Sensor, comparator, actuator, model: these are not abstract concepts. They are the four structural elements that determine whether an agent fails gracefully or catastrophically. The atlas maps those elements across three operating domains: Software Engineering, Customer Experience, and AIOps.
The Alignment Extension takes the coordination pressure layer further: Innovation Tax thresholds, the Fifth Ceremony intervention model, and the full coordination cost analysis.
What agents can do safely, what they can assist with, and where humans still need to stay in the loop.
Where the work actually lives across touchpoints, backstage processes, systems, and infrastructure.
Where ownership, policy, telemetry, and weak interfaces start slowing everything down, where the Innovation Tax compounds. Alignment here means coordination health, not AI safety alignment.
Every production agent is a cybernetic system. The diagram maps the full control loop that governs agent behavior across any domain. Understanding this structure is what separates agents designed for production from agents that run until something breaks.
Based on Dubberly Design Office / Russell & Norvig
The control loop maps directly to production-grade CNCF graduated projects. This is architecture, not metaphor.
| Role | Function | CNCF Technology | What It Does |
|---|---|---|---|
| Sensor | Percept Collection | OpenTelemetry Cilium (eBPF) | Collects traces, metrics, and logs from the environment |
| Goal | Desired Outcome | Git Manifests SLOs | Defines the target state or performance objective |
| Comparator | Policy Evaluation | OPA Kyverno | Evaluates current state against policy and goals |
| Actuator | Action Execution | K8s Control Plane Crossplane | Executes changes to bring system toward goal state |
| Model | Historical State | etcd Prometheus | Stores world state and time-series for prediction |
Where the work lives across the system for the selected lens.
What agents can handle, what they can support, and where trust still breaks down.
Where the hidden cost of coordination starts to compound, shown here by operational domain (SWE, CX, AIOps) and scope. Note: this is a different view from the Alignment Extension's org-scale heatmap, which measures coordination pressure across organizational scale (startup → enterprise). Both are valid lenses on the same underlying dynamic.
These notes explore the same questions from different angles.
The goal isn’t a single framework, but a set of lenses for thinking about how agents, systems, and organizations will evolve together.
Taken together, these pages form a loose map of the terrain.
Some explore agent capability, others explore system structure or alignment pressure.
Viewed together they offer different perspectives on the same emerging question: how increasing autonomy reshapes coordination inside real organizations.
Ready to move from the index to the analysis?
The Alignment Extension takes the coordination pressure layer further: Innovation Tax thresholds, the Fifth Ceremony intervention model, and what sustainable velocity actually requires.
→ Explore the Alignment Extension