You Set the Direction—So Why Won't Your Engineers Move? An Org-Design Decision Playbook

• Who this is for: IT managers who now decide org direction and team structure (especially the newly promoted), tech leads / EMs, and execs or HR rethinking organization design
• Assumed knowledge: A working feel for software teams. Every framework is unpacked in one line at first mention, so no prior reading is required
• Reading time: About 20 minutes

Overview

“I set the direction. I explained it in the meeting. And yet my engineers aren’t moving.” Nearly every newly promoted manager hits this wall. The cause is usually not the content of the direction—it’s the way it was decided, the flow of the decision itself.

When you decide org direction or structure, the classic playbook (diagnose the present → design → plan → implement) is still the indispensable foundation. Translating strategy into the shape of the organization, aligning structure, rewards, and people—without that design, even a good direction disintegrates on the floor. But the classic flow has a built-in trap. If you lock the destination and final form up front and run it as one straight line, by the time it’s finished the assumptions have gone stale. It drifts away from your engineers’ lived sense of the work today, and no one owns it. That’s the waterfall failure mode.

What fixes this is weaving in modern decision methods. The PR-FAQ, which writes the finished state as prose first to surface contradictions in assumptions. The Type 1 / Type 2 distinction, deliberate only on irreversible calls and fast on reversible ones. OKRs, which re-tune the route every quarter. Team Topologies, which draws team boundaries along the flow of value. These don’t replace the classics—they are threads that patch the classics’ weaknesses: rigidity, drift, and slowness.

And in the AI era, this stopped being a matter of taste. McKinsey’s 2025 survey found that 88% of organizations use AI in at least one business function, while only about a third have scaled it across the enterprise. What moves the needle on impact (EBIT) is not “adopting tools” but fundamentally redesigning workflows—yet only 21% of AI-using organizations have actually done it¹. A line that separates winners from the rest is being drawn in territory you can’t reach by tweaking the org chart.

This article aims to give a newly promoted manager something they can apply to their own team size and phase right away: (1) the classic playbook and its pitfalls, (2) the modern methods worth weaving in, (3) a situation-by-situation cheat sheet, and (4) responses to objections from both classicists and modernists. The conclusion, stated up front: build on the classic foundation, weave in the modern, and never treat a decision as “decided once and done.” That is the most realistic path to raising both buy-in on the floor and execution speed at the same time.

Why “correctly decided” direction stalls on the floor

Start with a familiar scene. Just promoted to lead a product team, you’re handed the quarter’s org direction. You spend weeks analyzing the market and the status quo, package it into clean slides, and present it at the all-hands. No one objects. And yet the next week, the codebase hasn’t changed. Engineers are quietly burning down their usual tasks.

What’s happening here isn’t a problem of ability or motivation. It’s a structural one: the flow of the decision isn’t connected to how the floor actually thinks. The classic playbook for org design roughly runs in this order:

1. Diagnose the present (how do we work now?)
2. Design (the structure and processes we should have)
3. Plan (the migration roadmap)
4. Implement (rollout and embedding)

The flow itself is correct. The problem is that new managers run it like a waterfall—lock the final form first, push 1→4 in a straight line. Two diseases follow. One is “fixed ends, hollow middle”: the destination (strategy) and the landing (org chart) look impressive, but the why that should sit between them never reaches the floor. The other is “staleness”: while you spend weeks on design, the tech and the team’s situation move, and the finished direction no longer matches reality.

The very person who systematized the classic flow warns about this. Lynda Gratton, who codified work redesign, argues we should reframe it not as “nailing the right answer in one shot” but as “managing a process of redesign and change.” The core of her four steps—(1) understand what matters, (2) reimagine new ways of operating, (3) model and test, (4) act and create—is that each step embeds feedback loops and co-creation with the people doing the work². In other words, the classic flow was always meant to run as a learning loop, not a straight line. What trips up new managers is stripping out that loop and turning it into a waterfall.

