What this document is
The Strategy argued why a firm like this must change. This document draws what it changes into. It describes the company not as a set of departments on an org chart but as a designed system with three layers and a clear job for each: a layer that sets direction, a layer that measures and manages, and a layer that does the value creating work, with meetings as the wiring that carries information and decisions between them.
Northwind Services is an illustrative composite, not a client. Its numbers are constructed to show how the method works, not to report a measured result. Northwind is a mid market B2B managed services firm of roughly 400 people, selling a mix of projects and retainers to IT and operations buyers at organizations of 1,000 to 10,000 staff. It has a commercial function, a delivery function, a finance function, and a small internal operations group, with three layers between an individual contributor and the chief executive. Everything attributed to Northwind below is a worked illustration of the architecture, never a report of something observed.
This is the architectural document. Where the Strategy is meant to convince, this is meant to be built and run. It follows one pattern in every chapter. It describes how companies like this typically operate and why that shape evolved. It states how the illustrative company works today, including where a real organization must capture its own current state from the field rather than assume it. It describes the future state specifically. It names what the machine does and what the human does. It lists the architectural changes underneath. And it closes on the business impact and the notes for building it.
One honesty marker before we start. A blueprint gives you the target architecture, the company as designed. It does not by itself give you the messy current reality of every function, the actual Tuesday that goes wrong. Wherever a chapter needs that lived current state, this document either shows the constructed Northwind illustration and labels it as constructed, or it marks the place where a real organization has to go and look. It never quietly fills the gap with plausible fiction, because a blueprint built on invented current states designs the wrong building.
Chapter 1: The shape of the company
In plain English. A company has to do three different kinds of work that do not mix well. It has to decide where to go. It has to keep track of whether it is getting there. And it has to actually do the work that customers pay for. The Blueprint separates these into three layers, Lead, Measure, and Do, and treats the meetings between them as the mechanism that keeps the layers in sync. Most companies smear these three together inside every role and every meeting, which is why direction, measurement, and work constantly interfere with each other.
The control layer is more than measurement alone. It measures, it manages against what it measures, and it mentors the people doing the work so that capability grows rather than only performance being scored. Measure, manage, mentor is one job in three parts, and a company that does the first without the other two produces scorecards nobody learns from.
Current reality. In a typical company the three kinds of work are tangled. The person doing the work also reports on it, also interprets the numbers, also argues about strategy, all in the same meetings. This tangling was rational when information was scarce, because separating the layers would have required moving information between them constantly, and moving information was the expensive thing. So companies collapsed the layers to save on information movement, and paid for it in confusion about which conversation they were even having.
What this means at Northwind. In the illustration, Northwind has committed to the separation on paper while its weekly rhythm still mixes the layers. The leadership review is nominally a decision forum, and in the constructed picture it spends most of its time establishing what is true rather than deciding what to do, which makes it a synchronization meeting wearing a decision meeting's name.
This is exactly the kind of observation a real organization has to capture rather than assume. How cleanly the three layers actually run, how often a delivery review turns into a strategy debate, how often a strategy meeting collapses into status reporting: these come from sitting in the actual meetings and writing down what happened. The Northwind version above is a constructed illustration of the shape of the finding, not a substitute for doing the looking.
Future state. The three layers run as distinct systems with distinct jobs. Lead decides direction and holds the values and filters that every choice passes through. The control layer holds the scorecards, manages against them, and develops the people underneath them. Do runs the value engines and the service lines. Each layer has its own cadence and its own conversation, and no meeting is allowed to quietly become a different layer's meeting. The connective tissue between them is deliberate, not accidental, because the whole point of separating the layers is to make the flow between them explicit and therefore improvable.
What the machine does. It maintains the state each layer needs from the others continuously, so the layers no longer have to meet merely to synchronize. Do produces telemetry the control layer reads without a status meeting. The control layer produces a current picture the Lead layer reads without a reporting deck. The machine is the wiring, so the meetings can be about judgment rather than about catching up.
