Ethan Mollick, whose research on AI in organizations is among the most cited in the field, made the point explicitly in June 2026: “A corporate position that workers should ‘just use AI to do stuff’ has never been enough. AI use in companies is a leadership problem.” He is right, and it explains something I see consistently in fractional CTO conversations right now — the incoming requests are coming from companies that have already deployed AI tools and are realizing the tools don’t self-organize.
The pattern is specific: AI tools are in place, adoption is uneven, and nobody has answered what “good” actually looks like in this particular organization, with this particular team, doing this particular work.
sequenceDiagram participant T as Tooling-only participant Ev as AI Rollout participant L as Leadership-first Note over Ev: Tools deployed company-wide T->>Ev: Configure, train, measure L->>Ev: Assign ownership — fractional CTO Note over Ev: Uneven adoption surfaces T-->>Ev: No owner, no definition of good L->>Ev: Adoption design and workflow changes Note over Ev: Six months in T-->>Ev: AI investment underdelivers L->>Ev: Governance, redesign, ROI visible
What the Tools Don’t Do
Companies who have rolled out AI tools company-wide — Copilot, Claude for Work, ChatGPT Enterprise — frequently arrive at the same place a few months into the rollout. Adoption is variable. The highest performers are getting substantially more value than the average user. Productivity data is difficult to interpret. And nobody has addressed the deeper question of what people should actually be doing differently as a result of having these capabilities.
The tools add capability and leave the organizational design work to leadership. In companies with strong technology leadership already in place, that work gets done. In companies where the CTO role is vacant, part-time, or occupied by someone whose background is primarily technical rather than organizational, the work doesn’t happen — and the AI investment produces far less than it should.
This is not a technology failure. It’s an organizational design failure, and it requires organizational design to fix.
At CloudVirga, the Irvine-based mortgage SaaS company where I served as a senior consultant on their Angular-based loan origination system, the developers were handed virtual development environments — VMs where every dev tool lived behind a remote desktop. Clicking into the VM to write a line of code took three or four seconds to respond. The frustration was constant and the morale cost was visible. The fix wasn’t a better VM image or a faster network configuration. I raised it to the senior directors and pushed hard for physical hardware to replace the virtual devices. The issue became contentious — there were reasons the VMs had been chosen — but the machines were ultimately replaced, and team morale jumped overnight.
The point is the parallel, not the hardware. The unlock wasn’t technical. It was social and political — someone with enough standing had to name the problem, take a position, and absorb the friction of overturning a decision that other people had reasons for. AI adoption problems sit in the same category. The tools are already capable. What’s missing is someone willing to make the organizational call that the tools imply but don’t require — and to defend that call against the cultural inertia that wants to leave the workflow exactly as it was.
What the Fractional CTO Does in This Context
The fractional CTO engagement in the AI era is less about architecture decisions than it used to be and more about three specific functions:
AI adoption design. Not “we have Copilot configured” but “what does a developer’s actual workflow look like now, what’s changed, and where are the highest-leverage adjustments?” This requires someone who has seen it play out across multiple organizations and can apply pattern recognition rather than theory. The answer is different for a 12-person engineering team at a growth-stage company than for a 60-person team at a PE-backed mid-market company, and the tool suite is only a small part of what determines the answer.
Workforce and process redesign. When AI tools meaningfully accelerate individual output, the questions of how teams are structured, how work is allocated, and how performance is measured all change. Most companies haven’t addressed this. Productivity is up in some areas, capacity has increased, and the team structure looks the same as it did before the tools were deployed. The fractional CTO’s role is to push the organizational changes that the productivity gains make possible and that don’t happen without deliberate decision-making.
Governance and policy. Who decides which AI tools are approved for use? What’s the policy on AI-generated code in production? What are the data handling rules for AI tools that have access to company data or customer data? What’s the compliance posture for AI-assisted workflows in regulated contexts? These decisions don’t emerge naturally. They require someone to own them, and they require someone with enough technical depth to make them correctly rather than conservatively.
Why the Fractional Model Works for This
The fractional model is well-suited to AI leadership problems for a reason that doesn’t apply as clearly to other technology leadership work: the problems are strategic and episodic rather than sustained and operational.
Most companies in the AI adoption phase don’t need a full-time CTO presence. They need someone who can establish the policy framework, work through the organizational design questions, and check in as the program evolves. That’s a one to two day per week engagement, not a full-time one. The fractional structure provides the engagement level the problem actually requires rather than a full-time hire solving a part-time problem.
The fractional model also provides cross-company perspective that a full-time CTO rarely has. A fractional CTO working across four or five companies simultaneously sees where the common patterns are — what adoption decisions tend to go well, what mistakes recur, and what solutions travel from one context to another. That pattern recognition is often the most valuable thing the engagement provides. The specific technical judgment matters less than the ability to recognize the situation and know how it typically resolves.
What Triggers the Conversation
In the fractional CTO conversations I’m having in 2026, the most common entry point is a company that has done the tooling work and is now facing the organizational gap. They know the AI tools are capable. They can see that some team members are using them effectively. They can’t articulate why adoption isn’t broader and can’t specify what “broad adoption” would actually look like in their context.
The answer is almost never about the tools. It’s about the organizational design around the tools — roles, expectations, workflows, and governance — and the absence of someone accountable for making those design decisions deliberately rather than letting them emerge from the bottom up.
Reach out if you’re in that position. The first conversation is usually diagnostic — most organizations benefit from simply being able to articulate where the gap is before figuring out how to close it.
