The Cleric is the person who keeps complex systems reliable, repeatable, and governable.
They work in DevOps, SRE, platform engineering, FinOps, security, or any role that turns good intentions into operating practice. They make sure the build pipeline exists, documentation gets written, approvals have an owner, and the retrospective actually happens.
Without AI, the Cleric is a stabilizing force.
With AI, the ground is moving underneath the rules they spent years refining.
Those rules were designed for a world where people wrote the code, a human read the diff, and documentation was produced by someone who understood the system. AI changes what "done" means, what "reviewed" means, and what a test result proves.
The Cleric often notices this first. They are also often the last person given permission to question the old standards.
When rigor turns inward
The Cleric’s anti-patterns appear when operational rigor stops serving the work and starts serving itself.
The Faithful Automator extends trust from approved tooling to the output produced by that tooling. If the model says the code is correct, the review is approved. If the AI linter reports no issues, the change ships. If a test generator reports sufficient coverage, nobody writes another test.
The tool said so. What more is needed?
The Faithful Automator is not lazy. They may work extremely hard to build and maintain the automation pipeline. The problem is where that effort goes. They protect the workflow instead of questioning its conclusions. When the system produces an unreliable result, they respond with "but it passed all the checks" rather than investigating what the checks failed to see.
The Ritualist of Redundancy adds new AI ceremonies without retiring the old ones. The team now has sprint planning, AI-assisted sprint planning, a retrospective, an AI-generated retrospective summary, code review, and an AI review scanner.
Every ceremony arrived with a reasonable justification. None was removed. The calendar is full, and the value of each meeting is unclear.
The Doomsday Archivist preserves every output, log, prompt history, model version, and intermediate file forever. The instinct is understandable: keep the provenance. The result is a retention problem disguised as diligence. Storage costs grow, sensitive material accumulates, and nobody knows which artifacts matter.
Question the current proof
The failure is not gullibility. It is the difficulty of knowing which professional judgments still apply when the thing being judged has changed.
The counter-move is not to abandon process. It is to test each practice against the current reality.
Does the review process catch AI-generated bugs, or does it only confirm that the code has the expected shape?
Does the test suite verify behavior, or does it verify that the model wrote tests matching the implementation it just generated?
Does the documentation standard produce understanding, or does it produce paragraphs attached to every function?
Does retaining every prompt help someone maintain the system, or does it create a searchable pile nobody can interpret?
The answers may vary by project. A security boundary deserves a different review than a routine internal script. A generated test may be useful evidence, but it should not count as independent validation of the code that inspired it. A prompt may be worth keeping when it records an architectural decision and worthless when it records a one-line formatting request.
Process needs a purpose that can be stated plainly.
Retire what no longer works
Start by inventorying the checks that govern AI-assisted work. For each one, name the failure it is supposed to catch. Look at recent incidents, escaped bugs, and confusing changes. Did the process catch them? If not, add a check that addresses the actual gap instead of adding another status box.
Keep automated checks for what machines can judge well. Use human attention for assumptions, tradeoffs, security implications, and behavior under conditions the test does not cover.
Remove ceremonies that duplicate one another. Set retention limits for prompts and intermediate artifacts. Preserve the reasoning that future maintainers need, not every keystroke produced on the way there.
Process discipline is a strength. The moment a team repeats a ritual because it is familiar rather than because it works, the Cleric’s monster starts feeding.