What the human does. Humans own the conversations that require judgment: setting direction in Lead, deciding what a divergence means and what to do about it in the control layer, and doing the work and the delivery judgment in Do. The layer separation is precisely so that human attention in each layer is spent on that layer's actual decisions rather than on reconciling the other layers.
Architectural changes. One shared source of truth for the state of the work, readable by all three layers, so the layers stop maintaining separate representations of the same reality. Defined interfaces between layers, so it is clear what Do owes the control layer and what the control layer owes Lead. And a meeting architecture, covered in its own chapter, that is designed rather than inherited.
Business impact. The direct impact is management leverage: fewer meetings whose only purpose is to move information between layers, and clearer decisions because each conversation knows which layer it belongs to. The indirect impact is speed, because a divergence in Do reaches the layer that can act on it without waiting for the weekly cycle that used to carry it.
Implementation notes. This chapter is the frame for all the others, so it does not get built directly. It gets built by building the layers underneath it. The test that it is working is simple and behavioral: people can say, of any given meeting, which layer it belongs to, and meetings stop sliding into the wrong layer.
Chapter 2: Lead, and the five filters
In plain English. The Lead layer answers one question, where are we going, and gives everyone a way to check any decision against that direction without escalating it. The Blueprint expresses this as a mission statement and five decision filters. The mission says what the company is for. The filters are the test any choice has to pass.
Current reality. Most companies have a mission statement nobody uses and make decisions by seniority and instinct. Direction lives in the heads of a few leaders, which means every nonobvious decision has to travel up to those heads and back down, which is slow and does not scale. The mission on the wall and the decisions in the meetings are two different things.
What this means at Northwind. A usable mission has a specific shape. It names an intersection: two or more things that are usually managed separately, plus the gap between them that keeps producing failures nobody owns. It explains why the company exists by naming the seam it sells into. That shape is the transferable part. The words below are an illustration of the form, not a template to copy, and any real company would have to write its own from the seam it actually occupies.
The Northwind illustration: Northwind exists at the intersection of the systems a customer runs and the operations those systems are supposed to carry, so the gap between what is technically running and what is actually working stops becoming the customer's problem to discover.
Five filters sit under the mission. Every meaningful choice has to pass them: is this getting us closer to our goals; does it align with our values; is it good for our customers; is it something we can actually do; and does it make the company less dependent on any single person to keep working. The fifth filter is the durability test, and it is the one most often left out, which is why so many services firms are a set of individuals rather than a company.
Whether those filters are actually applied to live decisions today, or whether they live in the document and not yet in the room, is a current state fact a real organization must capture rather than assume. The honest way to capture it is to take five recent decisions of consequence and ask which filter was applied out loud at the time.
Future state. The mission and the filters are the working test for every meaningful choice, applied out loud, at the point of decision, by whoever is making it, without escalation. A proposal that fails a filter is reshaped or dropped on the spot. The fifth filter, durability, is treated as a genuine design constraint and not a slogan, which means choices are consciously weighed for whether they leave the company more able to run without the person currently holding the thing together.
What the machine does. It can surface the filters at the moment of decision, hold a visible record of how past decisions were tested against them, and flag when a proposed change appears to conflict with a filter, so the conflict is examined rather than missed. It cannot apply the filters, because applying them is judgment, especially where two filters pull against each other.
What the human does. Humans set the mission, own the values, and make the actual call when the filters conflict, which they regularly will. What is best for one customer this quarter and what makes the company less person dependent are not always the same direction, and choosing between them is exactly the kind of judgment the Strategy says stays human and accountable.
Architectural changes. The filters become a real object in the system, attached to decisions rather than sitting in a document, so that the reasoning behind a choice is captured where the choice is made. This is the beginning of enterprise memory: not just what we decided, but which filter decided it.
Business impact. Decision speed and decision consistency. When direction is a shared, applied test rather than a set of leaders to escalate to, more good decisions get made faster and lower in the company, and fewer decisions quietly drift away from the strategy. The durability filter, applied consistently, compounds into a company that keeps working when individuals move on.
Implementation notes. Writing the mission and the filters is the easy half; the work is adoption. The test that it is working is that you can pick a recent real decision and hear which filter carried it, and that at least once a proposal everyone liked was reshaped because it failed a filter. Until a filter has actually killed or changed a decision, it is decoration.
