Hands-On Workshop · Leveraging AI for Workforce Efficiency

Workshop 1:
The Customer Complaint

A frustrated long-time customer. A shipment with damaged units. Draft a response that protects the relationship — without admitting fault you don't owe.

⏱ 10 MIN Customer Service / Operations All Industries

01 / The Scenario

You are the Customer Service Manager at a manufacturing company in eastern North Carolina.

This morning, you opened your inbox to a sharp email from Carolina Equipment Supply, a customer of 10+ years. Their latest shipment arrived with 3 damaged units out of 50. They're asking for a full replacement shipment and a 15% discount on their next order.

You pulled the records. Your packaging passed inspection. Photos from the customer show crush damage consistent with freight handling — not a plant defect. The freight carrier is responsible, not you. But this customer has been with you for a decade and does roughly $180K a year in business.

Your RoleCustomer Service Manager
CustomerCarolina Equipment Supply
Customer ContactDenise Whitaker, Purchasing Director
OrderPO #CES-2041, 50 units
Issue3 units damaged on arrival — crush damage, not defect
Customer DemandFull 50-unit replacement + 15% discount on next order
Likely CauseFreight handling (carrier: Reliance Logistics)
Relationship History10+ years, $180K annual spend, first damage claim
Your GoalReplace the 3 damaged units, file a freight claim, preserve the relationship — without admitting plant fault or eating a 15% discount

02 / Step-by-Step

1

Open your AI tool

⏱ 1 minute

Open whichever AI tool you prefer and start a new conversation:

ChatGPT Claude Gemini Copilot

If your company has deployed an enterprise AI, sign in with your work account for better data protection.

2

Try the "lazy prompt" first

⏱ 1 minute

Type this exact prompt and see what you get:

Lazy Write an email to a customer who complained about damaged products.
Notice The output is generic. It could be for any customer, any product, any situation. It probably apologizes for things you didn't do wrong. This is why most people's first AI experience is underwhelming — the prompt is the problem, not the tool.
3

Try the structured prompt

⏱ 2 minutes

Now paste this into the same conversation:

Structured You are an experienced Customer Service Manager at a manufacturing company in eastern North Carolina. Draft a response email to a frustrated long-time customer who has asked for a full replacement shipment AND a 15% discount on their next order after receiving damaged product. Context: - Customer: Denise Whitaker, Purchasing Director, Carolina Equipment Supply - PO #CES-2041, 50 units, 3 arrived damaged - Photos show crush damage consistent with freight handling, not a plant defect - Our packaging passed inspection before leaving the dock - Freight carrier is Reliance Logistics — they're responsible, not us - Customer is 10+ years with us, roughly $180K annual spend, first damage claim ever Tone: Professional, warm, firm. Do NOT admit plant fault. Do NOT agree to the 15% discount. Strategy: Offer to replace the 3 damaged units immediately as a goodwill gesture, and explain that we'll file the freight claim on their behalf. Protect the relationship. Constraints: Under 200 words. Request a quick call to walk them through the photos and the freight claim process. Sign it from: Jordan Price, Customer Service Manager.
Notice the Difference Same AI. Completely different output. The AI didn't get smarter — the prompt got clearer. You gave it a role, a situation, a strategy, and a line you won't cross. That's the job.
4

Iterate — talk back to the AI

⏱ 2 minutes

Revise the draft by typing this follow-up in the same conversation:

Revise Good, but the opening is too cold. This is a 10-year customer — start by acknowledging the relationship and thanking them for flagging it quickly. Also soften the freight claim explanation so it doesn't read like we're pointing fingers. Keep everything else.
Notice You didn't start over. You talked back to the AI the way you'd talk back to a sales rep who drafted something for you. Revisions are free.
5

Try a firmer tone

⏱ 2 minutes

Now flex the other direction. In the same conversation, type:

Firmer Now give me a firmer version. Assume the customer has pushed back after the first email and is still insisting on the 15% discount. Hold the line — no discount — but stay professional and keep the door open. Make it clear the damage happened in freight, and we've already started the claim on their behalf.
Notice Same core facts. Different tone. You now have three versions — a warm first reply, a firm follow-up, and whatever the AI gave you first. Pick the one that fits how the conversation actually goes.
6

Spin off the internal version

⏱ 1 minute

One more use of the same thread. Try this:

Internal Now draft a short internal note to my plant manager and our shipping lead summarizing the situation, what I'm offering the customer, and what I need them to do (pull QC records for PO #CES-2041 and prep 3 replacement units for expedited shipping). Keep it under 100 words.
Notice Same source of truth, two deliverables, one conversation. External reply and internal action memo — both drafted and ready in under ten minutes.
7

Reflect with your table

⏱ 2 minutes

Discuss:

  1. Which version would you actually send to the customer?
  2. What would you change before hitting send?
  3. What's one recurring external email at your company you could run through this same sequence next week?

03 / What You Just Learned

The first prompt is rarely the best one. Iteration is where quality lives.

Giving AI a role changes the output completely. "You are an experienced Customer Service Manager" outperforms "write an email" every time.

Context is king. The relationship history, the freight evidence, the annual spend — all of that made the draft land.

Tone is a prompt, not a rewrite. Warmer, firmer, more diplomatic — ask, and the AI adjusts.

Talk back to the AI. Don't start over. Revise it the way you'd revise a coworker's first draft.

A Note on Real-World Use

When you do this with a real customer situation at your own company: use your company's enterprise AI tool if one is deployed, never paste customer names, PO numbers, pricing, or contract details into a free public tool, and always read the draft carefully before sending. AI gets things close, not perfect. You are still the author. AI is the assistant.