Which Element of Intent Management™ Is Breaking Down?

When AI-enabled work consistently disappoints — when outputs miss the mark, reviews take forever, teams keep correcting the same problems, or people quietly stop using the tools — the instinct is often to blame the technology. A different model, a better prompt, a more capable system. Sometimes that's right. More often, the problem is in one of four specific places, and knowing which one changes everything about how you fix it.

Intent Management™ is built around four elements: Outcome Definition, Evaluation Criteria, Decision Authority, and Communication Cadence. When AI-enabled work breaks down, it's almost always traceable to one of these four things failing. Here's how to diagnose which one.

Outcome Definition

The first thing to check is whether the outcome was defined clearly enough in the first place. This goes beyond knowing what the deliverable is — it means the standard for what "good" looks like was explicit, shared, and specific enough to evaluate against.

Signs this is the problem: outputs are technically correct but miss the actual need, different reviewers keep applying different standards, or the team is spending significant time trying to figure out what the output was supposed to accomplish before they can evaluate it. The fix is doing the work of definition before the next project starts — not in the debrief after it fails.

Evaluation Criteria

Even when outcomes are well-defined, work can still go sideways if the criteria for evaluating outputs aren't explicit. Outcome Definition answers the question of what you're trying to produce. Evaluation Criteria answers the question of how you'll know when it's good enough — the specific, observable characteristics that distinguish acceptable from unacceptable output.

Signs this is the problem: reviewers are spending a lot of time on outputs that technically meet the brief, feedback cycles are long and contentious, or the same type of output gets approved in some contexts and rejected in others with no clear explanation. The fix is developing rubrics — not elaborate ones, but explicit enough that two reviewers looking at the same output would reach the same conclusion about whether it passes.

Decision Authority

This is the most frequently overlooked element. Decision Authority is about who has the right to say yes, who has the right to say no, and what happens when AI output is good enough by some standards but not others. In human workflows, this is usually handled by organizational hierarchy and intuition. In AI-enabled workflows, it breaks down because the volume of outputs requiring a decision increases significantly, and the people responsible for making those decisions often haven't been explicitly authorized to do so.

Signs this is the problem: outputs are stuck in review, people are escalating decisions that should be made locally, or there's inconsistency in what gets approved depending on who reviews it. The fix is explicitly mapping who approves what, at what quality threshold, and what the escalation path is when something doesn't clearly pass or fail.

Communication Cadence

The fourth element is about what happens over time. Intent drifts. Priorities shift. What was a reasonable standard six months ago may not fit the current context. Communication Cadence is the structured practice of checking whether the intent is still calibrated — whether the outcomes, criteria, and decision authorities defined at the start of a workflow still match what the organization actually needs now.

Signs this is the problem: the team is following the process correctly but producing outputs that no longer seem relevant, there's a growing disconnect between what the AI workflow produces and what the business actually uses, or feedback keeps surfacing the same issues that were supposedly resolved in earlier cycles. The fix is building regular calibration checkpoints into the workflow — not long reviews, but deliberate moments where the team asks whether what they're doing still matches what they're trying to accomplish.

Most AI workflow problems are not technology problems. They're intent problems — places where what the organization is asking for and what it actually needs have come apart.

Diagnosing Your Situation

In practice, these elements compound. A weak Outcome Definition makes Evaluation Criteria impossible to get right. Unclear Decision Authority undermines even well-defined criteria. And without Communication Cadence, any of the other three elements will eventually drift out of alignment with the actual need.

The diagnostic question to start with is: where in the workflow does the work first go wrong? If it's in the output itself, start with Outcome Definition. If the output seems reasonable but review keeps getting stuck, look at Evaluation Criteria and Decision Authority. If the work was running well for a while and then started degrading, Communication Cadence is usually the culprit.

Naming the right problem is most of the work. Once you know which element is breaking down, the fix is usually more obvious than the symptom makes it appear.