Chapter 3: The three pillars
In plain English. The company organizes itself not by department but by three pillars of business value: Leveraged Sales, Transferrable Delivery, and Profitability. Every pillar shows up twice in the architecture, once in the control layer as a scorecard and once in the Do layer as the engines that produce it. The pillars are the spine that connects what we measure to what we do.
Current reality. Most companies organize by function: a sales department, a delivery department, a finance department, each with its own goals that frequently conflict. Sales books work delivery cannot staff. Delivery optimizes utilization in ways that hurt renewal. Finance chases margin in ways that starve growth. The departments are real and the shared outcome is nobody's job. This structure is another artifact of information scarcity: departments were containers for specialized knowledge back when knowledge was expensive to move, and their walls became the places value leaks out.
What this means at Northwind. The Blueprint replaces departments with pillars as the organizing logic. Leveraged Sales, Transferrable Delivery, and Profitability are each defined as a value to be produced, with engines beneath and a scorecard above. The word transferrable in the delivery pillar is doing real work: it signals delivery designed to be repeatable and transferable rather than heroic and person dependent, which ties straight back to the durability filter.
In the constructed Northwind picture, the seam between the commercial function and the delivery function is where the leakage shows: the commercial system of record and the delivery system of record disagree about scope, dates, and value, and someone reconciles them by hand every cycle. That is what a department wall looks like from the inside, expressed as recurring rework.
How far a real organization already runs on pillars versus still running on department lines underneath the drawing is a current state fact that has to be captured, not assumed. A practical way to capture it: ask ten people which pillar their work feeds and see how many answer with a department name instead.
Future state. The pillars are the primary structure. Leveraged Sales means growth that does not consume proportional effort: sales that leverage marketing, reputation, and repeatable motion rather than pure headcount. Transferrable Delivery means delivery that is systematized enough to hand off, scale, and survive the departure of the person who invented it. Profitability means the economics are engineered, not hoped for. People understand their work as feeding a pillar, and the scorecard for that pillar is the shared truth that stops the old department level optimization wars.
What the machine does. It keeps each pillar's scorecard live and honest from the underlying work, so the pillar's health is always visible without assembly. It can also surface when the pillars are pulling against each other, when a Leveraged Sales push is about to overrun Transferrable Delivery capacity, early enough to decide deliberately rather than discover painfully.
What the human does. Humans own the tradeoffs between pillars, which are genuine and constant. Booking more, delivering more transferably, and protecting margin are three goods that compete, and balancing them is leadership judgment. The architecture makes the tradeoff visible; it does not make it for you.
Architectural changes. Work, people, and systems get mapped to pillars rather than only to departments, so that the pillar view is real and not a reporting overlay. The scorecards become the managed objects. This is a significant change to how the company sees itself and will meet the most resistance, because departments defend their walls.
Business impact. This is where the largest structural value sits. Organizing by shared outcome rather than by functional silo attacks the single biggest source of the information tax, the leakage at department seams. It also builds the shape a buyer, a lender, or a successor all value for the same reason: leveraged, transferable, profitable is a company rather than a collection of people.
Implementation notes. Defining the pillars is not the work; making them the real operating structure rather than a picture is the work. The test is that a person can name which pillar their work feeds and what that pillar's scorecard currently reads, and that a cross pillar tradeoff was recently made deliberately, in the open, rather than by one department quietly winning.
Chapter 4: Do, the engines and the service lines
In plain English. The Do layer is where the value gets made. It has two parts. The value engines are the repeatable business motions that produce each pillar: Marketing and Bookings for Leveraged Sales, Project Governance and Account Management for Transferrable Delivery, Revenue for Profitability. The service lines are what the company actually sells and delivers, offered through a common service catalog so every line has the same shape.
Current reality. In most services firms the delivery work and the business motions that surround it are run by feel and by senior people who carry the method in their heads. The engine is a person, which means it does not scale, does not transfer, and breaks when the person leaves. The service catalog, if it exists, is a sales artifact rather than an operating one.
