Engineering Beyond Agile: The CTO Role Has Outgrown the Org Chart
Ep.10 - Three datasets, one pattern: the bottleneck is the people architecture.
The job used to be legible. You owned the technical direction. You managed a hierarchy. You reported to the CEO. The salary staircase priced the role by the size of headcount that sat beneath you: manage 101–200 people, earn a 90% premium. That made sense when complexity scaled with headcount.
It doesn’t any more. And CTO Craft × Albany’s 2026 Compensation Survey just produced four years of data proving it. B
ut there was more standing out to me.
The role is changing faster than the structure around it
Three things surfaced in my view this week, and they all point at the same gap.
Jay Parikh, EVP at Microsoft, laid out a six-step agent platform: Build, Contextualise, Run, Govern, Improve, Surface. Not a concept deck or roadmap. But an operational stack for running agents at enterprise scale. The factory floor is built. It’s designed to compound. Agent behaviour, outcomes, feedback — all of it flows back into the system. Improves under human oversight. Parikh’s own framing: “AI alone won’t change your business. The system running it will.” The platform is the floor, not the ceiling. What sits on top of it — the operating model, the governance, the people who design the system — is where the work lands. That work didn’t exist three years ago. Now it’s the CTO’s job, and nobody updated the job description.
Microsoft’s 2026 Work Trend Index tells you why that work is so hard. Only 1 in 4 AI users say their organisation has clear leadership alignment on AI. The WTI’s own manager study found that when lea
ders actively model AI use, teams report a 30-point lift in trust and a 22-point lift in critical thinking. Without that signal from the top, adoption stalls. Agents in the M365 ecosystem have grown 15x year-over-year — 18x in large enterprises. The agents are already there. The question is whether the organisation has noticed.
The WTI’s recommendation: “redesign workflows before applying AI.” Fair. But who does the redesigning? Microsoft’s own answer was Camp AIR, a three-week internal programme where product dev teams stopped feature work, mapped their workflows, and redesigned how work moved. It worked. Cycles compressed. Prototyping accelerated. But that’s Microsoft, a company built on change and growth mindset.
If the three week program ends, what happens in week four? When the protected time ends, the coaches leave, and the org chart reasserts itself? A three-week intervention produces a three-week result unless the structure around it changes. That is a CTO-level problem. An architectural problem. And it lands on a role the org chart still describes as “manages engineering.”
Then Albany’s data: 84% of AI accountability sits at CTO level. 36% at EM level. The CTO now owns AI strategy, engineering delivery, architecture, and org design. A scope that has expanded in every direction while the title and the comp benchmark haven’t moved.
My conclusion: The role has outgrown the box it sits in.
What faith and money are actually diagnosing
CTOs don’t leave for money. 68.5% of departing CTOs cite loss of faith in leadership or strategy. The CTO who stops believing in the Intent of the organisation, its strategic direction, its never-compromise list, is not underpaid. They are unaligned. When that faith breaks, no raise repairs it.
CTOs own Intent. They sit where strategic direction is set — where the organisation decides what it will not compromise on. When Intent is incoherent, they leave. It’s that simple. They can’t trace a line from the board’s priorities to the engineering team’s work, so they stop trying.
I wrote about this in “The Manager Becomes Five.” When you flatten the EM layer, the five accountabilities do not disappear. They scatter. Architectural direction migrates into the spine, the codified trade-offs that make the operating model enforceable, if you have built one. Context routing dissolves into the cell model, if you have adopted it. Scrutiny is absorbed by the evidence surface, the telemetry that tells you whether the system is doing what you intended, if it exists.
And if it does not? The Director inherits three of the five. They get loaded up with work the structure was supposed to absorb. And they get paid a Director’s salary while carrying a load the organisation has not priced.
The CTO who stays and stops caring
The stickiest retention signal in the dataset is not salary. It’s equity. Stayers hold LTIP at 77%, leavers at 63%. Albany’s own verdict: boards that respond to flight risk with a salary increase are solving the wrong problem. The data has said so for four consecutive years.
Comp surveys measure retention. They do not measure alignment. Equity retains people. It does not align them.
A CTO with a vesting schedule stays. They attend the board meetings. They present the roadmap. They hire the engineers. But a CTO who stays without faith in the strategy, who no longer believes the Intent they are nominally responsible for, is worse than one who leaves. They coast. They defer hard decisions. They stop challenging the board. The political cost isn’t worth it when the golden handcuffs are doing the holding.
Equity buys retention. Intent buys alignment. The comp survey measures one of those. Ghost Decisions is about the other.
The most expensive CTO is not the one who leaves. It is the one who stays, stops believing, and lets the ghost decisions accumulate. Choices nobody makes explicitly. Intent leaking into Execution without passing through Structure, because the person who should be authoring the spine has stopped caring whether the spine is right.
The load rises, the structure doesn’t
CTO burnout is disproportionate in the Albany dataset. When the structural layer thins, the CTO absorbs what used to be distributed. They do not work more. They carry more.
The “Manager Becomes Five” argument works in reverse. The five accountabilities that fragment out of the EM role do not always land cleanly in the spine, the cells, or the evidence surface. In organisations that have flattened the middle without building the structure to replace it, those accountabilities could float upward. The CTO becomes the context router, the architectural memory, the career sponsor. They’re also doing resource allocation and product arbitration, because who else is going to? All at once. All again. At a higher altitude with a larger blast radius.
Parikh’s platform is designed to absorb exactly this. The governance layer, the continuous improvement loop, the observation and eval infrastructure. But the platform absorbs the technical load. It does not absorb the organisational load. Who sets the Intent that the platform enforces? Who redesigns the workflows the WTI says need redesigning? And who’s authoring the spine? That is still a person. That person is the CTO. And 38% of them are considering leaving, a number that has barely moved in four years of this survey.
Four years in, that stops being a trend and starts being a feature of the system. Now add the Agentic transformation to that, and the load doesn’t just rise — it changes shape.
The org chart doesn’t see this
The salary staircase prices headcount. A CTO running 200 engineers in a traditional hierarchy earns the premium. A CTO running 30 people organised into 10 cells, each context-complete, each shipping inside the spine, each emitting evidence that closes the loop, earns half as much, by the benchmark.
The second CTO is doing harder work. More consequential work. They are designing the operating model the first CTO’s organisation will eventually need to adopt. But the org chart doesn’t measure operating-model design. It measures the number of people below you on the chart.
The structural answer isn’t a raise, a retention package, or a three-week programme. It’s a redesign of how the role connects to the organisation: what the CTO is accountable for, what the spine carries so they don’t have to, and what the evidence surface makes visible so faith isn’t required — proof is. I’ve been building that architecture. More on that soon.
The agents are scaling, the platform is built, a playbook’s been published on the intranet. The question is whether organisations will redesign the box the CTO sits in, its scope, its structural support, its accountability boundaries, or keep describing the role with an org chart that was drawn before agents existed.
The role has outgrown the chart. Data says so. The leaders I met at DeveloperWeek Management in San Francisco said so. And they will surface this, possibly by departures.
Not for money. For faith.
More goodness in the book
This piece is part of the thinking behind my book, Ghost Decisions: How to Lead When AI Moves Faster Than You → https://ghostdecisions.com”
