets that never compound.

With intent

Opportunities filter. Choices align to the thesis. Value compounds because every deliverable strengthens the one before.

AI Vision

A vision is not a strategy. It is the statement the strategy serves.

A vision

Advisory margins have compressed 7 points. We will build the firm where partner judgement compounds at the margin rather than where output volume wins the quote. Today named. Three years from now named. Humans Decide named.

Not a vision

We will use AI to improve efficiency and enhance customer experience. Names no Today. Names no Three years from now. Names no Humans Decide line. Nothing the firm could disagree with.

Strategy and Vision

The layer that moves and the layer that stays.

Strategy

What the organisation will do.
The layer that moves.

→

Vision

What the organisation is becoming.
The layer that stays.

Without a vision, every AI decision is taken on first principles.

Work Stream and Value Stream

The function you keep and the work you redesign.

Work Stream

The function you came in with.
Your home through M03.

→

Value Stream

The work the programme redesigns.
Your unit from M04.

From M04 the room works by Value Stream, not by function.

From Function to Flow

Why this is a change of posture, not just a change of teams.

Optimise the flow, not the step

A function makes its own step faster. A Value Stream improves the whole journey to the customer. Speed up one step alone and the bottleneck simply moves.

Design the handoffs

Value leaks where work passes between functions. The Value Stream makes those handoffs the thing you design, not the thing you inherit.

Own an outcome you share

You work with Members who do not report to you, toward a result none of you owns alone. The posture moves from defending a function to delivering a flow.

AI inside a function moves the bottleneck. AI across a Value Stream moves the outcome.

The Mandate

What is the mandate, and why does it need signing?

Contains

Scope of authority, accountability and decision rights. The room knows exactly what they may choose.

Binds

The CEO to resource the work. The Value Streams to deliver it. Neither side can quietly step away.

Excludes

The unbounded remit. Mandate is a line the Value Streams do not cross without going back to the Sponsor.

The Three Horizons

What are we choosing between: now, next, later?

Now · 0–3 months

High confidence, narrow reach. The prototype that ships in this programme.

Next · 3–12 months

Medium confidence, wider reach. The workflow the Value Streams carry into the quarter.

Later · 12+ months

Lower confidence, highest ambition. The organisational reshape the thesis aims at.

The Operating Model Canvas

What does an operating model actually look like on a page?

Ownership

Who is accountable for each deliverable, named.

Decision rights

Who approves what, and at what authority level.

Rhythms

The cadences at which review, update and audit happen.

Escalations

When and how exceptions route upward.

Resources

Named tools, named budget, named people.

Constraints

The lines this operating model does not cross.

Accountability as Structure

Why does accountability need to be designed, not declared?

Declared

"We'll be accountable." Aspirational language. No named owner, no logged decision, no review rhythm. Accountability in theory.

Designed

Named owner. Named decision right. Named escalation. Logged. Reviewed at a known cadence. Accountability in practice.

Workflow as Architecture

What is a workflow, and why does it need architecture?

Task

One step. One person. Done in minutes. A single unit of work with no decisions and no handoffs.

Process

A sequence of tasks with known inputs and outputs. Still linear. No branching on context.

Workflow

End-to-end. Decision points. Exceptions. Interfaces between roles. The actual shape of work once AI is embedded. Every workflow names five components. Trigger, AI tool, human checkpoint, output, governance note. An AI tool is any software component that produces output using AI: a plugin, an integration, a custom service. A human checkpoint is the named review point where a person reads AI output and makes a call before it travels.

Quantified Value

What does 'value' mean when we say a use case has it?

Time saved

Hours per week, per engagement or per deliverable. Measurable and stackable.

Money

Revenue won, cost avoided, margin recovered. In currency, not percentage.

Risk reduced

Incident count, exposure window, compliance gap. Fewer things that could break.

Quality gained

Accuracy, consistency, senior-review pass rate. What the client or reviewer notices.

Governance Sequencing

Why does governance follow use cases, not precede them?

Governance first

Theoretical risks. Blanket policies. Adoption blocks before any real work has been tried. Everyone waits; no one learns.

Use cases first

Real risks visible in live work. Governance designed against what actually happens. Policies earn their authority by responding to evidence.

Architecture as Portfolio

What does a portfolio view reveal that individual views can't?

Individual view

Each participant's architecture looks complete on its own. Their function is covered. Their deliverables hang together.

Portfolio view

Same work rendered across the Value Streams. Gaps emerge (functions no one owns). Overlaps emerge (two teams building the same thing). The architecture gets real.

Playbook vs Prototype

Why does scoping precede building?

Prototype

What we build inside the programme. Code. Prompts. Workflows. A working demonstration of the thesis in one place.

Playbook

What survives the programme. Pattern. Principle. Template. The thing someone uses Monday morning without the facilitator in the room.