Part A: The classic playbook—”the foundation that turns strategy into form” is still non-negotiable

Before reaching for modern methods, be clear about what the classics guarantee. Skip this, and the modern methods below become “just a way to create chaos.”

1. Integrated design—before the org chart, align structure, rewards, and people

The most common failure in org design is drawing the org chart and calling it done. The classic literature consistently warns against this. Jay Galbraith’s Star Model holds that performance is never determined by structure alone; it emerges only when five levers—strategy, structure, processes (information flows), rewards/evaluation, and people—are aligned with one another. Misalign any one and friction appears, with a “shadow organization” forming behind the formal one³.

A concrete example IT companies hit constantly: you declare “we’re going flat and autonomous,” yet the evaluation system still runs along individual vertical reporting lines. Because structure (flat) and rewards (vertical) aren’t aligned, engineers say “agile” out loud while optimizing for what gets rewarded—reporting up. That’s the textbook case of direction that won’t move.

To design at larger scale and scope, you need the mindset of a TOM (Target Operating Model). McKinsey’s new rules for operating-model redesign call for translating strategy into structure across the whole organization, designing in units of value streams, and choosing among multiple options with explicit trade-offs rather than a single plan. Their “Organize to Value” maps 12 elements—purpose, talent, leadership, rewards, and more—as one system, checking whether each choice aligns with strategy⁴. In one bank’s case, teams meant to be end-to-end value streams turned out, in reality, to be a repackaging of the old structure: strategy still handed off to product management, which approached solution architecture, which delivered to developers⁴. Change the name of the structure, and nothing changes unless you design the linkage between the elements—this is the heaviest lesson the classics teach.

2. The classic flow’s pitfalls (the ones new managers fall into hardest)

The value of the classics is solid. But spell out the traps a new manager is prone to.

• Too heavy to act: Just diagnosing the present for a TOM can take three to six weeks. “We’re already agile, this is overkill,” you think—and end up doing nothing.
• Locked by perfectionism: You treat the first design as “the answer” and forbid course-correction while running. Gratton’s loop is lost².
• Drift from the floor’s reality: The design stays at abstract boxes and lines, never connecting to the codebase and workflows engineers touch daily.

That’s the “foundation.” It’s necessary, but on its own it can’t withstand uncertainty. Next, the threads you weave into the foundation.

Part B: Weaving in the modern—raising adaptability under uncertainty

The essence of modern methods is to supplement the classics’ “soundness of design” with speed of adaptation to change. Not replacement—weaving.

1. Lock the destination first—backcasting and the PR-FAQ

One reason the classic flow drifts from the floor is that consensus ends in the vague “sounds about right” of slides. Amazon’s Working Backwards / PR-FAQ cuts sharply here. The move is simple: write the press release and FAQ as prose first, as if the product or project were already finished. Starting from the customer, you ask “What needs to be true for this to succeed?” to surface contradictions in assumptions up front. In the meeting you use no slides; you run a “narrative meeting” where for the first 15–20 minutes everyone reads the document in silence, focusing on testing the truth rather than selling⁵.

Where this earns its keep in an IT company is clear. When deciding a new feature or platform direction, don’t dive into design—write a PR-FAQ describing “the state six months out, with engineers using this platform.” Contradictions like “for that, we need an auth foundation first” or “this assumption doesn’t exist in-house yet” become visible before a line of code is written. Think of it as landing backcasting (reasoning back from the desired endpoint) into a verifiable, written form.

2. Be deliberate only on the irreversible, decide the reversible fast—Type 1 / Type 2

The single biggest habit that kills a new manager’s execution speed is treating every decision as heavy. In his 2015 shareholder letter, Jeff Bezos split decisions into two kinds. Type 1 decisions are irreversible or nearly so—”one-way doors”; once through, you can’t get back. Decide them carefully, with consultation, slowly. Type 2 decisions are reversible—”two-way doors”; if it’s wrong, reopen the door and walk back. High-judgment individuals or small groups can and should make them fast. Bezos’s warning: the larger an organization gets, the more it applies the heavy Type 1 process to Type 2 decisions, producing slowness, excessive risk aversion, and too little experimentation⁶.

