People recognize it
The team can describe what triggers the work, who touches it, where it gets stuck, and what a finished result looks like.
Free 10-minute assessment
A good first AI workflow is repeated, understandable, testable, and reviewable. Use this checklist to find work that is useful enough to matter without making your team's first experiment needlessly risky.
Choose the work before the tool
The team can describe what triggers the work, who touches it, where it gets stuck, and what a finished result looks like.
A knowledgeable person can inspect the result before an error becomes a customer, financial, legal, security, or operational problem.
The team can compare time, completeness, rework, consistency, or decision quality before and after a small experiment.
Interactive checklist
Check each statement that is already true. A low score is useful information—it means the workflow needs framing, not that the team has failed an AI test.
Choose a smaller, more repeated process with clearer inputs, visible errors, and an accountable reviewer. Learning will be faster and safer.
Quirky but realistic example
FINAL_v7_REAL_FINAL.xlsxSuppose a shared inbox receives customer requests. One person copies each request into a spreadsheet, interprets its urgency, looks up missing details, and prepares a daily priority list. That workflow is a better AI-training candidate than "automate customer service."
Know when to wait
If experts cannot agree on the expected result, adding AI will amplify the ambiguity. Define the decision and acceptance criteria first.
Do not make a high-stakes legal, financial, clinical, personnel, security, or policy decision the team's first practice exercise.
A cross-functional workflow without an accountable reviewer becomes a tool experiment with nowhere to land.
Use the result
Collect three representative examples, name the owner and reviewer, and write down the current time or rework burden.
Clarify the missing inputs, output, review point, test data, or success measure before choosing any technology.
Find one repeated handoff, document, classification, summary, or quality check inside the larger process.
Questions
A good first candidate is repeated, has recognizable inputs and outputs, has an accountable owner, can be tested with non-sensitive examples, and includes a clear human-review point before mistakes cause harm.
Usually not. Begin with a bounded, representative workflow whose errors are visible and correctable. The team can apply what it learns to higher-stakes work after its review habits and success measures are stronger.
No. First identify the job, source material, constraints, expected output, human-review points, and success criteria. Tool choice is easier once the team understands the work.
Use descriptions and sanitized examples for early framing. Any later use of confidential, regulated, or personal data requires the organization's approved tools, policies, access controls, and professional review.
Turn the score into a practical session
TrainedToThink workshops help leaders and teams frame the work, practice with representative examples, name the human-review points, and decide what to test next.
Map this workflow Build your workshop blueprint