A note on the quiet summer: the manuscript went through a full editorial pass from June through August, and that consumed the writing hours. It’s back. So is this.
The last piece ended on a claim. The most important leadership skill of the next decade is the skill of engineering context that makes intent executable on contact.
The obvious follow-on question is who does that work.
Most leadership teams answer it as a sourcing problem. Where do we find these people, what do they cost, which competitor is hoarding them, how long is the search. It is a comfortable question because it has a budget line attached and a function that owns it.
It is the wrong question. In most Fortune 500 organizations, the people who can already do this work are inside the building, on payroll, badged, and spending the majority of their week on something else.
The Two-Week Prototype
I have worked alongside several professionals who fit the composite I am about to describe. Call her Maya.
Maya is a senior individual contributor at a Fortune 500 company. Director level, four years in seat, reporting into data and analytics. She has strategic credentials — an MBA from a top program, six years of strategy work before it — and technical credentials, ten years of engineering and data work underneath all of it. She moves between business judgment, technical execution, and strategic framing in the same conversation, against the same artifact. She does in one motion what the operating model was built to accomplish across three meetings and two functions.
The marketing organization had been working a customer segmentation question for months. The traditional path was a six-month engagement: marketing leadership, an outside consultancy, the data team, a series of working groups, a segmentation framework documented in a deck, then a build cycle to operationalize the framework in the marketing platform.
Maya built a working prototype in two weeks.
It ran against real customer data. It produced segments the marketing team recognized as true and had not seen articulated before. It could be pointed at a campaign question and answer it the same afternoon.
Then the operating model did what the operating model does. The prototype was not a decision. It was an input to a decision. To become a decision it needed a business case — the value quantified, the investment sized, the risks enumerated, the governance path cleared, the steering committee briefed. Maya spent the following months producing documentation describing the value of a thing that already existed and was already working.
The prototype kept running because Maya kept it running. Then Maya left.
She was recruited by a smaller competitor whose operating model did not require the translation step. She rebuilt the capability in two months. The competitor launched into the market position the Fortune 500 had been planning around. By the time the business case cleared review, the position was gone.
Not a Culture Problem
The exit interview will not say any of this. The exit interview will say she wanted a more innovative environment, or faster decision-making, or a bigger platform. HR will code it as culture. The pattern will get escalated as an engagement finding.
And the response will be cultural, because that is what the diagnosis licenses. Innovation programs. Employer branding. Hackathons. Dual-career ladders for individual contributors. Refreshed learning pathways. None of these are bad. Not one of them touches the thing that produced the departure.
A professional who joins because the cultural pitch resonated still finds herself, in month four, writing a business case for a capability she has already built and demonstrated. The pattern resumes. The cultural intervention adds cost without changing the structural condition underneath it.
Underneath the cultural framing is a trust question the operating model never names out loud. The business case is a test. It exists to verify that the strategic side of the professional is real — that someone who can write the code can also be trusted with the judgment. She passes the test. The model concludes her strategic capability is genuine. And then it requires the test again on the next artifact, because the apparatus built around the translation step does not have a path that skips it.
Why the Step Existed, and Why It Stopped Making Sense
The translation step was not bureaucratic drag. It was a rational response to real economics.
Building used to be expensive. Validating used to be expensive. A wrong strategic hypothesis discovered in month seven of a build cost millions. So organizations put a deliberate, defensible, document-heavy phase in front of the expensive phase, and staffed it with people fluent in strategy who handed off to people fluent in execution. Two languages, two professionals, two rooms. The handoff was the price of insulating the costly phase from an unvetted idea.
That economics has reversed. Building is cheap. Validating is cheap. The strategic hypothesis now gets tested fastest in the technical phase, not before it. The insulation is protecting against a cost that no longer exists — and charging full price for the protection.
Which means the professional who holds both fluencies at once stopped being a rare luxury and started being the shape the work requires.
And they exist in volume. I have watched this from the classroom side for years. The undergraduate I taught in 2017 learned technical skill as a discipline separated from the business contexts it would eventually serve. The undergraduate I taught in 2023 arrived already prototyping working data products in second-year coursework, against business cases they had chosen from real companies they wanted to work for. They did not learn two disciplines that a career would later integrate. They learned one discipline applied to one artifact, from the start.
What that graduate looks like in your funnel is unfamiliar. The résumé reads too junior for the work they can actually produce. The portfolio shows working capability rather than completed projects. The keywords the applicant tracking system was tuned to weight are not there. The leader who tells me they cannot find both-fluencies professionals in production volume is not wrong about their own funnel. They are wrong about the market. The funnel was built to filter for traditional signal — right schools, right early employers, right titles. These candidates are graduating from different schools, taking different early paths, and building a different signal pattern entirely.
They exist. The recruiting apparatus was not built to see them.
The Same Failure, Twice
So the pattern shows up on both ends at once, and almost never gets read as one thing.
On the retention side, the operating model cannot hold a professional whose work bypasses the step, because the step does not become optional once she proves she does not need it.
On the hiring side, the same model cannot reliably hire her replacement, because the requisition inherits the old role architecture. The job description gets updated with new competencies — translate complex data into business insight, present to executive leadership, advanced SQL and Python, MBA preferred — and every one of those competencies describes the translation work rather than the artifact work. The role takes nine months to fill. The hire matches the description well. Within six months the hiring manager can see the gap, and there is no construct in the operating model for saying the description itself was wrong. So the role gets re-scoped. Expectations come down. The strategic conversations that needed a both-fluencies professional go back to being staffed by translation across two rooms.
The retention failure and the hiring failure are one structural problem seen from two angles. And both are quiet enough, individually, to be absorbed as ordinary friction.
Three Diagnostics You Can Run This Month
None of these requires a program, a consultant, or a budget cycle.
Pull your last three senior technical or data departures. Ignore the stated reason. Look instead at the final six months of calendar and output. Ask what share of that time was spent producing evidence for a conclusion that had already been demonstrated. That share is the alignment tax, and you have a name and a salary attached to it.
Then pull your open senior requisitions in those same functions. Read each description and mark every competency that describes moving intent between two groups of people. If the marked lines outnumber the ones describing what the person will actually build or decide, you are hiring for the step rather than the work.
Then take the last capability someone in your organization prototyped and prove out, and trace the approval path it had to walk before it could be used at scale. Count the reviews that added information, and count the reviews that only confirmed something the working prototype had already shown. The second number is what you are charging your best people to produce.
Context engineering is not a capability you procure. It is a capability you either make room for or spend down.
The talent to run the new operating model is not missing. In most Fortune 500 organizations it is already inside the building, already working with the right data, already producing the thing the model has no role for. The funnel was built to find it. The operating model was built to waste it.
Next in this series: “The Slope.” Two comparable companies. Same markets, same budgets, same leadership pedigree. Three years later one of them is commissioning benchmarking work to understand why it is behind — and the deck it gets back will point at the wrong thing.
Matt Keane is a Chief Data and AI Officer, Professor of Data Science and Analytics, and AI researcher with 20+ years of Fortune 500 transformation experience. His upcoming book, The Alignment Tax, explores how organizations can eliminate the alignment tax and build competitive advantage in the AI era.