The small t / Big T language comes from Webster and Westerman at MIT Sloan — their GenAI risk slope. I built my own: a Business Transformation risk slope, resolved into three levels — Change, Disruptive Transformation, Scaling Transformation.

Small t is Level 1, Change: a better tool, a task automated. Low-risk, and on its own it never moves the P&L. Big T is the real crossing — Level 2, Disruptive Transformation, where margin first moves, and Level 3, Scaling Transformation, where new revenue does.

Here's why most go under at Level 2, and it isn't ambition. Level 2 requires genuine disruptive transformation — silos taken out of processes and tools, one version of the truth, the value chain rebuilt for how the work now flows, including where AI does it. That's hard, so most take the shortcut: they bolt AI onto the old model and skip the rebuild. The models that make Level 2 hold — the ones I lay out in the work itself — aren't optional and they aren't improvised; they're evidence-based, and they're the difference between margin that moves and an expensive layer of intelligence sitting on a broken operating model. Skip them, and Level 2 drowns you.

There's a second danger, and it's where my method earns its place: even done right, a transformation this deep sends the human system into resistance — and I've spent a career, and a book, on how to land the change without that resistance sinking it. But make no mistake about the order. The reason Level 2 fails is the rebuild skipped, not the people. Manage the people, skip the rebuild, and you still drown. Do the rebuild on the evidence — that's the crossing.

Webster, M., & Westerman, G. (2025). Generate value from GenAI with 'small t' transformations. MIT Sloan Management Review. · Wachter, R. P. (2024). Der Widerspenstige: Eine heitere Geschichte aus der Welt der Transformation. Independently published. · Business Transformation risk slope and the Change / Disruptive / Scaling levels — Ruth Pauline Wachter (2026).

They operate at different depths of the water, and that's the part most people miss.

Change management is a Level 1 tool. It's built to embed a change that's already defined — communicate it, train for it, manage the adoption curve. On my risk slope that's the shallows: it works for Change, where the end state is known and you're helping people get used to it. Necessary, and not nothing. But it's a tool sized for shallow water.

Transformation — Level 2 and Level 3, Disruptive and Scaling — is a different depth and needs a different tool entirely. Here the end state isn't defined yet; you're rebuilding how the business creates value, not embedding a settled change. Reach for change-management methods at this depth and you'll find they don't carry you — they were never built to. Communication and training don't rebuild a value chain, take silos out, or redesign an operating model for AI. That's the most common and most expensive mistake at Level 2: bringing Level 1 tools to a Level 2 crossing, running town halls and adoption plans over a transformation that needed structural redesign — and wondering why nothing moved.

So it's not that one is better. It's that they belong to different depths. Change management embeds what's already decided. Transformation decides — and builds — what isn't there yet. Use the shallow-water tool in deep water, and you don't cross. You sink, slowly, in a very well-communicated way.

The Change / Disruptive / Scaling levels of the Business Transformation risk slope — Ruth Pauline Wachter (2026).

Because the direction you're driving from cascades through the entire organization — and there are only two directions. Away from something, or toward something. The distinction is one of the most robustly evidenced in motivation psychology: a goal you move toward sustains energy and performance, while a threat you move away from drains both, because the mind spends itself scanning for danger instead of reaching for a target.

And that orientation doesn't stay at the top. A leader's regulatory focus measurably shapes the focus and behavior of the people below them — promotion-focused, "toward" leadership produces transformational behavior and primes the same drive in others; avoidance-focused, "away" leadership produces control and management-by-exception, and primes prevention and passivity down the line. A CEO driving from fear — of lost share, of the board — narrows into micromanagement and risk-aversion, and a culture forms that only raises its hand when failure is already certain.

This is why most transformations that "fail on communication" are really failing here: the change was framed as a threat ("transform or we go under"), and a threat, however true, pushes people into the very passivity that sinks the crossing. The work, before any system or model, is to convert that drive — from away into toward — by naming a shore worth reaching, so the organization moves toward something it can see. That conversion is psychological, it is measurable, and it is where I start.

Elliot, A. J., & Church, M. A. (1997). A hierarchical model of approach and avoidance achievement motivation. Journal of Personality and Social Psychology, 72(1), 218–232. · Johnson, R. E., King, D. D., Lin, S.-H., Scott, B. A., Jackson Walker, E. M., & Wang, M. (2017). Regulatory focus trickle-down: How leader regulatory focus and behavior shape follower regulatory focus. Organizational Behavior and Human Decision Processes, 140, 29–45. · Applied to transformation: Ruth Pauline Wachter (2026).

This is the Level 2 paradox, and it's the most misread moment of a disruptive transformation. You introduced AI agents to take work out — and in the deep middle of the crossing, the work briefly climbs instead. It isn't a sign the transformation is failing. It's a sign it's working, and that you cannot stop here.

