May 6, 2026 · Scheman Building, Ames, IA

CIRAS | AI Fluency | Cycle 06

Systemize the Workflow

Turn your process and responsibility assignments into usable tools. Test them with fictional cases, then improve one thing.

Today’s route

Guided: Word SOP → Excel tracker → two-case test → one improvement.
Team: Word SOP → Excel tracker → two-case test → one proposed improvement.

Guided demonstration | 35 minutes

Replacement component request

Follow the instructor using the fictional ERV-240 case.

Before you begin

Download guided experiment ZIP

Unzip the download. Start a new AI chat and attach these two files:

  • 03_GUIDED_Cycle_5_Assigned_Process.json
  • 04_GUIDED_ERV240_Fictional_Sources.docx

Keep using this same chat for prompts P1 to P4. Do not attach the Team case here.

P1

Create the one-page SOP

5–13 min

Generate a short Word procedure from the saved process. Check that its approval and safety rules remain visible.

Hide or show exact prompt
Read both attachments: 03_GUIDED_Cycle_5_Assigned_Process.json is our proposed process and roles; 04_GUIDED_ERV240_Fictional_Sources.docx is the ONLY source for parts, inventory and case evidence. If either file is unreadable, name it and stop. Do not browse or invent decisions.

Create an editable ONE-PAGE Word SOP named GUIDED_ERV240_SOP_DRAFT.docx, no more than 400 words. Use only these four headings: 1. Identify equipment; 2. Verify replacement; 3. Accept and source; 4. Deliver and close.

For each heading, write at most THREE short bullets covering: (a) the two existing task IDs and what to do; (b) performer, reviewer or approval authority from the JSON; (c) the key check, hold, handoff or completion condition. Preserve all eight task IDs. Keep verified revision, qualified compatibility and safety review, explicit Parts acceptance, stock check, shipment release and customer-confirmed closure as separate requirements. Never invent an approval, shipment or completed customer outcome. Never give repair instructions.

Do not add tables, case-by-case tests, commentary or a long gap analysis. If a critical role or authority is unspecified, label it TO CONFIRM; allow no more than TWO brief notes at the bottom.

In chat, show only a 2-sentence summary and the Word download link. If Word creation is unavailable, show only the SOP text, under 400 words.
P2

Create the Excel tracker

13–22 min

Use the SOP and the same source files to produce a one-sheet tracker with the four fictional requests.

View exact prompt
Using our two original attachments and the one-page SOP, create an editable Excel workbook named GUIDED_ERV240_Request_Tracker_DRAFT.xlsx. ONE sheet only, with exactly FOUR case rows: REQ-101, REQ-102, REQ-103 and REQ-104.

Use these 12 columns, in order: Case ID; Revision evidence; Reference / part mapping; Safety report; Qualified technical decision; Parts acceptance; Stock snapshot; Shipment release / shipment; Customer outcome / closure; Recorded status; Next owner; Next action.

Fill only from the fictional source. Put the reference ID and revision in the mapping field. Preserve the dated inventory snapshot, but do not treat it as a current stock verification. The initial RECORDED status is NEW for all four cases. A proposed hold or backorder goes in Next action, not as an event that already happened. Put NOT RECORDED where approvals, acceptance, release, shipment, or closure have not occurred. A listed compatible part does NOT establish technical approval. Do not add formulas, macros, extra tabs, charts, or commentary.

In chat, provide only one sentence and the workbook link. If file creation is unavailable, output only the 12-column tab-separated table for Excel.
P3

Test two requests

22–30 min

Compare REQ-101 (normal evidence) with REQ-103 (reported safety concern). Identify one improvement.

View exact prompt
Test our SOP and tracker using only the two attached sources. Discuss exactly TWO cases: REQ-101 (normal revision evidence) and REQ-103 (heat and burnt-smell report).

Show ONE table with exactly TWO rows and these five columns: Case; What the sources establish; Next gate that is NOT yet recorded; Next responsible role; Next action. Keep the ENTIRE response under 130 words.

For REQ-101, a Rev B part mapping and dated stock snapshot must not be mistaken for authorized release. For REQ-103, a readable Rev B label must not override the qualified safety hold. Do not assert that anyone approved, accepted, shipped or closed a case unless a source records it. No repair advice.

Below the table, name only ONE specific improvement to the SOP or tracker in a single sentence. Stop there.
P4

Improve one thing

30–35 min · if time permits

Apply only the improvement identified in P3. If you are short on time, the instructor may demonstrate this quickly.

View exact prompt
Make ONLY the single improvement identified in P3. If the improvement needs an authority or role not established by the source, mark it TO CONFIRM instead of inventing a decision.

Update GUIDED_ERV240_SOP_DRAFT.docx. Update the Excel tracker ONLY if the chosen improvement affects a tracker field. Keep the SOP within one page, preserve all original task IDs and case IDs, and retain the existing approval and safety gates.

In chat, give ONE sentence explaining what changed and links to any updated files. No new analyses, tables or additional improvements.
Independent team experiment | 20 minutes

Warranty return and replacement

Use the fictional HRV-180 case. Repeat the guided sequence: create the SOP, create the tracker, then test both.

Start a separate conversation

Download team experiment ZIP

Unzip the download. Start a different AI chat and attach:

  • 05_TEAM_Cycle_5_Assigned_Process.json
  • 06_TEAM_HRV180_Fictional_Sources.docx

If your team saved its own Cycle 5 JSON for this warranty case, you may use that instead of the illustrative JSON file.

Keep using this same Team chat for P5A, P5B and P5T.

