Free 10-minute assessment

Which workflow should your team bring to AI training?

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.

Short answer: start with a recurring process that has recognizable source material, a clear output, an accountable owner, and a human-review point before the result affects customers, employees, money, or policy.
Score a workflow Map the workflow

Choose the work before the tool

A useful first workflow has three qualities.

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.

Mistakes are reviewable

A knowledgeable person can inspect the result before an error becomes a customer, financial, legal, security, or operational problem.

Success is observable

The team can compare time, completeness, rework, consistency, or decision quality before and after a small experiment.

Interactive checklist

Score one real workflow.

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.

0/10
Not the first workflow to choose

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

The spreadsheet called FINAL_v7_REAL_FINAL.xlsx

Suppose 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."

  1. Frame the job. Turn an incoming request into a complete, evidence-backed triage brief.
  2. Name the judgment. Urgency, exceptions, and incomplete information still require an accountable person.
  3. Set the review gate. A supervisor checks the evidence and final priority before work is assigned.
  4. Test one small change. Compare completeness and review time on sanitized examples before changing the live process.

Know when to wait

Three reasons not to start with a workflow.

Nobody agrees what "good" means

If experts cannot agree on the expected result, adding AI will amplify the ambiguity. Define the decision and acceptance criteria first.

The first error would be expensive

Do not make a high-stakes legal, financial, clinical, personnel, security, or policy decision the team's first practice exercise.

The process has no owner

A cross-functional workflow without an accountable reviewer becomes a tool experiment with nowhere to land.

Use the result

What to do with your score.

8–10: bring it to the workshop

Collect three representative examples, name the owner and reviewer, and write down the current time or rework burden.

5–7: spend 30 minutes framing it

Clarify the missing inputs, output, review point, test data, or success measure before choosing any technology.

0–4: choose a smaller slice

Find one repeated handoff, document, classification, summary, or quality check inside the larger process.

Questions

AI workflow-readiness FAQ

What makes a workflow a good first candidate for AI?

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.

Should we begin with our most important workflow?

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.

Do we need to select ChatGPT, Claude, Copilot, or another tool first?

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.

Can we use confidential company material in the assessment?

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

Bring one recognizable workflow. Leave with one useful next step.

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