You ran the diagnostic. Now comes the harder conversation.
You know which decisions you are trying to make better. You named the specific questions your AI program needs to answer. You identified where value is actually lost in your marketing organization. You did the work that Vol. 1 described.
Now comes the harder conversation.
Your current operating model is not designed to support those decisions.
Not because your team is not capable. Not because your technology is inadequate. Because the model was built for a different set of questions. The workflows were designed before AI was a factor. The decision rights were assigned based on human bandwidth, not information velocity. The accountability structures were built for a world where data moved slowly and decisions had long lead times.
That world is gone. But most operating models are still running on its logic.
This is the TRANSFORM layer of The Enterprise AI Operating System. It is the most uncomfortable part of the framework. Vol. 1 (ILLUMINATE) was analytical. You gathered information, asked questions, named the gap. That part felt strategic and contained.
TRANSFORM is political.
When you redesign a workflow around an AI-supported decision, you change who owns the outcome. You change which team has the information advantage. You change whose judgment the organization trusts, and how fast. That redistribution of authority is why most AI operating model projects stall in month four. Not because the technology failed. Because the briefing room got quiet.
The organizations getting real returns from AI did not just deploy better tools. They rebuilt the model those tools run on. And they did it before the tools went in.
This article is about how to do that without the project dying in a steering committee.
The numbers on operating model transformation are specific enough to be embarrassing.
McKinsey, Gartner, and the organizations that have already done this work have produced data precise enough to make most enterprise AI programs uncomfortable.
Four numbers. One pattern. The organizations reporting EBIT impact from AI rebuilt the model first. The rest are still waiting for the tools to fix the model for them.
Three things the organizations that succeed do differently.
TRANSFORM is not a technology project. It is an organizational design project that happens to involve technology. Most enterprise AI programs treat it as the former. The ones getting results treat it as the latter.
Map the decision, not the task
The standard workflow redesign process starts with tasks. Which tasks can AI automate? Which steps can we remove? This produces efficiency gains at the activity level. It does not produce revenue impact at the organizational level. Because tasks are not the unit of value. Decisions are.
The TRANSFORM diagnostic starts one level up. For each decision identified in the ILLUMINATE phase, trace the workflow that currently produces the information that informs that decision. Who gathers it? In what form? At what frequency? How stale is it when it arrives? Who synthesizes it? Who presents it? Who approves it?
That workflow is what you are redesigning. Not the task that executes after the decision is made. The process that produces the decision itself.
When Lowe's redesigned its operating model around AI-supported decisions in 2025, the result was a 30% reduction in decision time. Not because the AI was faster than the humans it replaced, but because the workflow it ran on was designed around the decision rather than around the organizational chart. The technology was the last variable in the equation, not the first. Same investment. Different architecture. Different outcome.
Name who loses authority and address it directly
This is the part of TRANSFORM that every consultant skips in the deck.
When you redesign a decision-making workflow around AI, someone loses something. A team that used to synthesize market data loses the synthesis role. A director whose value was knowing which segment to target first loses the information advantage. A VP whose team controlled the briefing loses the briefing.
These are not trivial losses. They are identity losses. And they produce resistance that looks like technical objections but is actually political. "The data quality is not good enough." "The model does not understand our context." "We need more time to validate." These objections are sometimes legitimate. They are very often proxies for a harder conversation.
The organizations that navigate TRANSFORM successfully name this dynamic explicitly in the room. Not as an accusation. As a design problem. The question is not "who is resisting the change." The question is "how do we redesign accountability so that the people who lose one form of authority gain a different, more valuable one?"
The answer almost always involves moving people up the decision chain. The analyst who used to gather data now interprets AI-generated summaries and challenges the model's assumptions. The director who used to synthesize becomes the quality gate on AI-generated strategy. The VP who used to control the briefing now owns the decision architecture itself. Different authority. More consequential authority. That transition has to be designed, not hoped for.
Gartner's July 2025 research found that organizations adapting change plans continuously based on employee responses are four times more likely to achieve change success. That is not change management. That is organizational design. The difference matters more than most AI programs acknowledge.
Run one workflow redesign before you build the program
The most common TRANSFORM failure is scope. An organization decides to redesign its entire marketing operating model around AI simultaneously. They form a cross-functional task force. They hire a transformation consultancy. They produce a 90-day roadmap. Twelve weeks later, the task force is still meeting and nothing has changed.
The high-performing organizations in McKinsey's 2025 dataset did not start with a program. They started with one workflow. One specific decision. One team. They redesigned that workflow completely, ran it for 60 days, measured what changed, and used that result as the proof point to fund the next one.
This is not a limitation of ambition. It is a mechanism for speed. An organization that has successfully redesigned one decision workflow around AI has something most do not: proof. Internal proof that the sequence works. Real numbers from their own operation. A team that has done it once and can train the next team.
One workflow. Sixty days. One proof point. Then the next one.
A transformation program has governance, budget, a name, and a roadmap. It does not have proof. One workflow redesign has proof. That proof is worth more than any roadmap because it changes the internal conversation from "should we do this" to "where do we do it next."
The failure arc for operating model transformation is different from pilot failure. It is louder. And it arrives faster.
Pilots fail quietly, in month ten, when the budget gets redirected. Operating model transformations fail loudly, in month four, in a meeting where a VP uses the phrase "we're not ready."
Stage 1 / Month 1-2: The scope expands before the first workflow is redesigned
Stage 2 / Month 2-4: The resistance surfaces as technical objections
Stage 3 / Month 4-6: The program gets paused for reassessment
Stage 4: The lesson learned is wrong
Same AI tools. Different operating model. Completely different outcomes.