The mechanism is simple once you see it. Agents run around the clock; they don't tire and they don't wait. So they push more of the existing business through the value chain, faster — and every output still lands on the humans left in the loop to check it. The remaining user-in-the-loop and human-in-the-loop roles, sized for the old volume, are now the bottleneck for a volume several times larger, running 24/7. More AI, more throughput, more verification load on fewer people. The paradox isn't that AI created work. It's that AI scaled the work and left the humans at the old scale.

This is why a half-finished disruptive transformation is the dangerous place to stand: you have taken on the cost and the strain of agents, but stopped before the step that resolves it. The resolution isn't fewer agents — it's agentic AI: agents orchestrating agents, so the verification and coordination that's drowning your people is itself carried by the system. That is not a luxury upgrade. At Level 2, it is the only way through to the far bank. Stop short, and you have built a faster machine that burns out the few people still holding it together.

I wrote about this at length, from inside my own company, in the European Business Transformation Review.

A structural shift. The useful question isn't whether to engage, but at what pace and under what governance. The hype is real — most vendor claims dramatically overstate what AI delivers in a specific organizational context. But the underlying capability curve isn't reversible. Organizations that build genuine AI competence now — not just buy AI tools — gain a compounding advantage. And the risk isn't moving too slowly. It's moving too fast: deploying without verifying the architecture, without governance, without understanding what the model actually does.

Four pressure points are converging at once.

Competition stopped being a slow game. Rivals with AI-assisted pricing, demand forecasting, and segmentation operate at a structurally different speed — and the gap compounds every quarter they're ahead.

Customers expect real-time everything. Delivery status, invoice queries, stock, order changes — instant, accurate, self-service. Companies still running these through manual or batch processes aren't just slow; they're losing customers to whoever isn't.

Ecosystems moved to live data flows. Supply chains, logistics, financial systems — those who exchange live signals across their ecosystem have a structural cost and reliability advantage. AI is the layer that makes that data actionable, not just visible.

New AI-native products become possible — but not yet for most. Services where the intelligence is the product, not a feature, require a verified, stable operating model underneath. Without it you're building on sand. Genuine AI-product capability begins at Westerman's Level 2 — repeatable, governed AI built into core operations. Level 1 organizations that try to build AI products aren't innovating; they're accelerating their own instability. This is why Big T transformation is the prerequisite, not the outcome. The companies leading in AI-enabled products in 2027 are the ones redesigning their operating models now — not the ones buying the most licences.

Because most of them are a shortcut — and the shortcut is exactly what doesn't work.

The promise of AI agents is real: they can move cost out and lift EBIT margin. But that effect only appears when the operating model underneath has been rebuilt to let them work — silos removed in both process and tools, one version of the truth, the value chain redesigned for how the work now flows. That rebuild is disruptive transformation. Most companies skip it. They want the margin effect of agents without the disruptive change that produces it, so they bolt AI onto the same broken model — onto silos, onto three conflicting versions of the data, onto a process no one redesigned — and the agent inherits all of it. The result isn't transformation. It's an expensive layer of intelligence sitting on an operating model that was never built to carry it. MIT's NANDA research found that 19 in 20 enterprise AI efforts deliver no measurable return — and this is the largest reason why.

Three further reasons compound it. First: AI is procured as a product when it has to be integrated as a capability — a licence is not organizational AI literacy. Second: the ROI framing is wrong — AI is measured against cost reduction, when much of its value is decision-quality, which takes longer to show. Third: organizations skip the verification layer — they act on AI outputs without the capacity to know when the AI is wrong. That's the Verification Tax: the hidden cost of catching errors no one budgeted for.

The pattern underneath all four: AI doesn't rescue a broken operating model. It exposes it — faster, and at scale.

Risk slope & levels — Webster & Westerman (2025), MIT Sloan. Naming, transformation markers & the revenue/EBIT/risk overlay — Ruth Pauline Wachter (2026).

AI governance is the framework that decides who deploys which AI tools, how outputs are validated, how errors are handled, and who bears liability when AI gets something wrong. Boards should care because AI failure is now an executive risk, not a technology risk — regulatory exposure, reputation loss, and operational dependency on unverified systems are all governance failures. A board without a clear AI governance position is one incident away from a very uncomfortable conversation with shareholders.

Most organizations sit at one of three stages, yet most governance frameworks collapse them into one.

Paper-in-the-Loop (PITL). The process runs on paper — approvals, sign-offs, physical handovers. Before any AI conversation is relevant, this has to be resolved. You cannot govern AI outputs in a process that isn't yet digitalized. PITL is not a starting point for AI; it's a prerequisite problem.

User-in-the-Loop (UITL). The process is digitalized and a human sits inside the tool — clicking, approving, forwarding. Technically in the loop, but transactional: the system presents, the user confirms, judgment is minimal. This is the most common state in European mid-sized companies, and it is frequently mislabelled as Human-in-the-Loop. It isn't.