What this means at Northwind. The structural idea worth carrying is not the particular lines a company sells. It is that each service line is assigned a designated strategic role, so the portfolio is a deliberate set of jobs rather than an accumulated list of things the firm happens to do. A line that anchors recurring revenue and a line that wins deals nobody else can win are different assets and should be managed differently.
Northwind's illustrative lines and their roles:
- Platform Operations: the foundation. It is what everything else attaches to, and it earns the right to be in the customer's environment at all.
- Managed Support: the recurring anchor. It carries the retained revenue base and the daily relationship.
- Operations Assurance: the outside in lens. It looks at the customer's operations the way an unsentimental outsider would and reports what is actually true, which is the role no incumbent provider is naturally able to play for itself.
- Automation Studio: the differentiator. It is the capability that opens doors the other three then fill.
The service catalog gives all four lines a common shape: Assess, Operate, Manage, and Govern and Advise. Those verbs are illustrative. The point is that there are a small number of them, that they describe stages of a relationship rather than products, and that a customer can move along them in a known order.
The current maturity of each engine, which business motions are already systematized and which are still person dependent, is a current state fact that has to be captured from the delivery organization. Nothing in an architecture document can tell you which of your engines is secretly one person.
Future state. Each value engine is a defined, repeatable motion with an owner, not a person improvising. Each service line delivers through the common catalog shape, so a customer can move from Assess to Operate to Manage across any line and the company can cross sell along a known path. Delivery is transferable by design, which is the whole point of the middle pillar. The differentiator line is positioned to open doors the other lines then fill, turning a unique capability into a pipeline engine.
What the machine does. Inside the engines, the machine runs the uncertainty reducing parts: assembling the deal picture for Bookings, tracking project state for Project Governance, watching account health for Account Management, keeping revenue and collections current for the Revenue engine. Inside the service lines, it handles the continuous sensing that managed services in particular demand, where the monitoring load is far beyond what a human roster can watch and machine scale is not optional.
What the human does. Humans do the expert delivery: the design decisions, the calls that require reading a situation rather than a signal, the difficult conversation with a customer whose expectations and contract have drifted apart, and the relationships the work sits inside. They also own the engine designs and improve them. The machine runs the motion; the human owns the craft.
Architectural changes. The engines become defined objects with owners and playbooks rather than tacit practice. The service catalog becomes an operating structure that shapes delivery, not just a sales list. Cross line delivery paths get built so the Assess to Manage journey is real. This connects to the Playbooks and Resource Matrix infrastructure in a later chapter.
Business impact. Transferable, repeatable engines are the difference between a firm valued as a book of relationships and a firm valued as a machine. Operationally, systematized engines mean growth without proportional senior person burnout, which is the Leveraged Sales promise made real.
Implementation notes. This is the layer with the most real world texture and therefore the layer where field capture matters most. The honest current state of each engine and each line, what is systematized and what is heroic, has to come from the delivery organization before the future state per engine can be specified in full. This chapter frames the layer; the per engine and per line detail is what Deliverable 3 works out, function by function.
Chapter 5: Measure, the scorecards
In plain English. The control layer holds one scorecard per pillar and manages the company against them. Leveraged Sales is measured by pipeline, win rate, and deal velocity. Transferrable Delivery by utilization, customer satisfaction, and on time delivery. Profitability by margin, collections, and cost per delivery. These nine numbers are how the company knows whether each pillar is healthy. One scorecard per pillar, and no more, is itself a discipline: nine managed numbers, not ninety.
Current reality. Most companies drown in metrics and manage on almost none of them. Numbers are assembled by hand for meetings, they lag reality by days or weeks, different systems disagree about the same figure, and by the time a number is trusted it is old. The scorecard is a periodic artifact produced by human effort, which means it is expensive, stale, and slightly wrong, and everyone half knows it.
What this means at Northwind. The constructed Northwind picture shows what that costs. Delivery leads assemble progress into a shared format on a recurring cycle, a summary is produced from that collection for the leadership review, and the two systems of record are reconciled by hand because they disagree. In the illustration, the status collection consumes 96 hours a month across all participants, and the lag from the moment a project slips to the moment an executive knows about it is 9 days. Both figures are constructed for this example, not measured anywhere.
