Jeff Clarke, Dell's CEO, made the structural argument from the DTW 2026 keynote stage with the bluntness it deserves.
"The error some companies will make is they're going to take agents and try to put it on their existing process, and it'll make them better, and you're going to get incremental productivity gains — 20, 30, 40%. But if you want the productivity gains of 10X, 100X, you have to rebuild the workflow."
The proof point under the abstract claim was concrete. Fifty Dell engineers spent a year writing a piece of software shipped today on Alienware notebooks. Two AI engineers replicated thirty percent of its functionality in a week. That contrast is the rewire argument compressed into a single sentence and the receipt nailed to it.
This is the structural half of the rewire. AI sitting on top of an unchanged operation yields twenty to forty percent. The orders-of-magnitude returns the AI narrative has been promising for two years require the workflow to be rebuilt around the assumption that intelligence is now a primary input the operation can consume. Dell took three years to do this work — simplify, standardize, and automate the company first; then connect the data sources; then build the agentic framework; then expand broadly.
Most companies skipped the sequence. Most companies are still on incremental returns.
The previous piece on this site named this as the explanation for every stalled pilot. The argument is correct.
It is also half of the picture.
The second seventy-percent number
Julia Dhar runs the behavioral-science practice at BCG. In a Fortune piece this month she names a finding that belongs alongside McKinsey's twenty-one-percent rewire data in every conversation about AI transformation.
The corporate-transformation failure rate has held at roughly seventy percent for forty years. It has not improved despite better strategy frameworks, better technology, and better consulting methodology. The diagnosis, drawn from six thousand surveys, fifty executive interviews, and fifty years of behavioral research:
"Change doesn't fail because people resist. It fails because leaders misunderstand how people really change."
The specific mechanism is the false-consensus effect. Around seventy percent of executives report feeling positive about a change they know nothing about. Leaders assume their disposition is universally shared. Employees more commonly feel anxious, overwhelmed, or frustrated. The asymmetry produces a recurring pattern: executives believe alignment exists; employees experience the change as an imposition; the rollout stalls; the executive concludes "resistance to change"; the cycle repeats.
There are now two seventy-percent numbers in this discussion. Both are doing structural work.
The composition
The rewire is structural and behavioral. The structural work is the rebuild itself — workflows decomposed, AI given a sensible job, the foundation built underneath. The behavioral work is the design of how the rebuild lands in the people who will execute it daily.
The two halves are not optional. Skipping the structural work produces the seventy-nine-percent McKinsey number: most companies see no enterprise return because they installed AI on top of an unchanged operation. Skipping the behavioral work produces the seventy-percent BCG number: most companies that begin a rebuild fail to land it because the leadership treated the case as self-evident and skipped the design of the employee experience.
A company can fail at either independently. The companies that win the AI era are the ones doing both halves.
What behavioral failure looks like inside a structurally sound rewire
It is worth being specific about what behavioral failure looks like inside an AI rewire that is otherwise correct.
The leadership has read the literature. The workflow has been genuinely redesigned. The AI has been deployed inside the redesigned process. On paper the new workflow is faster and the metrics confirm it.
But the supervisors do not trust the AI's recommendations because they were not part of designing what good looked like. The high-skill performers feel their judgment is being supplanted by a system whose reasoning is opaque to them. The new metrics have been published, but the team is still being evaluated against the old ones. The training has been delivered, but the employees do not feel competent against the new expectations. The rollout produces friction the leadership cannot see because — by the false-consensus effect — they project their own clarity about the change onto the people working inside it.
Six months later the AI is in place, the dashboards show it is working, and the team is operating around it rather than through it. The leadership concludes AI does not fit the culture. The diagnosis is wrong. The structural work was right. The behavioral half was missing.
The prescription
Dhar's mindset shift is operator-altitude and worth borrowing verbatim:
"Leaders of a successful change treat employees as the customers of that change. They obsess about their people's experience of change, just as they obsess about their customers' experience of their product."
Every senior operator already understands customer experience as a category of work. Mapping the same discipline to internal change reframes the rollout from "communicate the plan" to "design the change so the people in it succeed."
Two of the behavioral mechanisms Dhar names are immediately usable. The IKEA effect — employees who contribute to the design of the change commit harder to executing it; the team that helped design the AI rewire defends it the way they defend any system they built. The endowed-progress effect — leaders who surface early wins compound momentum; the team that sees the first three small improvements generates the next ten itself.
These are not soft skills. They are the design of the work surrounding the technical rewire.
The service-company translation
A mid-market service company executing an AI rewire has both failure modes available to it, and the behavioral burden is heavier than at Dell's scale.
Dell took three years to get the structural foundation in place. The behavioral work could lean on a strategic foundation that had been built deliberately over the same window.
A service company doing the rewire on top of an operation that has never been simplified, standardized, or instrumented is doing both pieces of foundation work simultaneously. The structural rebuild is harder because the starting point is messier. The behavioral burden is heavier because the change is more visible to the people inside the work — they feel both the operational shift and the technical shift in the same six months.
This is why the rewire conversation in service companies cannot stay at the structural altitude where the Clarke and McKinsey literature lives. The structural argument is correct and necessary. The behavioral work is what determines whether the structural correctness ever produces the return.
The position
Three independent data points are converging on the same picture. Clarke at Dell names the structural rebuild and shows what it returns at scale. McKinsey quantifies how few companies have actually done the rebuild. Dhar at BCG names the behavioral mechanism that explains why most attempts at the rebuild stall mid-execution.
The previous piece argued that most companies fail because they skip the rewire. This is true.
The companies that begin the rewire and still fail are the second story. They are not failing because the strategy was wrong. They are failing because they assumed the people inside the change felt about it the way they did.
That is the other half of the work.