Deployability

What does 'it works' actually mean?

Works

Produces the intended output on real inputs, reliably enough to trust.

Safe

Fails in ways you can tolerate. Respects data boundaries. Won't embarrass the firm.

Supportable

Someone knows how to maintain it who isn't the person who built it. Documented.

Adoptable

Users will actually use it. Fits the work. Easier than the old way within the first few tries.

Scaling Is Sequencing

Why copying isn't scaling.

Copying

Same prototype cloned across teams. Breaks at the edges where context differs. Governance lags behind. The Value Streams revert.

Sequencing

Prototype → pattern → playbook → programme. Each layer generalises. Governance keeps pace. The Value Streams compound.

Commissioning

What does it mean for this system to be commissioned?

Architecture ratified

All 28 deliverables signed off. The CEO has said "yes, this is what we've built."

Ownership named

Every deliverable has a named owner who is not the facilitator. The operating model has left the programme.

Capability self-sustaining

The next quarter's work is scheduled and resourced by the firm, without needing the facilitator to convene it.

The Alumni Network

What does alumni mean for the room?

Alumni

Ratified practitioners, not students. The firm commissioned the work; you built it. You graduate as architects of your own operating model.

Network

Peer support across firms. Alumni from other programmes encounter the same questions six months later. The network is where those answers live.

Pledge

An ongoing commitment: to share, to review, to help future programmes. Alumni is a position, not a certificate.

AI Readiness: Baseline Rubric

How ready are you, with evidence?

Capability

1, no named practitioners. 3, pockets of use, untaught. 5, documented practice, trained Value Streams, audit trail.

Culture

1, AI seen as threat. 3, quiet curiosity, no permission. 5, AI-use is expected, signalled from the top, normalised.

Data infrastructure

1, knowledge in heads and inboxes. 3, some systems, weak retrieval. 5, institutional knowledge queryable, governed, current.

Governance

1, no policy. 3, informal guidance, no enforcement. 5, written policy, named owners, escalation paths, audit.

Tooling

1, consumer accounts only. 3, enterprise licences, no workflow integration. 5, integrated stack, procurement discipline, measured ROI.

What Changes Next Week

What will you do with this baseline between now and M02?

1 · Draft your strategic thesis

150 words. What kind of organisation does this firm intend to become with AI. Specific enough that a competitor couldn't sign it.

2 · Name one off-limit

One thing AI will not do in this firm. Write it in a sentence. Explain why.

3 · Come back with evidence

One client, one engagement, one workflow, where the thesis is testable. Due before session opens.

Your Operating Model Canvas

Fill each zone with your current state, target state and named owner.

Ownership

Who is accountable for each component today. Who will be accountable at target. Named person, not role.

Decision rights

Who approves what today. Who approves what at target. Authority level named for each.

Rhythms

The cadences that exist today. The cadences that must exist at target. Frequency and purpose for each.

Escalations

How exceptions route upward today. How they must route at target. Trigger and named recipient per path.

Constraints

The lines the operating model does not cross. Today and target. Explicit, not implied.

Your Workflow Canvas

Map one workflow end to end. Six fields, every field named.

Trigger

What starts this workflow. The event, the input, the request.

AI tool

Which tool does the work. Name the tool, the version, the configuration.

Human checkpoint

Where a human reviews before output is released. Who, what they check, what fails.

Output

What the workflow produces. The artefact, the format, the destination.

Governance note

The policy that covers this workflow. The data class, the risk, the mitigation.

Owner

The named person accountable for this workflow running safely.

The Use Case Brief Rubric

Six sections. Every section passes the CFO test before the brief is complete.

Problem

Named, specific, with the workflow it lives in. Not a category, a concrete friction.

Solution

The AI workflow proposed. The tool, the checkpoint, the output. Specified, not sketched.

Effort

Hours to build. Hours to run per week. Ongoing cost. Named.

Value

Time, money, risk, quality. Quantified. Estimation method stated.

Risks

What could fail. How it would be noticed. Specific, not generic.

Mitigations

For every risk, the control in place. Human checkpoint, data boundary, escalation.

Value, Under Challenge

Present the value claim. The room probes. The estimate survives or it changes.

The claim

The value your use case produces. Time saved, money made, risk reduced. With the number, the unit and the estimation method.

The probe

Three tests every estimate faces. Named unit economics per user, per week or per transaction. Named comparable benchmark the number can stand next to. Named uncertainty range the estimate survives within. Where the estimate breaks.

Your Prototype Scope

Name what you will build, what you will not. Six fields.

Target workflow

The one workflow the prototype delivers end to end.

Inputs

What the prototype consumes. Format, source, owner.

Outputs

What the prototype produces. Format, destination, quality bar.

First prompt

The opening instruction given to the AI. One paragraph, specific.

Dependency

What must exist for this to work. Data, access, tool, permission.