Translated to an IT company: choosing your database engine or a public API spec (you can’t break backward compatibility) is Type 1—line up multiple options and trade-offs, decide carefully. Internal directory structure, a temporary CI setting, the team’s review workflow—those are Type 2—the lead decides on the spot and changes it next week if it doesn’t fit. Just placing one question—”can I walk this back?”—at the entrance of every decision changes an organization’s speed dramatically.

3. Keep the destination fixed while re-tuning the route—OKRs

Once you’ve fixed the endpoint by backcasting, adjust how you approach it every quarter. OKRs (Objectives and Key Results) separate the objective you don’t change (O) from the measurable results (KR) you revisit regularly. This is where the classic “stick to the plan you decided” and the floor’s “the situation moves” get reconciled. Don’t move the destination; re-tune the route—that is how “continuous evolution” is implemented.

4. Draw team boundaries along the “flow”—Team Topologies

A modern pattern that cuts teams by the flow of value rather than the org chart is Team Topologies. It sorts teams into four types—stream-aligned (deliver value directly to customers), platform (a foundation that accelerates other teams), enabling (temporarily lift another team’s capability), and complicated-subsystem (a complex area needing specialist knowledge)—and makes the way teams relate explicit through three interaction modes (collaboration, X-as-a-Service, facilitating)⁷. The aim is to reduce ambiguous hand-offs and clarify each team’s “API.”

The value for a new manager is being able to instantly locate your own team’s role with one question: “are we a product team or a platform team?” It also pairs well with Dynamic Reteaming—recombining teams as the situation demands rather than fixing them.

5. Under AI, “workflow redesign,” not “tool adoption,” decides the outcome

This is the most important thread right now. According to McKinsey’s 2025 survey (fielded June–July 2025; 105 countries, 1,993 respondents), 88% of organizations use AI in at least one business function, while about two-thirds have not scaled it across the enterprise. And what moves impact (EBIT) most is fundamental workflow redesign—yet only 21% of AI-using organizations have rebuilt at least some workflows from the ground up. Among the “high performers” who are getting results, 55% have redesigned—about 2.8 times the 20% rate of other firms¹. Stalled AI adoption usually takes the shape of a pilot that “just bolts on an in-house tool and never scales.”

Looking further ahead, McKinsey describes the AI-agent-native organization as a shift from the “org chart” of traditional hierarchical delegation to a “work chart” based on exchanging tasks and outcomes. Decision-making and communication go flat, and the expected shape is one manager orchestrating a small human team alongside many agents⁸. The message repeats here too: the real challenge isn’t the technology—it’s redesigning workflows, leadership, and culture⁸.

The implication for a new manager is plain. The moment you shrink “responding to AI” into “which tool do we adopt,” you join the pilot-stalled majority. The question to ask is “how do we rebuild the flow of this work itself?”

Part C: When to use what—a situation-by-situation cheat sheet

Here is everything above, classic and modern, put into a form a new manager can map onto their own situation and pick from. Left to right: “what you’re deciding,” “the classic foundation you can’t skip,” “the modern method to weave in,” and “the trap new managers fall into.”

What you’re deciding	Classic foundation you can’t skip	Modern method to weave in	Trap new managers fall into
New product / feature direction	Translate strategy into the flow of value4	Write the finished state first with a PR-FAQ to surface contradictions5	Mistaking the vague “consensus” of slides for a decision
Team split / structure	Align structure, rewards, and people via the Star Model3	Make boundaries explicit with Team Topologies’ four types and interaction modes7	Drawing only the org chart, leaving rewards/evaluation as-is
Irreversible tech choice (DB, platform, public API)	Evaluate tech and governance together via a TOM4	Treat as Type 1; compare multiple options and trade-offs carefully6	Treating reversible calls as heavy, slowing everything down
Quarterly priorities	Cascade from strategy to KRs	Re-tune with OKRs; decide reversible calls fast in small groups6	Never revisiting mid-quarter the direction you fixed at the start
Diagnosing a dysfunctional org	Pinpoint misalignment via Step 1 “Understand”2	Redesign workflows from the ground up (don’t settle for adding tools)1	Moving boxes on the org chart and calling it done

