Home – Cycle 1 – Cycle 2 – Cycle 3 – Cycle 4 – Cycle 5 – Cycle 6
Cycle 4 – Steer It – Experiment: Creating Alignment
How to Use This Page
- Run the guided experiment with the instructor.
- Start with a solid request and treat the first response as a first shot, not a final answer.
- Look at where the response landed before deciding what to ask next.
- React to what you actually received: question it, redirect it, narrow it, expand it, or ask it to improve something.
- Then choose one independent experiment and steer the conversation yourself.
Guided Experiment
Scenario: Improving Customer Service Coverage
Goal: Start with a good request, then use each response to decide where the conversation should go next.
Step 1 – Take the First Shot
Situation: Your company is receiving more complaints about slow customer service response times. Leadership is considering extending customer service hours, but does not want to simply add another shift.
What you know:
- Current customer service hours are 8 AM to 5 PM.
- 28% of customer inquiries arrive outside those hours.
- Most after-hours inquiries arrive between 5 PM and 7 PM.
- Complaints about response time have increased 20% over the last six months.
- Adding a full second shift would be too expensive.
- Employees are already concerned about workload.
- Leadership wants to improve responsiveness within 60 days without significantly increasing staffing costs.
Run this prompt:
You are advising a company that needs to improve customer service responsiveness. Customer service currently operates from 8 AM to 5 PM, but 28% of inquiries arrive outside those hours, with most arriving between 5 PM and 7 PM. Complaints about response time have increased 20% over the last six months. Adding a full second shift would be too expensive, and current employees are already concerned about workload. Analyze the situation and recommend three practical ways the company could improve customer service responsiveness within the next 60 days without significantly increasing staffing costs. Explain the advantages and disadvantages of each and recommend which option you would pursue first.
Step 2 – Where Did It Land?
Read the response before deciding what to ask next.
- What do you like?
- What concerns you?
- What seems too generic or unrealistic?
- What idea deserves more attention?
- Did the response make you think of something you had not considered?
Step 3 – Take the Next Shot
Choose one reaction from the discussion and tell the AI what you want it to do next. Do not restart unless you need to. React to the response you already have.
Then repeat: Read the new response, decide where you want to go next, and continue until the result is useful for the situation.
Step 4 – Decide When It Is Good Enough
Compare where the conversation ended with the first response. Did you know at the beginning that this was where you wanted to end up? What did you learn from the intermediate responses that affected your next move?
Compare / Reflect
- How did the first response influence what you asked next?
- Did you ever change direction because of something the AI surfaced?
- Were your follow-up prompts as detailed as the starting request?
- Could you have predicted the entire conversation before you started?
- When did you decide the result was good enough to use?
Independent Experiments
Choose one scenario below. Start with the provided request, review what the AI gives you, and then decide where to take the conversation. There is no predetermined sequence of follow-up prompts.
Scenario 1 – Overtime Keeps Increasing
Situation: Overtime costs have increased substantially even though production volume has barely changed.
What you know:
- Overtime hours are up 24% over the last six months.
- Production volume is up only 3%.
- Headcount is essentially unchanged.
- Absenteeism has increased from 4% to 7%.
- Supervisors report more schedule changes and production interruptions.
- On-time delivery has declined from 94% to 89%.
- Management believes the operation may simply need more employees.
- Leadership wants to understand the problem before approving additional hiring.
Starting Prompt:
You are advising a manufacturing company whose overtime hours have increased 24% even though production volume increased only 3%. Headcount is essentially unchanged, absenteeism increased from 4% to 7%, supervisors report more schedule changes and production interruptions, and on-time delivery declined from 94% to 89%. Management believes additional hiring may be necessary. Analyze the situation, identify the most likely contributors to the overtime increase, and recommend what management should do next before deciding whether to add employees.
Your job: Run it. Evaluate the response. Decide what you want to challenge, investigate, clarify, or pursue. Tell the AI and continue from there.
Scenario 2 – A Large Customer May Not Be a Good Customer
Situation: One of the company’s largest customers continues to grow, but operations is questioning whether the account is actually good business.
What you know:
- Annual sales to the customer increased 18%.
- The customer represents approximately 14% of company revenue.
- Their orders are smaller and more frequent than comparable customers.
- They regularly request expedited production.
- They require special packaging.
- They receive an 8% volume discount.
- Sales considers the relationship strategically important.
- Operations believes the account is creating significant disruption.
- Leadership does not want to damage an important customer relationship.
Starting Prompt:
You are advising leadership about one of the company's largest customers. The account represents about 14% of revenue and sales increased 18% this year, but its orders are smaller and more frequent than comparable customers, expedited production is common, special packaging is required, and the customer receives an 8% volume discount. Sales considers the account strategically important while operations believes it is increasingly disruptive. Evaluate the situation and recommend how leadership should approach the customer relationship without unnecessarily putting the account at risk.
Your job: Run it. Decide which part of the response deserves more attention. You might explore profitability, operations, pricing, negotiation, strategic value, missing information, or something else the response causes you to notice.
Scenario 3 – A Quality Problem Without a Confirmed Cause
Situation: Returns have increased significantly for one product family, but the cause has not been established.
What you know:
- Returns increased from 2% to 5% over six months.
- Most returns involve the same general defect.
- Production suspects variation in incoming material.
- The suspected supplier is the company’s lowest-cost supplier and has excellent delivery performance.
- Changing suppliers would increase material costs.
- A new supplier would require approximately 6-8 weeks to qualify.
- Adding additional incoming inspection would slow production.
- Nobody has established that the supplier is actually causing the defect.
- Leadership wants action because customer complaints are increasing.
Starting Prompt:
You are advising a manufacturer experiencing an increase in customer returns from 2% to 5% for one product family. Most returns involve the same type of defect, and production suspects incoming material variation. The current supplier is the lowest-cost option and has excellent delivery performance. Changing suppliers would increase costs and require 6-8 weeks of qualification, while additional incoming inspection could slow production. The supplier has not yet been confirmed as the cause. Recommend how the company should respond to the quality problem while balancing urgency, cost, production impact, and the uncertainty about the root cause.
Your job: Run it and evaluate not only the recommendation but the reasoning behind it. If the AI makes an assumption, jumps too quickly to a solution, or surfaces an interesting possibility, use that to determine your next move.
Scenario 4 – Use Your Own Situation
Task: Apply iteration to something you might actually ask an LLM to help you think through or produce at work.
- Choose a real work situation.
- Give the AI enough information to take a good first shot. Use RTIO if it helps you organize your request.
- Run the request and read the response before planning what comes next.
- Decide what you want to change, question, explore, or pursue.
- Tell the AI and review where the next response lands.
- Repeat until you decide the result is useful enough.
Not Sure Where to Go Next?
- What do you disagree with?
- What interests you?
- What’s missing?
- What assumption should you challenge?
- What deserves more attention?
- Did the response make you think of something new?
- Is it good enough yet?
Goal: You are not trying to predict the whole conversation. You are using each response to decide what to do next.
Developed in collaboration by:
CIRAS (Center for Industrial Research And Service, ISU) • IIAI (Iowa Initiative for Artificial Intelligence, UofI) • TrAC (Translational Artificial Intelligence Center, ISU)
![]() |
![]() |
![]() |
This course was developed through a collaborative effort to advance AI literacy across organizations in Iowa.