Out of scope

What this prototype will not do. Named explicitly, so the boundary holds.

AI In Your Organisation Today

Silent capture. The room sees the themes, not the names. Honesty lands here or in the board report later.

Tools

The AI tools your people use today. Name them. Paid accounts, personal accounts, embedded features. Include employees using external AI tools without IT approval. Vendor tools that embed AI without disclosure.

Data

The data your people put through those tools. Client material, internal documents, personal information. Include automation touching client data without a named reviewer.

Checkpoints

Where a human reviews before output leaves the firm. Where no checkpoint exists today. Include anything branded AI used without a named owner.

Your Governance Canvas

Five sections. Every section named with specifics, not principles.

Decision rights

Who approves AI deployments. Who approves new tools. Who approves new use cases. Named.

Risk categories

The categories of AI risk this firm holds. Data, factual, reputational, regulatory. With thresholds.

Escalation paths

When an incident occurs, who gets the call. Name, role, phone. Paths for each risk category.

Data handling

Which data classes can flow through which tools. Approval level per tool per class.

Review cadence

How often the framework is reviewed. Who reviews it. What triggers an interim review.

Your Playbook Scope

One playbook per function. Six fields name what it contains and what it does not.

Target function

The function this playbook serves. The named owner on receipt.

Workflows in scope

The specific workflows covered. Referenced from the validated library, not paraphrased.

Prompts

The prompt library entries attached. Named by id. One per covered workflow minimum.

Governance

The governance checkpoints that apply. The sections of the framework referenced.

Tools

The approved tools. The configuration guidance. The data classes each handles.

Worked examples

One end-to-end example per covered workflow. Real inputs, real outputs, real checkpoints.

Scope, Under Challenge

Present the scope. The room probes. What is missing. What is over-scoped.

The scope

The target function, the workflows, the prompts, the governance coverage. As you have it written today.

The probe

Three tests every scope faces. Does it hold at M10 demo day. Does it name a measurable output. What cuts first if the deadline slips. Where the boundary is too wide or too narrow.

The Demo Rubric

Three minutes. Four things a demo must show.

Input shown

The real input the prototype consumed. On screen, not described.

Output shown

The real output the prototype produced. On screen, with quality notes.

Handoff shown

The human checkpoint. Who reviewed. What they changed. What they approved.

Peer question

One question from the room at the end. Answered in one minute or noted for follow-up.

Your Scaling Roadmap

Place every initiative in Now, Next or Later. Every item has a named owner.

Now · 0 to 3 months

What the firm deploys in the next quarter. Named owner. Resource committed. Dependency cleared.

Next · 3 to 12 months

What the firm builds across the year. Named owner. Sequenced against the Now initiatives it depends on.

Later · 12 months plus

The longer reach. Named owner per initiative. Signalled now so budget and talent arrive in time.

The Six Levers

Every opportunity lives on one of these. Name the lever first, the opportunity second.

Revenue

New revenue won. Existing revenue protected. Margin recovered at the top line.

Cost

Cost avoided. Cost removed. Hours out of the run rate.

Risk

Incident exposure reduced. Compliance gap closed. Fewer things that could break.

Speed

Time from trigger to output shortened. Decisions reached faster. Queues cleared.

Quality

Accuracy gained. Consistency gained. Senior review pass rate improved.

Talent

Senior time reclaimed. Junior capability compounded. Hiring or retention helped.

What Moves To The Top

Five criteria. Rank the shortlist. Defend the top. Defend the bottom.

Value

Size of the payoff in the chosen lever. Quantified or best-estimated.

Effort

Hours to build. Hours to run. Dependencies cleared or outstanding.

Fit to thesis

Sharpness of match to the D02 thesis. High, medium, low.

Risk of delay

What gets worse if this slips a quarter. What compounds if this lands this quarter.

Owner readiness

The person who would own this today. Their current capacity. Their current commitment.

Self-Sustaining, Under Challenge

Walk every item in silence. Where all three are absent, the room sees the count, not the case. Thin rhythm lands here or in the operating review later.

The claim

Your roadmap. Every initiative, its owner, its cadence, its review point. As you have it today.

The probe

Three things every item must have. A named owner, first and last name. A named cadence with real dates. A named review point, not a meeting. Rewrite until the roadmap runs without external help.

Operating Principles

Eight structural findings on organisational AI value, drawn from surveys of more than 30,000 executives across 16 institutions between 2023 and 2026.

Tool deployment does not create value

The operating model determines the return

Governance precedes deployment

Workflow redesign produces the most value

People adopt what they understand

Alignment converts investment into capability

The advantage compounds

Structure sustains capability

More frameworks land here as the authoring pass proceeds. The module deck pipeline reads this section; every new <article class="framework"> becomes renderable across all 12 modules where it's referenced by slide data_source.