There is one way to use this table: don’t skip the classic column. Modern methods become “adaptive” only on top of a classic foundation; without the foundation, they become “disorder.” Running one full row—both cells—is the smallest unit of weaving.

Put the whole approach on one page and it looks like this. Thread the modern through each classic step, verify on the floor at the end, and loop back to the top—not a one-way waterfall, but a loop that keeps turning.

graph TD
    A[Diagnose the present<br>classic 4 steps] --> B[Lock the destination first<br>backcasting and PR-FAQ]
    B --> C[Integrated org design<br>align structure rewards people]
    C --> D[Deliberate only on irreversible<br>Type1 and Type2]
    D --> E[Re-tune each quarter<br>OKR]
    E --> F[Verify and learn on the floor<br>workflow redesign]
    F --> A

Objections and caveats—answering critics on both flanks

This playbook draws legitimate criticism from both classicists and modernists. Keep the responses ready so a new manager can explain them on the floor.

Classicist: “Modern methods create chaos and blur accountability. They’ll just confuse new managers.” A fair concern. The response: cleanly separate accountability for irreversible parts via Type 1 / Type 2. Even when you delegate reversible decisions to small groups, ultimate accountability for irreversible ones stays with a person (the manager). Don’t bolt on governance as separate oversight—embed it into the flow of the work. In an IT company, the already-ingrained code-review culture is the model: apply heavier review only to irreversible changes, within the daily flow, and carry that same idea straight over to org decisions.

Modernist: “Make everything agile and you lose strategic coherence.” Also fair. The response: a hybrid that fixes the destination (O) by backcasting and re-tunes only the route (KR) each quarter with OKRs. Precisely because you don’t move the destination, you can flex the route without coherence collapsing.

IT-specific: “We’re already flat and we run Dynamic Reteaming. We don’t need the classics.” A common bit of pride. The response: check exactly one thing with the Galbraith Star Model—are rewards/evaluation and people-placement aligned with that structure?³ However flat the structure, if evaluation stays vertical, “flat” is just a façade. For a new manager, one question suffices: “Can we evaluate, on flat criteria, the results produced by a flat structure?”

State the caveats too. The McKinsey 2025 data cited in the AI-native section is the latest available, but it is a single, self-reported survey, and you can’t conclude the future from it¹. The picture of the agentic organization is a prediction of the current direction, not empirical proof⁸. Read it as “this is how things are trending now,” and verify it in your own context.

A new manager’s first move

Summarizing the theory is useless if you can’t act on it. Here’s the smallest step you can try next quarter.

1. Pick your next org decision and ask “can I walk this back?” If you can (Type 2), decide it small and fast without waiting for consensus.
2. Write that decision’s “finished state” as a one-page PR-FAQ. Prose, not slides. Include one line: “what needs to be true to succeed?”
3. Check one Star Model question. For the structure that decision moves, are rewards/evaluation and people-placement aligned with it? If not, that’s why the floor isn’t moving.
4. Fix the destination; revisit the route every quarter with OKRs. Stop “decide once and done,” and replace it with a loop.

None of these wait on a multi-week TOM analysis—you can start today. Not discarding the classic foundation, but threading the modern through it one strand at a time—that is the most realistic way to keep direction from ending at “merely decided.”