Human-in-the-Loop (HITL). A human applies genuine judgment before a consequential output triggers an action. The distinction isn't where the human sits — it's what they do. Someone rubber-stamping AI output is User-in-the-Loop. Someone who reads, evaluates, and takes responsibility is Human-in-the-Loop.

The governance implication: the standard question — "is there a human in the loop?" — is the wrong one. The right one is: is that human exercising judgment, or are they a click in a workflow? Before deploying GenAI in any consequential process, an organization needs to know which stage it's actually at — not which it believes it's at.

Not with tool selection — that's procurement, not strategy. Start with what the organization wants to achieve and what it needs to get there. The right first questions: where do the biggest quality losses happen today through manual, error-prone processes? Where is decision speed insufficient? What data do we have, and what do we need? Then the governance question: how do we ensure what AI produces is reliable enough to act on?

The clearest signal is what I call Transformation Delivery Dissonance: your CIO and external consultants report on track, your functional leaders nod in steering committees — and your frontline rolls their eyes when no one senior is watching. As a C-level you then have three legitimate responses, and one that ends the transformation.

Add information to one side. Give leadership unfiltered ground truth — skip-levels, anonymous pulse checks, direct stream-call access — or give employees the strategic context they're missing, so their worry is informed rather than reflexive. Often the dissonance clears when one side simply knows what the other knows.

Change the enabling conditions. People won't raise concerns where concerns have historically been punished or ignored. Build explicit raise-your-hand mechanisms into the structure — escalation slots in stream calls, anonymous red-flag channels, a standing agenda item called "what is not working." Make concern-raising the expected behaviour, not the brave one.

Address the culture gap directly. Some dissonance isn't about information — it's about risk tolerance. Transformation requires risk affinity. If your organization operates at 120% risk avoidance — only raising hands when failure is certain — that's a culture problem inside the transformation, not a communication one. Name it and run it as a culture track.

The one response that ends it: trivialization. Deciding the eye-rolling is just resistance, that the functional people don't get the bigger picture, that a better town hall will fix it. Trivialization is how leadership converts recoverable dissonance into a failed transformation.

Six early indicators. (1) Leadership can't state the transformation logic in one sentence. (2) The transformation has its own team, but the line organization hasn't changed its behaviour. (3) Milestones are met on paper while the underlying capability doesn't develop. (4) The external consultants are running the program. (5) The board gets progress reports but no escalations. (6) Employees describe the transformation as something happening to them, not something they're part of.

The systematic intervention when a transformation has stalled, is actively failing, or has produced the wrong outcomes despite significant investment. Not a consulting engagement that produces another strategy deck — an operating-level mandate to diagnose what broke, stabilize what can be salvaged, and rebuild execution capacity under real accountability. It requires executive authority, not advisory authority. Someone has to be willing to make the decisions the previous team couldn't or wouldn't.

First, distinguish stalling from failure. Stalling is recoverable — missing resources, unclear priorities, a governance gap. Failure means the underlying assumptions were wrong. In both cases the worst response is to request more reports and wait. The right one: an external diagnosis from someone with no interest in justifying the previous approach — then a clear decision: continue, correct, or stop.

An interim CEO takes operating authority during a leadership gap, crisis, or transition — and owns outcomes, not recommendations. A Chief Transformation Officer holds structural accountability for delivery across the organization, with authority to intervene at any level where the transformation is stalling. Both are substantively different from a consultant: the mandate is to make things happen, not advise on what should happen. When a transformation is already behind, that distinction matters enormously.

Four conditions tend to decide it, and none of them are about competence. First: the people who would lead the change are the same people who designed the situation that needs changing — not a flaw, just a blind spot no one sees from inside. Second: the transformation requires decisions that are politically impossible for an internal candidate, who still has to live with everyone afterward. Third: the organization needs credibility with outside stakeholders — investors, supervisory board, banks — that an internal appointment can't carry. Fourth: the pace required is faster than internal development can deliver. When none of these are present, promoting from within is usually the better, cheaper, more durable answer. The honest test is whether you need capability you don't have, or distance you can't generate from inside. Those call for different solutions.

It's the line between recommending and owning. Advisory authority produces analysis, options, and a deck — and the responsibility for what happens next stays with you. Operating authority means someone is accountable for the outcome, with the standing to make the decisions the situation demands, not just describe them. Both have their place: when the problem is genuinely unclear, good analysis is worth more than premature action. But when a transformation is stalling, the missing ingredient is rarely more analysis — it's someone with the authority to act on what's already known, and to carry the consequence of acting. The question to ask of anyone you bring in isn't "what will you tell me?" It's "what will you be responsible for?" The answer sorts advisors from operators in one sentence.