Ethan Mollick at Wharton has been direct about something that most AI adoption narratives miss: AI use in companies is, above all, a leadership problem — not a technology problem, not a training problem, but a leadership problem involving fundamental questions about what people should do with their time, how work is organized, and how to keep human judgment at the center of AI-augmented work.
That framing explains something that surprised a lot of observers: fractional CTO demand has risen, not fallen, as AI tools have proliferated. More fractional technology executives were placed in the past twelve months than in the previous three years combined, according to recent LinkedIn data on fractional executive placements. The instinct — that easier code generation means less need for technology leadership — turns out to be exactly backward.
flowchart TD
A[Technology challenge] --> B{What kind<br/>of problem?}
B -->|Don't know what to build| K[Knowledge gap]
B -->|Decisions keep stalling| L[Leadership gap]
K --> KA[Technical advisor<br/>or consultant]
L --> FC[Fractional CTO]
FC --> R[Accountable executive<br/>owns the outcome]
class FC accent
class R good
class KA warn
classDef good fill:#163a26,stroke:#44cc77,color:#d7ffe6;
classDef bad fill:#3a1620,stroke:#ff5555,color:#ffd9d9;
classDef warn fill:#3a2e16,stroke:#ffaa33,color:#ffe9c7;
classDef accent fill:#15233b,stroke:#4488ff,color:#dce9ff;
The Misconception: Easier Code Means Less Leadership Needed
The logic was seductive. If AI can generate working code from natural language descriptions — and it can, to a degree that would have seemed implausible three years ago — then the most expensive and scarce ingredient in technology delivery gets cheaper. That should reduce the cost of building software, which it does. The leap was that it would also reduce the need for someone to lead technology strategy. That hasn’t happened.
The reason is that code generation has never been the core constraint in technology leadership. The hard work of a CTO is not writing code — it is deciding what to build, in what sequence, with what architecture, and with what team. It is evaluating vendor commitments, assessing technical debt, managing the organizational dynamics of engineering teams, and reporting technology risk to the board in terms that non-technical leadership can act on.
AI tools do not make those decisions. They surface more decisions faster, which is the opposite of reducing executive workload.
What Actually Increased in Complexity
Three things that didn’t exist as executive concerns five years ago are now standard issues for any company building software in 2026.
AI vendor commitment decisions. A company evaluating whether to build on a specific model provider’s API — whether to bet core product functionality on Claude, GPT-4o, Gemini, or an open-weight alternative — is making an architectural decision with multi-year implications. Switching costs are real. Data exposure obligations differ by provider. The pricing structures are not stable. An executive needs to own that decision and its consequences — not a developer choosing an API, and not a vendor’s sales team.
AI-generated code governance. When developer output increases significantly because AI is generating substantial portions of the codebase, the review, testing, and architecture coherence problem scales with it. Code that ships fast but nobody fully understands accumulates what some practitioners are calling comprehension debt — systems that work until they don’t, with no one who knows how to fix them without starting over. Managing that risk is an executive function.
AI platform strategy across non-engineering departments. Marketing teams, finance teams, customer success teams — all are purchasing AI tools independently, building prompting workflows, and automating processes without coordinating with engineering or IT. The result is a shadow AI infrastructure that creates data exposure, integration fragility, and compliance risk. Identifying it, governing it, and directing it toward actual business value requires someone with cross-organizational authority. That’s a CTO function, and it doesn’t appear on any AI vendor’s feature list.
The Fractional Model Fits the Current Demand Pattern
Most mid-market companies evaluating fractional CTO engagements in 2026 are not replacing a vacant full-time role. They are adding executive capacity to handle a category of decisions that didn’t exist, or wasn’t complex enough to require dedicated leadership, three years ago. That’s a pattern the fractional model handles well.
Consider the kind of business this favors. Years back I led architecture on a multi-year legacy migration at Geologistics, a global logistics enterprise running about $1.5B in revenue across 1,000 locations in 140 countries. The work itself was converting roughly 400 AS400 screens into a modern web platform. But the operational shape of that business is exactly the profile that today carries the most AI optionality. Routing decisions, exception handling, customs documentation, demand forecasting, freight classification. Every one of those is now a candidate for AI augmentation, and every candidate is a decision that compounds across regions and regulators. A company at that scale absolutely needs senior judgment when the options multiply, but rarely needs (or can justify) a full-time CTO on payroll. That gap is precisely where fractional engagements have been landing this year.
At LERETA, the second-largest property tax payment processor in the U.S. — processing roughly $18B annually — the architecture work that justified a $20M board investment in legacy modernization began with someone in the room who could translate technical reality into a form the board could evaluate and commit to. Not a developer. Not a consultant producing a report. An embedded technology executive who owned the narrative and the roadmap. That’s what the current AI-decision environment requires from mid-market companies that don’t have it yet.
When vendors, developers, AI integrators, and IT staff roll up to one accountable technology leader, the organization makes informed decisions instead of accumulating technical debt quietly across disconnected teams. That oversight function — once reserved for companies large enough to justify a full-time CTO — is now table stakes for any mid-market company making serious AI investments.
The Leadership Gap Is the Actual Constraint
Mollick’s framing is worth sitting with. The organizations struggling with AI adoption are not struggling because their developers lack access to AI tools. Most have access. They are struggling because nobody in an executive seat is accountable for AI strategy, AI governance, and the organizational changes that meaningful AI adoption requires.
That’s a leadership gap. Fractional CTOs are filling it — not by writing code, but by answering the questions that AI tools don’t answer: what to build, how to govern it, who is accountable for the outcome, and how to communicate all of that to a board that is increasingly asking questions the leadership team cannot currently answer.
If you are evaluating whether your organization has a knowledge gap or a leadership gap in technology, a direct conversation is the fastest way to find out.
