Workflow Automation Quiz
Describe up to 5 workflows and see which ones are the best candidates for AI automation. Each workflow is scored on frequency, duration, structure, output type, and team involvement.
Not every workflow should be automated
The most underrated skill in AI deployment is knowing what not to automate. Some workflows look like obvious candidates — they’re repetitive, take a lot of time, and happen often — but they rely on tacit human judgment that’s hard to specify in instructions. Others look unpromising but turn out to be ideal because they’re structured, rule-governed, and easy to review.
The four patterns that usually get incorrectly flagged for automation: workflows that require physical presence, workflows where errors cost more than the human review would, workflows that happen rarely (less than once a week), and workflows where the real bottleneck is a decision, not a task. Each of these looks automatable on paper and produces AI initiatives that either stall or generate work rather than remove it.
The five dimensions this quiz scores — frequency, duration, structure, output type, and team involvement — are designed together to separate real automation candidates from look-alikes. A workflow that runs often but has unstructured output will score lower than a workflow that runs less often with highly structured output, and that’s usually the correct call.
How to read your scores
15-20: strong automation candidate. The workflow is structured enough that a Claude AI project with proper instructions will produce reliable output, and it runs often enough to justify the engineering work. Examples from past engagements: proposal generation, compliance checklists, status report drafts, onboarding email sequences, weekly performance summaries.
10-14: deployable, but needs more design up-front. These workflows usually have one or two dimensions that pull down the score — output is semi-structured, or team involvement is high. They still work, but they require more instruction engineering and a clearer human-in-the-loop pattern to avoid surfacing garbage outputs. Expect a slightly longer build phase.
Below 10:skip this workflow for now. Trying to automate it produces more friction than value, and the time spent would be better invested in a higher-scoring candidate. Revisit it once three or four other projects are running cleanly — by then, you’ll have internal pattern library that may unlock it.