Tony Wood analyses how agentic AI dissolves the coordination layer that defines modern organisations. The coordination tax, the execution layer shift, the verification flywheel, and what genuinely human work survives..
| Dimension | Traditional UX | Agentic Experience Design (AXD) |
|---|---|---|
| Primary material | Attention and affordance | Trust and delegation |
| User state | Present, navigating | Absent, delegating |
| Design output | Screens and interfaces | Outcomes and constraints |
| Temporal model | Session-based | Relationship-based |
| Success metric | Task completion | Trust calibration |
Agentic AI systems do not automate tasks within existing organisational structures - they delete the need for organisations to be structured the way they currently are. When AI agent harnesses can go directly from insight to code in a single loop, the coordination layer that existed to manage human-to-human handoffs disappears. Not because the handoffs are automated, but because the handoffs no longer happen. The roles that existed to manage those handoffs - project managers writing PRDs, engine
The verification flywheel is a compounding dynamic that emerges when organisations shift their execution layer from human teams to agent harnesses. As coordination roles are removed, the remaining work becomes more verifiable by agents. As remaining work becomes more verifiable, agents can do more of it. As agents do more, less coordination is needed. Each turn of the flywheel accelerates the next. This flywheel pushes organisations relentlessly toward a structure where humans manage teams of ag
The work that survives the coordination tax elimination is the thin layer of genuine judgment under uncertainty: product vision (not the PRD, but the upstream conviction about what the product is and who it is for), brand as deep thought work (not the guidelines document, but what the company means), genuine care in customer relationships, engineering architecture as systems thinking, and designing agentic systems themselves. The typical product leader spends at best 5–10% of their week on produ
Agentic AI systems do not automate tasks within existing organisational structures - they delete the need for organisations to be structured the way they currently are. When AI agent harnesses can go directly from insight to code in a single loop, the coordination layer that existed to manage human-to-human handoffs disappears. Not because the handoffs are automated, but because the handoffs no longer happen. The roles that existed to manage those handoffs - project managers writing PRDs, engine
The verification flywheel is a compounding dynamic that emerges when organisations shift their execution layer from human teams to agent harnesses. As coordination roles are removed, the remaining work becomes more verifiable by agents. As remaining work becomes more verifiable, agents can do more of it. As agents do more, less coordination is needed. Each turn of the flywheel accelerates the next. This flywheel pushes organisations relentlessly toward a structure where humans manage teams of ag
Now count the hours doing the thing you were actually hired to do. Writing the code. Designing the interface. Closing the deal. Building the strategy. Creating something that did not exist before you sat down. If your ratio resembles the research - and it does, because the research describes a structural condition, not a personal failing - you spend roughly sixty per cent of your time on the first two categories and forty per cent on the third. On a good week. That ratio is not a management problem. It is not a tooling problem. It is not a culture problem. It is the Every organisation pays a tax that appears on no balance sheet. The AXD Institute calls it the The coordination tax is not a symptom of bad management. It is a design constraint imposed by the material properties of the execution layer. When the execution layer is human, every participant carries a private model of the world inside their head - partial, decaying, shaped by their role and their recent conversations. No two models are identical. The entire apparatus of organisational coordination exists to align these models closely enough that collective action becomes possible. Meetings exist because humans cannot merge mental models asynchronously. PRDs exist because the person with the insight is rarely the person with the implementation skill. Sprint planning exists because humans cannot dynamically reallocate effort without explicit negotiation. Status updates exist because the state of work, when performed by humans, is opaque to everyone except the person performing it. Strip away the management jargon and what remains is a simple truth: The coordination tax is not a bug in organisational design. It is the price of human collaboration at scale. And it is about to be repriced. Consider what happens when a new engineer joins a team. For weeks, they produce almost nothing. Not because they are incompetent, but because they are absorbing context. Every decision the team made in the past year lives in s