The gap between the 6% getting EBIT impact from AI and the 94% that are not is not a tools gap. It is a model gap. And the model gap traces back to four specific behavioral differences.
Where the framework meets the research.
Four evidence blocks. Named sources. Real outcomes. The same rigor that the ILLUMINATE diagnostic demands from your AI program applies here too.
McKinsey's 2025 State of AI survey tested 25 organizational attributes for their correlation with EBIT impact from AI. Workflow redesign ranked first. Not talent. Not technology investment. Not governance. The redesign of how work actually flows around AI-supported decisions. Only 21% of organizations using gen AI have done it. High performers are 2.8x more likely to have done so (55% vs 20% of the rest). The data is unambiguous: the organizations getting EBIT impact from AI rebuilt the model first. The tools came second.
Gartner's December 2025 survey of 110 CHROs found 78% agree workflows and roles must change to capture AI value. Their April 2026 survey of 782 infrastructure and operations leaders found only 28% of AI use cases fully succeed. The most cited failure factor: organizations expected AI to improve existing processes rather than redesigning those processes first. Organizations that continuously adapt change plans based on employee responses are four times more likely to achieve transformation success. TRANSFORM is not a technology rollout. It is a change management problem with technology as the instrument.
Lowe's did not deploy AI to make its existing decision process faster. It rebuilt the decision process, then deployed AI to support the rebuilt model. The result was a 30% reduction in decision time. Not because the AI was faster than the humans it replaced, but because the workflow it ran on was designed around the decision rather than around the organizational chart. The technology was the last variable in the equation, not the first. Same investment. Different architecture. Different outcome.
Gartner's prediction that 60% of AI projects will be cancelled by end of 2026 due to inadequate data foundations is widely cited as a technology problem. It is not. Data fragmentation, inconsistent governance, and siloed repositories are operating model problems. They are symptoms of organizations where data flows around departmental authority rather than around decisions. The organizations that have solved their data readiness problem did not fix their data infrastructure. They redesigned the operating model that determines who owns which data, who has access to it, and what decisions it is supposed to inform. The data problem resolved as a consequence.
What enterprise leaders ask before they transform.
-
TRANSFORM is Layer 2 of The Enterprise AI Operating System, developed by THE UN KNOWN. It is the operating model redesign phase that follows the ILLUMINATE diagnostic. Once an organization knows which decisions it needs to make better, TRANSFORM addresses what has to change in how the organization actually works before any AI tool is deployed. The core premise: AI deployed on top of a broken operating model does not fix the model. It makes the broken model run faster.
-
The primary failure pattern follows four stages. Organizations broaden a specific workflow redesign into a full transformation program before any proof point exists. Technical objections surface in month two that are actually political objections about authority redistribution. The program stalls in a steering committee when no one can articulate the revenue impact. The post-mortem identifies change management and data readiness as root causes, when the real cause is that the authority redistribution problem was never named or addressed in the design phase.
-
McKinsey's 2025 State of AI survey tested 25 organizational factors for their correlation with EBIT impact from AI. Of all 25, workflow redesign ranked first. AI high performers are 2.8x more likely to have fundamentally redesigned workflows (55% vs 20% of other organizations). AI deployed on top of existing workflows produces activity efficiency. AI deployed inside redesigned workflows produces decision quality, which is what drives revenue impact.
-
An AI-enabled organization has deployed AI tools on top of its existing operating model. It produces efficiency gains at the task level. An AI-native organization has redesigned its operating model around the decisions it wants AI to support. It produces different outcomes, not just faster ones. The decisions themselves are better informed, made faster, and tied to specific revenue accountability. Most enterprise AI programs are building AI-enabled organizations while measuring for AI-native results.
-
Resistance to operating model redesign is almost always a response to authority redistribution, not technology change. When a workflow is redesigned around AI-supported decisions, some people lose their information advantage, their synthesis role, or their position in the decision chain. The organizations that navigate this successfully name the authority redistribution explicitly in the design phase, redesign the new workflow so that people who lose one form of authority gain a more consequential one, and treat the resistance as a design problem rather than a change management problem.
-
The organizations reporting the fastest and most durable results from AI operating model work start with a 60-day sprint on one specific decision workflow. Not a transformation program. One decision. One team. Full redesign, deployment, and measurement in 60 days. The result becomes the internal proof point that funds and legitimizes the next sprint. McKinsey's 2025 research shows that organizations starting with one workflow and expanding iteratively reach meaningful EBIT impact faster than those starting with enterprise-wide programs.
-
The CMO should own the TRANSFORM layer, not delegate it. When IT or a digital transformation office owns the redesign, success is defined by system performance metrics. When the CMO owns it, success is defined by decision quality metrics: did we make the right call faster, and did it produce revenue? The organizations getting EBIT impact from AI have a marketing or business leader in the design chair, with technology as a resource rather than as the driving logic.
-
Data fragmentation and poor data quality are commonly treated as technology problems. They are operating model problems. Gartner predicts 60% of AI projects will be cancelled by end of 2026 due to inadequate data foundations. The organizations that have solved this did not fix their data infrastructure first. They redesigned the operating model that determines who owns which data, who accesses it, and what decisions it serves. The data problem resolved as a consequence of the model redesign.
TRANSFORM is the hardest layer. It asks people to give up something before they have received anything in return. The proof is on the other side. The discomfort is right now.
The Enterprise AI
Operating System.
Get the Full AI Stack by THE UN KNOWN®
No spam. No vendor pitches. Just the framework. Privacy policy
