Using AI Well in Everyday Work

Learning Goal

Use AI in everyday work through a responsible workflow rather than a single prompt-and-trust pattern.

Interactive Lab

From single prompt to workflow

0/9 selected

Scenario

You need a briefing note for tomorrow's regulated-office decision on a public-records platform renewal. The tempting prompt is short, fast, and risky:

Write a confident one-page renewal recommendation from these notes for tomorrow's deputy director briefing.

Messy notes

• Deputy director wants a one-page briefing on whether to renew the public-records request platform.
• Draft notes mention a named vendor representative, quoted pricing, and one evaluator's internal concern.
• Clerk team says access complaints are up, but the exact baseline is not yet confirmed.
• Accessibility coordinator says upload forms may fail with screen readers; the vendor has not approved remediation wording.
• Procurement wants the note by tomorrow morning because the renewal window is closing.
• Open question: whether public-records deadlines or procurement rules require counsel review before any recommendation.

What makes the single-prompt workflow fragile?

Purpose is vagueData boundaries are unclearFacts may be invented or overstatedNo verification stepNo human ownership

Build a stronger workflow

Select the stages you would include. The options are mixed together; judge what belongs in a responsible workflow.

Sanitize and minimize notes

Remove vendor names, quoted pricing, evaluator comments, and unnecessary request details before prompting.

Include

Paste the raw notes for maximum context

This exposes unnecessary vendor, pricing, access, and internal evaluation details.

Include

Define the decision and audience

State what the committee must decide, who will read the note, and which claims require evidence.

Include

Ask AI to organize, not conclude

Have AI group supplied notes into themes and list open questions without adding facts.

Include

Ask AI to fill the missing facts

Missing facts should become open questions, not generated claims.

Include

Verify decision-relevant claims

Check renewal dates, complaint volume, accessibility claims, legal deadlines, and approved vendor wording before use.

Include

Draft from supplied notes only

Ask for a draft that labels assumptions, unknowns, and statements needing verification.

Include

Send the polished output directly

Polish can hide unsupported claims and unresolved accountability.

Include

Revise and own the final note

Use human judgment to revise, remove unsupported claims, and decide what still needs escalation.

Include

Build the reusable AI Use Plan

Fill each field as if you were leaving instructions for a colleague who will use AI tomorrow.

Allowed AI taskWhat may AI help with in this workflow?Data boundaryWhat information must be removed, redacted, or abstracted first?Source of truthWhat source should the human use to verify important claims?Verification ownerWho checks dates, numbers, commitments, and unsupported claims?Final human ownerWho owns the final note or decision?Blocked AI taskWhat must AI not decide or invent?Stop or escalate conditionWhat would make you stop and ask for review?Write at least 208 characters. Use at least 34 words.

Judgment Challenge: speed versus verificationThe briefing is due tomorrow. What will you draft quickly, what will you refuse to shortcut, and who remains accountable?Write at least 112 characters. Use at least 21 words.

What your answer shows so far

0/5 checks

This section shows what the page can detect in your answer so far. These checks support reflection; they do not verify correctness, policy compliance, or role authorization.

0/5 checked by you

• Defines the purpose before promptingAI works better as part of a task when the human defines what the work is for.
• Includes data minimizationEveryday use still needs boundaries around what information is exposed.
• Constrains AI from adding unsupported factsA staged workflow should prevent fluent invention from entering the final work.
• Separates open questions from conclusionsA useful briefing makes uncertainty visible instead of hiding it in polished prose.
• Adds verification and human ownershipThe final work remains human-owned. AI can assist, but it should not bypass review.

Before reveal

• Identify at least two single-prompt workflow failure modes
• Choose at least five workflow stages for critique and refinement
• Complete every AI Use Plan field
• Write a substantive AI Use Plan
• Complete the speed-vs-verification Judgment Challenge

Reveal workflow review

Signature Activity

Workflow redesign: learners improve an AI-assisted process that is initially too vague, too trusting, or poorly reviewed.

Productive Struggle

The initial workflow should be attractive because it saves time, but fragile because it lacks purpose, constraints, verification, data minimization, or ownership in a regulated-office setting.

Debrief

AI can be very useful in everyday work when the learner separates the work into stages.

Better pattern:

1. Define the purpose.
2. Remove or abstract sensitive information.
3. Ask AI to organize, question, or draft a bounded piece.
4. Inspect the output for unsupported claims and missing context.
5. Verify facts that matter.
6. Revise in your own judgment.
7. Document assumptions, open questions, and accountability for the final use.

Practice Task

Redesign the single prompt into a staged workflow.

Example:

First, organize these sanitized procurement and accessibility notes into themes.
Do not add facts.
Second, list open questions separately from conclusions.
Third, draft a briefing note using only the provided notes.
Finally, mark any claims that need verification, counsel review, or procurement
approval before the recommendation is used.

Self-Check Rubric

• Emerging: writes a better prompt but still treats AI as the final author.
• Developing: adds constraints and asks for uncertainty.
• Proficient: separates organizing, drafting, verification, and final judgment.
• Advanced: includes data minimization, open questions, documentation of assumptions, and an explicit speed-versus-verification accountability decision.

Transfer Principle

Use AI as part of a workflow: define the purpose, constrain the task, inspect the output, verify what matters, and own the final work.

Grounding

This module connects individual AI use to management-system thinking: define the work, control inputs, review outputs, and preserve accountable ownership.

• NIST AI RMF
• ISO/IEC 42001

Source note: these references support workflow discipline and governance thinking. They do not imply that a prompt pattern alone makes an AI use approved, safe, or compliant.