Assess one workflow, not your worth

Choose a task you already understand: drafting a weekly sales summary, organizing customer comments, or preparing a first version of a campaign brief. Avoid starting with a vague ambition such as “automate marketing.”

Write one sentence: “When this input arrives, produce this draft for this person to review.” If that sentence is difficult to complete, document the workflow first. You do not need to become a programmer to make this decision.

Five checks that change the decision

Answer honestly. “I do not know yet” is useful information. The assessment should reveal work to do, not reward optimistic answers.

  • Workflow: can you name a repeatable input and the output you need?
  • Permission: are you allowed to use this data with the chosen tool?
  • Reviewer: who can recognize an incorrect or incomplete result?
  • Success: what evidence would show that the output is useful?
  • Consequence: can you keep the first attempt separate from customer messages, payments, and other external actions?

What your result means

The recommendation follows the limiting condition, not an average of your answers. A clear task does not cancel out missing data permission. A confident user does not replace a qualified reviewer.

  1. Unclear task: write the input, expected output, and owner before choosing software.
  2. Unknown permission: practice with fictional examples while you confirm approved tools and data use.
  3. No reviewer or success check: define how a person will inspect the output before using it.
  4. Consequential decisions or automatic actions: narrow the scope to a draft, simulation, or internal comparison and get appropriate oversight.
  5. Clear task, approved data, review owner, and checks: run a bounded pilot with human approval before any external action.

Example: a weekly sales summary

A sales manager wants a draft summary from an approved export. The pilot uses a copy of last week’s data, keeps deal IDs, and asks for wins, losses, missing owners, and questions. It does not email clients or change CRM records.

Success means every quoted total reconciles to the source, every highlighted deal exists, and unknown fields stay unknown. Record both drafting time and correction time. A fast answer with incorrect totals is not a successful report. If permission for real records is unclear, use a fictional spreadsheet first.

Run a small experiment you can inspect

Save the original input. Write the task brief before prompting. Keep the output and your corrections together so someone else can understand the result. Then repeat with a second example, including one awkward case such as missing dates or contradictory notes.

Continue only if the reviewer can explain what worked and what still needs checking. A successful sample is evidence about that task under those conditions; it is not permission to automate every similar task.

Why there is no percentage score

A single percentage hides important differences between a low-consequence draft and an action that affects another person. This is an editorial planning aid with visible rules, not a validated organizational assessment.

For a broader risk-management program, NIST provides a voluntary AI Risk Management Framework. Its companion playbook explicitly is not a checklist to follow in its entirety. This page is an independent practical worksheet, not a NIST certification or endorsement.

Before you get started

Do I need technical skills to begin?

No. Understanding the task and recognizing a useful result matters more for this first exercise. Start with a draft you can check rather than a workflow you cannot explain.

Does a positive result mean our business is ready for autonomous AI?

No. The recommendation applies to the workflow and conditions you described. It does not approve autonomous access, spending, messaging, or consequential decisions.

Should I enter confidential company information?

You can complete the assessment using a generic task description. Do not enter customer records, passwords, or confidential details. Confirm permission separately before using real work data in an AI tool.

What if my team is completely new to AI?

Use fictional inputs and compare the result with a simple answer you can verify. Learning how to specify and check a task is a useful first outcome.

Sources and how this guide was made

Product guidance is grounded in the sources below. The tools and fictional teaching materials were created for this guide. We do not present these examples as independent product benchmarks or guaranteed outcomes.

Product names belong to their respective owners. Something Big Is Happening is an independent publication. Check current plans, permissions, and availability in the official documentation.