From a Promising Task to a Coherent Use-Case Portfolio

OpenAI’s use-case guide becomes most useful when its three lenses—workplace friction, use-case primitives, and impact versus effort—are applied to the same piece of work. This checkpoint connects those lenses and adds the guide’s workflow-mapping idea. The goal is a practical judgment: can a team describe one bounded opportunity well enough to see where value is created, where risk travels, and what must be known before anyone chooses an architecture?

The answer requires resisting a common shortcut. Friction, primitive, impact, and effort are not four names for “a good AI idea.” They answer four different questions. Keeping them separate makes the eventual design more honest.

One candidate, four different judgments

Friction asks why the work is difficult. A team might spend too much time on repetitive transformations, wait for specialist help, or fail to get started because the work is ambiguous. This diagnosis describes the current condition. It does not yet say that AI is the best remedy. The same delay could arise from poor data, unclear ownership, an approval policy, or a missing conventional integration.

A primitive asks what kind of help the task needs. Content creation may turn context into a draft. Research may gather and organize information. Coding or data analysis may lower the skill barrier to a technical or analytical first pass. Ideation may create options or structure a problem. Automation may apply a stable routine repeatedly. A candidate can use more than one primitive, and the label remains descriptive: it says what the assistance does, not how important it is or how independently it should run.

Impact and effort ask whether the opportunity deserves investment now. The use-case guide’s matrix distinguishes high-impact, low-effort work from narrow self-service, high-value work that requires considerable investment, and high-effort work with little apparent value. “Impact” should refer to an outcome that matters to the organization or its users. “Effort” should include more than a demonstration: data preparation, permissions, integration, testing, adoption, support, and maintenance may all be part of the work.

Workflow mapping asks where the opportunity sits in a larger chain. A task can look valuable in isolation while producing an output that no later step trusts or uses. Conversely, a small task may unlock several downstream decisions. Mapping exposes the inputs, outputs, dependencies, reviewers, permissions, and failure costs that a quadrant cannot show.

flowchart LR A[Work problem] --> B[Friction diagnosis] B --> C[Primitive or combination] C --> D[Impact and effort judgment] D --> E[Workflow brief] E --> F[Agent fit question]

Consider a candidate such as “use AI to prepare a weekly customer update.” The friction might be repetitive collection and formatting, a skill bottleneck in interpreting data, ambiguity about which signals deserve attention, or all three.