TS: Transformation Theatre
Menu
Home
Archives
Tags
About Me
RSS
July 16, 2026
Transformation Theatre
TL;DR: A company's own job postings tell you whether its AI transformation is real. Most senior AI roles hire people to coordinate other people around the technology: a new layer of human go-betweens to manage the thing that removes the need for go-betweens. That is automating the status quo, not rebuilding. A real rebuild works differently: the knowledge that lives in the heads of the middle (why each control exists, which invisible correction is required) has to be captured into the AI systems first, and headcount falls only as the capture completes. Doing that requires people who understand why things work, not just how the forms move; the technology can reproduce every form for free, so understanding is the only thing left worth paying for. And it requires the firm to redesign itself, a decision no delegated department can make, yet firms absorbed every previous technology wave precisely by delegating it to a department, so demolition-and-rebuild is not in the repertoire. Watch what gets deleted, not what gets hired.
Contents: What the Job Postings Reveal · The Rebuild Is an Encoding, Not a Deletion · Syntax People, Semantic People · A Firm-Level Problem Wearing an IT Badge · The Missing Playbook · The Test That Matters<br>The clearest evidence of whether a company has started its AI transformation is sitting in public view: its own senior AI job postings. The last post argued that real transformation runs shrink, rebuild, expand, and that the shrink comes first, because you cannot rebuild an organization's information flows while thousands of people maintain the old ones. No shrink, no transformation. This post applies that diagnostic to where organizations actually are today. Almost everyone is automating the status quo rather than rebuilding, and the nibbling is not stupidity. It is the predictable output of how these firms are led, what they are incentivized to do, and what their management's careers never taught them.<br>What the Job Postings Reveal<br>If the cycle is real, you should be able to see where organizations are in it just by reading their job postings. So I did: three senior "AI" roles at three of the most sophisticated institutions in the same market, posted in the same period. Three postings are an illustration, not a survey; what they illustrate is the same assumption, three times over.<br>The first: "Director of AI Products." The verbs are align, coordinate, translate strategy into enablement plans, drive communication, lead through influence across matrixed teams: a chief of staff for AI activities. The role doesn't build anything. It routes knowledge about AI between the people who do. This organization is responding to AI by constructing a brand-new middle layer: hiring a human router to manage the technology that eliminates human routing.<br>The second: an "AI Technology Director." Closer to the edge: the posting wants an owner, someone who writes requirements and reviews code, takes prototypes to production. Better. But the job itself is to establish the software development lifecycle, CI/CD handoff processes, and reusability standards, and to coach juniors to execute within established frameworks. Every one of those exists to coordinate many human developers. The role optimizes the delivery pipeline, and delivery itself is the old paradigm: a producing team ships software to a consuming business.<br>The third: a senior AI architecture role. Ten-plus years of enterprise architecture, strength in stakeholder management, and the giveaway phrase: "review and govern solution designs through architecture checkpoints." A checkpoint is a coordination gate: it exists because downstream human teams can't be trusted to stay within constraints on their own. This is the translation-and-gatekeeping layer, staffed and formalized.<br>Three institutions, three senior AI roles, and not one describes the roles the future actually needs. All three assume the stratified model: a software team delivers, a business consumes, and humans in between route, gate, and translate.<br>What's missing? In the rebuilt organization, the central technology team doesn't deliver finished applications at all. It delivers primitives (data access, security guardrails, identity, compute, cost controls), a paved road. The working software is then grown and continuously evolved at the edge, by the people who own the outcome, directing AI to do it. Software stops being a product that ships and becomes a living practice. "Production" as a sacred, gated state stops making sense when the owner can regenerate an application as easily as editing a document. Governance doesn't disappear. It moves out of meetings and into the platform, enforced by the substrate instead of by checkpoints.<br>We spent thirty years treating this pattern (business users building their own tools) as a pathology called shadow IT, because unaided humans built...