Writing
The books made the argument that people, not technology, decide whether transformation works. The pieces below carry that argument into the AI era. Longer whitepapers first, shorter essays below.
How these are made
I use AI to help me research and write all of these papers. I built the process so I can keep up with what's happening in the field, and I'm sharing what it produces. If you'd like to know more about how the process works, get in touch.
The papers are © 2026 Gerald C. Kane and licensed under CC BY-NC 4.0: share and adapt them for noncommercial use with credit to me as the author, but don't present them as your own.
The Transform Imperative
Why the Real Value of AI Lies Beyond Efficiency, and What C-Suite Leaders Must Do About It
Most enterprises use AI to make existing work cheaper, and competitors can copy that within a year. Building on Zuboff's distinction between automating, informating, and transforming work, this paper argues the three layers stack and that durable advantage sits in the third. It covers how to split the AI portfolio across the layers, how a CFO defends the productivity J-curve, and why the C-suite can't delegate the transform layer.
The Relational Premium
Where Durable Enterprise Value Migrates When AI Makes Production Cheap
As AI absorbs the technical and transactional layers of professional work, value moves to the work that stays human: judgment, trust, and the relationships clients actually pay for. This paper makes the case for where that value lands and what it means for how firms staff and develop people.
Nobody Gets Promoted for Deleting the Report
The Third Bin in Every AI Task Inventory
AI task inventories sort work into automate and keep. They leave out a third bin, work that should stop, because eliminating it produces no artifact, has no owner, and earns no one a promotion. The paper proposes a Stop Inventory and four rules for running one: run it narrow, eliminate before automating, decide in advance where the freed hours go, and keep work decisions separate from people decisions.
What the vendors are changing
The Headless Enterprise
How Agent-First Software Will Reshape Enterprise Strategy, Pricing, and Power
The per-seat pricing model that has defined enterprise software for twenty years is repricing, because the agent is becoming the user. What the vendor shift to agent-first software means for pricing, portfolio, and governance decisions, and why the next twelve to twenty-four months are the window to make them on your own timeline instead of the vendor's.
The End of the AI Subsidy Era
How Metered Pricing, FinOps Discipline, and Architecture Choices Will Decide Who Captures Enterprise AI Value Through 2027
Enterprise AI has been priced below what it costs to run. That's ending, and usage-based billing is the first visible sign. How metered pricing, cost discipline, and architecture choices will decide who captures AI's value through 2027.
They Dropped the Exit Fee
Continual Learning Was Supposed to Give AI Labs Their Moat. The Contracts Already Say Otherwise.
Continual learning was supposed to lock enterprises into the AI labs. The contracts point the other way: buyers now pay to keep their data out of training, and the records from ERP, health-records, and cloud migrations show that switching costs sit inside the organization. Five actions for any board being asked to sign away training rights, starting with putting a number on contractual versus organizational switching cost.
It Only Has to Fail Once
Why the Expected Value of Automated Defense Depends on the Blast Radius
Ben Thompson argues that automated defense carries negative expected value for incumbents because one bad change can take everything down. Read through Perrow's work on complexity and coupling, the expected value turns out to depend on design: how much of the system a single change can reach, and what unpatched vulnerabilities cost while they wait. What senior leaders should decide in the next ninety days about AI that changes production software.
Whose Permission Does Your Incident Response Need?
A Model Provider's Safety Classifier Now Decides What Your Responders May Examine
When a company runs incident response on a vendor's AI model, the vendor's safety classifier decides what evidence responders may examine about the company's own systems. No one on the org chart owns that dependency. Drawing on the July 2026 Hugging Face intrusion, the paper sets out what boards, CISOs, and audit committees should settle before the next incident, from a self-hosted fallback model to written decision rights for containment.
When the argument becomes an operating question
These papers are public arguments. The client work begins when a leadership team needs to decide what one of them means for its own strategy and work.
Send me the decision in front of your team. I'll tell you whether a briefing, diagnostic, or longer engagement makes sense.