How each of the nine numbers is produced today, how current it is, how much manual assembly it takes, and how much the underlying systems agree: that is the current state a real organization has to capture, and it is worth capturing precisely, because that cost is the size of the prize. The Northwind numbers show the shape of the answer. They are not the answer.
Future state. Each of the nine numbers is live, trusted, and produced from the underlying work without anyone assembling it. The control layer stops spending its time producing numbers and starts spending it interpreting them: not what is the utilization, but why did it move and what should we do. The scorecard becomes a sensing instrument the company acts on rather than a report it produces.
What the machine does. It produces and maintains the nine numbers continuously from the source work, reconciles the systems that feed them, and flags divergence early. This is close to pure uncertainty reduction, the exact work the machine does best and the exact work that consumes the most human hours today.
What the human does. Humans interpret and decide. A number moving is not a decision; deciding what the movement means and what to do about it is judgment, and owning the resulting action is accountability. Mentoring belongs here too: the person who explains to a delivery lead why a number moved is doing the part of management that no dashboard performs. The control layer becomes a room full of decisions rather than a room full of reporting.
Architectural changes. The reality versus representation discipline from the Strategy lives here most sharply. The scorecard is a representation, and the architecture must keep it honest against the reality it claims to describe, which means the numbers are wired to source and continuously reconciled, and any number no one has recently checked against the world is treated as suspect. Enterprise memory captures not just the numbers but their movements and the decisions taken in response.
Business impact. Two large effects. The direct one is the elimination of the assembly cost, the human hours currently spent producing numbers rather than acting on them. The larger one is decision quality and speed, because a company that sees its nine numbers live and trusted catches problems while they are cheap, which shows up in every downstream metric from margin to satisfaction.
Implementation notes. Start by measuring the current cost and lag of the nine numbers, because that is both the business case and the baseline. The test that the future state is arriving is that the effort of producing the scorecard falls toward zero while the time spent deciding on it rises, and that a divergence was caught and acted on before it would previously have been noticed.
Chapter 6: Meetings, the connective tissue
In plain English. Meetings are the wiring between the three layers, the way information and decisions travel up and down. Saying so is unusually honest, because in most companies meetings are the largest hidden cost and nobody treats them as architecture. Here they are a designed part of the system, which means they can be designed down.
Current reality. Meetings are where companies pay the information tax most visibly. A large share of them exist only to move information between layers and people: to get everyone onto the same page, to report status upward, to push direction downward. They are expensive, they are frequent, and most of their content is synchronization that a shared, live source of truth would make unnecessary. They persist because, historically, a meeting was the cheapest way to sync human mental models.
What this means at Northwind. In the constructed illustration, the chain is visible end to end: a status collection feeds a summary, the summary feeds a leadership review, and the review spends most of its time establishing what is true. Three activities, one of which is a decision forum, and the decision forum is the one being crowded out.
The full meeting inventory is current state that has to be captured rather than assumed. Which recurring meetings exist between which layers, how much of each is synchronization versus decision, and how many hours the whole apparatus consumes. That inventory is uncomfortable and valuable, because it is a direct map of the information tax. This document deliberately does not supply a cadence, a headcount, or a meeting list for a real company, because inventing one would be inventing the very thing the exercise exists to discover.
Future state. Meetings split cleanly into two kinds, and the split is the whole redesign. Synchronization meetings, whose job was to get everyone current, largely disappear, because the layers are already current from the shared source of truth. Decision meetings, whose job is judgment, remain and get better, because the people in them arrive already informed and spend the whole meeting deciding rather than half of it catching up. The meeting stops being where information travels and becomes where judgment happens.
What the machine does. It removes the reason synchronization meetings exist by keeping every layer continuously current, and it prepares decision meetings by assembling the picture, surfacing the divergences, and drafting the options in advance, so human time in the room is spent on the decision and not the setup.