Conclusion

• The reason direction stalls on the floor is usually not its content but the flow of the decision. Turn the classic flow into a one-way waterfall and you get fixed ends with a hollow middle, plus staleness.
• The classic playbook (the four-step diagnosis, the TOM, the Galbraith Star Model) remains the non-negotiable foundation for translating strategy into organizational form and aligning structure, rewards, and people²⁴³.
• Weave in the PR-FAQ (write the finished state first), Type 1 / Type 2 (deliberate only on the irreversible), OKRs (fix the destination, re-tune the route), and Team Topologies (cut boundaries along the flow), and adaptation speed rises⁵⁶⁷.
• Under AI, what decides the outcome is not “tool adoption” but fundamental workflow redesign. The reality—88% using it, about two-thirds not scaled, only 21% having redesigned—shows it⁸¹.
• It folds into one line: build on the classic foundation, weave in the modern, and don’t treat it as decided-once-and-done. Even a new manager can start turning this loop from the very next decision.

Related articles

• Should You Become a Generalist, or Defend the Division of Labor? — AI-Era Role Design That Changes With Company Size - Organizes the answer to “which structure to choose” by company size
• What Separates People Who Use AI From People Used By It Is Organizational Learning Design - The talent-development side that sustains continuous evolution
• From Two-Pizza to One-Pizza Teams: What Gets Lost When AI Halves Your Team - Concrete trade-offs in team-composition decisions
• Building an AI-Native Engineering Team: OpenAI Codex Guide Explained - A practical approach to building AI-native teams
• From Tight Coupling to Loose Coupling in Team Collaboration: Designing How We Work in the AI Era - Rethinking team design through the lens of coupling

References

References are listed in the order their citation numbers appear in the text.

1. The State of AI: Global Survey 2025 - McKinsey & Company (2025). 88% use AI in business but about two-thirds have not scaled it. Workflow redesign has the biggest effect on EBIT impact, yet only 21% of AI-using firms have done it (55% among high performers vs. 20% of others, ~2.8x). n=1,993, 105 countries, fielded June–July 2025. 【Reliability: Medium-High】 ↩︎ ↩︎² ↩︎³ ↩︎⁴ ↩︎⁵

2. The Four-Step Process for Redesigning Work - Lynda Gratton, MIT Sloan Management Review (2022). Reframes org redesign as a learning process rather than a one-shot answer, in four steps. 【Reliability: Medium-High】 ↩︎ ↩︎² ↩︎³ ↩︎⁴

3. Implementing Your Business Model With the Galbraith Star Model - Strategyzer / Jay Galbraith. The Star Model: alignment across the five levers of strategy, structure, processes, rewards, and people. 【Reliability: Medium】 ↩︎ ↩︎² ↩︎³ ↩︎⁴

4. The new rules for getting your operating model redesign right - McKinsey & Company (2025). New rules for operating-model redesign (strategy translation, value streams, integrated design, multiple options with trade-offs) and the 12-element “Organize to Value.” 【Reliability: Medium-High】 ↩︎ ↩︎² ↩︎³ ↩︎⁴ ↩︎⁵

5. Working Backwards PR/FAQ Process - workingbackwards.com (by former Amazon executives). A narrative method that writes the finished state first and surfaces contradictions via “What needs to be true?” 【Reliability: Medium-High】 ↩︎ ↩︎² ↩︎³

6. 2015 Letter to Shareholders - Jeff Bezos, Amazon (2016). The Type 1 (irreversible, one-way door) vs. Type 2 (reversible, two-way door) distinction, and the trap of large orgs over-applying Type 1. 【Reliability: High】 ↩︎ ↩︎² ↩︎³ ↩︎⁴

7. Team Topologies — Key Concepts - Matthew Skelton & Manuel Pais (book, 2019). Designing organizations along the flow of value via four team types and three interaction modes. 【Reliability: Medium-High】 ↩︎ ↩︎² ↩︎³

8. The agentic organization: Contours of the next paradigm for the AI era - McKinsey & Company (2025). The shift from org chart to work chart, flat decision-making, and the claim that the challenge is not technology but redesigning workflows, leadership, and culture. 【Reliability: Medium】 ↩︎ ↩︎² ↩︎³ ↩︎⁴