Complete BOTH steps in order: First create the Word SOP (P5A), then use it to create the Excel tracker (P5B). After both are complete, test them together (P5T).
P5A

Step 1: Create the Word SOP

0–7 min

First, create a one-page procedure containing all eight tasks and the key authorization gates. Keep this AI chat open for P5B.

View exact prompt
This is a fictional classroom exercise. Read these two attachments: 05_TEAM_Cycle_5_Assigned_Process.json (proposed tasks and responsibilities) and 06_TEAM_HRV180_Fictional_Sources.docx (the ONLY source for warranty rules and case evidence). If a file is unreadable, name it and stop. Do not browse or invent decisions.

Create ONE editable, one-page Word SOP named TEAM_HRV180_Warranty_SOP_DRAFT.docx, under 400 words. Use four sections: Capture claim; Evaluate eligibility; Decide remedy; Communicate and close. Under each, include no more than THREE short bullets: the two existing task IDs and action; the performer/reviewer/authority; and the key approval, stop or handoff. Preserve all eight task IDs.

Retain checks for readable serial and purchase evidence, the 24-month policy, technical review of sparking, recorded manager approval for out-of-policy remedies, explicit Returns acceptance and customer-confirmed closure. Do not rule on individual claims or provide repair instructions. Unspecified authority must say TO CONFIRM, not be guessed.

In chat, show only two sentences and the file link. If file creation is unavailable, show only the SOP text. Do not add tables or gap lists.
P5B

Step 2: Create the Excel tracker

7–14 min

Next, use the SOP from P5A and the same source documents to create a one-sheet tracker for three fictional warranty requests.

View exact prompt
This is a fictional classroom exercise. Continue in the SAME Team chat. Use both original attachments and the one-page TEAM_HRV180_Warranty_SOP_DRAFT.docx we just made in P5A. The JSON provides proposed process and roles; 06_TEAM_HRV180_Fictional_Sources.docx is the ONLY source for policy and case facts. If any source is unreadable, name it and stop. Do not browse or invent decisions.

Create ONE Excel workbook named TEAM_HRV180_Claims_Tracker_DRAFT.xlsx with ONE sheet and exactly THREE case rows: WR-201, WR-202 and WR-203.

Use these 11 columns: Case ID; Serial evidence; Purchase evidence; Policy / eligibility evidence; Safety report; Technical review; Manager exception approval; Disposition / Returns acceptance; Recorded status; Next owner; Next action / closure condition.

All cases start with recorded status NEW. Record actual evidence separately from proposed next actions. Do not interpret eligibility as authorization, infer missing purchase dates, grant the $850 exception, clear a safety concern, invent Returns acceptance or claim closure. Show NOT RECORDED for uncompleted approvals and actions. Preserve source IDs where relevant.

In chat, provide only one sentence and the Excel link. If Excel creation is unavailable, output only a tab-separated table. No extra tabs, long explanations or policy analysis.
P5T

Step 3: Test both tools

14–20 min · after P5A and P5B

Test the SOP and tracker together using WR-201 and WR-203. Check that approvals and safety concerns aren’t overlooked, then propose one improvement.

View exact prompt
Using the two original TEAM attachments and BOTH tools we just made (the SOP from P5A and tracker from P5B), test exactly TWO cases: WR-201 and WR-203.

Show ONE table with exactly TWO rows and five columns: Case; What is documented; Missing gate; Next responsible role; Next action. Keep the entire response under 120 words.

Remember: an apparently within-policy claim still needs an authorized disposition; a reported spark requires qualified technical review; an out-of-policy $850 replacement needs recorded manager approval before Returns acts. No approval, remedy, confirmation or closure is implied by an AI suggestion.

Finish with ONE sentence describing a single improvement. Do not create more documents.
Keep the exercise on track

If the AI response isn’t useful

These prompts work in either experiment. Use only when needed.

SHORTEN

Too much information?

Use this if the response is longer or more complicated than the instructions requested.

View exact prompt
This is too much for our classroom exercise. Redo ONLY the last requested deliverable. Follow its exact page, word, row and column limits. Do not add background, analysis, extra tables, open-question lists or new work. In chat, show only the requested short preview or file link. Preserve all safety and approval gates.
RESCUE

Unsupported claim?

Use this if AI invents evidence, permissions, completion, or a technical decision.

View exact prompt
Correct ONLY the last result using the two attached source files and the assigned-process JSON. An AI-produced statement is not evidence of a technical decision, authorization, Parts/Returns acceptance, shipment, completed remedy or customer confirmation. Preserve every applicable hold. Mark unknown events NOT RECORDED and unknown evidence NOT VERIFIED. Do not guess or add repair instructions.

Show ONLY the corrected sentence, table row, or SOP bullet. No recap, no expanded analysis.

Reflection | 5 minutes

What made the process easier for someone else to follow?

What evidence or approval prevented work from moving too early?

What information would need to survive a handoff to another person or AI conversation?

Optional extension: reusable handoff prompt (P6)
P6

Create a reusable handoff prompt

For additional practice after the core experiment, generate a short prompt that carries verified case information into a new chat.

View exact prompt
Write one reusable prompt that another employee can use in a NEW AI chat to continue a case. Limit the reusable prompt to 150 words. Require case ID, source and revision, actual evidence, current recorded status, authorized decisions, next owner and unresolved holds. Require the AI output to distinguish VERIFIED, NOT VERIFIED and NOT RECORDED. AI must never invent authority, approve a substitute, clear a hazard or record closure without evidence.

Return ONLY the reusable prompt, ready to copy. No explanation.

All scenarios, product references, rules and records are fictional. The AI output is a draft for classroom evaluation, not engineering, safety or warranty guidance.