What the human does. Humans make the decisions the remaining meetings exist for. The scarce resource, senior attention, moves off synchronization and onto judgment, which is the Strategy's whole thesis applied to the calendar.
Architectural changes. The meeting architecture becomes explicit: each remaining meeting has a defined layer, a defined decision it exists to make, and a defined set of prepared inputs. Meetings without a decision get retired. Enterprise memory captures decisions and their reasoning from the meetings, so the record of why is not lost when the room empties.
Business impact. This is one of the most immediately felt impacts of the entire program, because everyone experiences the meeting load personally. Reclaimed senior hours convert directly into either capacity or leverage, and better prepared decision meetings improve the quality of the decisions that steer the whole company.
Implementation notes. Begin with the meeting inventory, because it is the fastest way to make the information tax visible and the reclaimed time is the most tangible early win. The test that it is working is that synchronization meetings are being retired, that the remaining meetings start with everyone already current, and that reclaimed hours are real and countable.
Chapter 7: The infrastructure of memory
In plain English. Beneath the visible layers sit four supporting structures: Value Engines, Systems, Playbooks, and Resource Matrix. Together they are the company's memory and reusable machinery, the place where how we do things is written down so it does not have to live in people's heads. In an AI native company this layer is not support, it is foundation, because it is what the machine reads to do its work.
Current reality. In most firms this layer barely exists. Method lives in senior heads, systems are a disconnected pile of tools, playbooks are stale documents nobody opens, and who can do what is known informally by a few people. The company's memory is human and therefore lossy, and it walks out the door regularly.
What this means at Northwind. Naming all four structures is the intent to make memory explicit. The constructed Northwind picture shows the cost of not having them: two systems of record that disagree about scope, dates, and value, reconciled by hand, is precisely what a missing shared foundation looks like in the daily work.
The current state of each of the four, how real the playbooks are, how connected the systems are, and whether the resource matrix reflects actual capability or aspiration, is exactly the kind of fact a real organization must go and capture. This layer is usually the least mature in any company and the most important to get right for an AI native design, which is a bad combination and a good reason to look before designing.
Future state. The four structures become the live foundation the rest of the system runs on. Playbooks are the current, used definitions of how each engine and service runs, and they are what the machine executes and improves against. Systems are integrated enough to hold one shared source of truth. The Resource Matrix is a live map of capability that staffing and delivery decisions read from. Value Engines are the defined, owned motions from the Do layer, catalogued and reusable. This layer is where the company's knowledge stops being tacit and becomes operational.
What the machine does. It reads this layer to do everything else: it runs the playbooks, it draws on the systems for the shared truth, it uses the Resource Matrix to reason about staffing, and it helps keep all of it current by noticing when reality has drifted from the written version. Crucially, the machine also helps maintain this layer, which is what keeps it from going stale, the failure mode that kills every playbook library.
What the human does. Humans author the method, make the judgment calls the playbooks cannot encode, and own the capability the Resource Matrix maps. They decide what good looks like; the layer records it and the machine runs it.
Architectural changes. This is the deepest architectural work, because it is the substrate. It means real integration toward a shared source of truth, playbooks that are living objects rather than documents, and a capability map that is maintained rather than assumed. Enterprise memory in the full sense lives here.
Business impact. The impact is leverage and transferability, the two things the whole company is being built for. A company whose method is explicit and machine runnable scales without proportional senior effort and survives the loss of any individual. This layer is quietly where much of the enterprise value is created.
Implementation notes. This layer is built gradually and continuously rather than in one push, and it is built by the Transformation System as each engine and service is redesigned: every redesign that reaches Operating Standard writes itself into this layer. The test that it is working is that a new person or the machine can run an engine from the playbook rather than from a senior person's memory, and that the layer is getting more current over time rather than less.
End of Deliverable 2. This Blueprint describes the company as a designed system. The Operating Model (Deliverable 3) makes each layer and engine runnable day to day. The Transformation System (Deliverable 4) is how we get from the current state, including the current states this document deliberately marked as things a real organization has to go and capture, to the future states described here, one tested change at a time. Northwind Services remains an illustrative composite throughout: its figures are constructed to show the method, and nothing here reports a measured result.