Fix the organisation you have — one component at a time.

What is true right now, written where the work is.

Thirty-three prose pages across Organization, Value Streams, Operating Model and Technology each open on a Current tab. It is typed by hand — the AI does not write your present. It may be left empty, so a component nobody has surveyed yet is visibly blank rather than quietly guessed at.

Internal Pressure closes by naming where an internal strain compounds a known external force — the pairs that turn two manageable problems into one urgent one.

• Value Stream Swimlane draws every stream across the departments that perform it. Departments are lanes; a step's sideways position comes from who does it, never from where it sits in the sequence — so a handover is an arrow crossing a lane boundary, and a backward arrow is a real rework loop. The lanes are not reordered to hide it.
• Analyse Department Processes opens every department-owned process as a card and records eight things about it: the trigger, the outcome and who receives it, the systems people actually work in, the bottlenecks and rework loops, the decisions that still need a human and on what basis, concrete AI opportunities named per step, volume and frequency, and the accountable role.
• External and Internal Pressure read the forces rather than the position. Every pressure carries its evidence, an intensity of High, Medium or Low, a horizon of Now, one-to-two years or three-to-five, and the one thing it forces you to do — closing with the cost of ignoring the five strongest for another year.
• Headcount per department and average annual cost per employee are typed figures on the department and the role, not numbers parsed out of generated prose. A department left empty takes the sum of the roles under it, so the two views cannot drift apart.

The gap, per component — not one verdict on the whole organisation.

The Delta tab compares your Current against New, the target the platform writes for that same page, and needs both to say anything. A gap on the Application Landscape reads nothing like a gap on Mission, which is exactly why the comparison sits per page instead of once per organisation.

• Delta is generated from the two texts in front of it, so it moves when either side moves. Correct the Current and the gap changes with it.
• Process Impact turns the same comparison into a number you can act on: agent-run processes with no AI opportunity written down yet — work the target hands to agents with nothing recorded about how.
• The pressure ranking supplies the order. A wide gap under low pressure is not the first thing you fix, and the two pages together say which is which.
• The page documents sidebar, the page instructions box and the refine box render on all four tabs and act on the tab in front of you — refining while reading Delta refines Delta.

Three horizons, risks and dependencies — on the page itself.

The Plan tab sets out the route from Current to New in three horizons — the first ninety days, the rest of year one, and beyond year one — and closes with the risks and dependencies that route carries. It needs the target to exist, because a route without a destination is a wish list.

• Ninety days first, on purpose: the horizon a sponsor can still hold someone to, and the one that decides whether the programme survives its first quarter.
• The plan is prose you can edit, regenerate, or refine with an instruction — the same controls as every other page, acting on the Plan tab.
• OKRs cascade company objectives to every department, each department objective tracing to exactly one company objective — so no plan carries an objective with nothing above it.

Ownership lands where the content lives.

A component's plan sits on that component's page, beside the text it is about. There is no second place where it also lives and slowly diverges. Each target carries a status of Not started, In progress or Final, with a Mark as final control beside it.

• Sign-off is computed, not stored as a flag: the status is worked out on every read by hashing the live text and comparing it against the hash captured at sign-off. An edit made after sign-off drops the badge back to In progress instead of leaving Final standing over content that has since changed.
• Named menu views give each stakeholder their own surface — a board member, a department head and an external consultant see the pages that concern them, saved once under a name and assigned rather than rebuilt per person.
• Permissions resolve in three layers — base role, then the view, then the user's own overrides, which still win — merged per field, and the server enforces the same map, so a hidden control is not merely hidden.

Every plan on one board.

Thirty-three plans on thirty-three pages answer a component question well and a programme question not at all: what are we doing, in what order, and how far have we got. Growth → Roadmap is that board — every page whose Plan tab holds content, with a priority, a status and an order you set yourself.

Owners, dates, effort and sub-steps are not in this version. The board answers order and progress; it is not a project-management tool pretending to be one.

• The rows are not stored. They are computed on every read from the saved plans, so clearing a plan removes its row with no sync step — and there is no state in which the board can claim a plan that a page does not have.
• The board's own table holds only the tracking — status, priority, position — and a row is written the first time one of those changes. Opening the board writes nothing at all.
• Regenerating a plan keeps its row exactly as it was: the tracking belongs to the page, not to a particular revision of the text.
• The plan text cannot be edited on the board, deliberately. It belongs to the page's own Plan tab, which already carries the refine box and the regenerate button — two editors over one string is how the two come apart.