Day-to-day playbook (verified)
Customer Reply (No System Prompt)
Time to introduce Bayt Coffee — a small Cairo specialty roaster Hagar will be working with on a side project later in the course. They've asked for help with customer-support replies. We're going to ship a tight system prompt for them in the next lesson, but first we need to see what the model produces when there's no system prompt at all.
The setup
A real customer message arrives:
"Order #4821, paid Monday, still not shipped. I needed it for a client meeting tomorrow. This is the third time something like this happens."
Hagar throws this at Claude with no system prompt and just the simplest possible user message:
A customer is angry their order is 3 days late. Reply.
Captured output
# Response to Late Order
Dear [Customer Name],
I sincerely apologize that your order is 3 days late. I completely understand your frustration, and you have every right to be upset.
Let me look into this immediately and find out exactly what happened with your shipment. Could you please provide your order number so I can track it down right away?
Here's what I'll do:
- Investigate the delay and get you a specific delivery date
- Explore options to expedite shipping at no cost to you
- Apply a [discount/refund/credit] to your account for the inconvenience
Your time and trust are valuable, and we've fallen short of our commitment to you. I'm personally making this my priority to resolve.
I'll follow up within [X hours] with a concrete update, or please feel free to reach me directly at [contact info].
Again, I'm truly sorry for this experience.
Sincerely,
[Your Name]
[Title]
---
**Key elements**: Acknowledge their feelings, take responsibility, offer immediate action, provide compensation, and give a specific timeline for follow-up.
Captured from Claude Sonnet 4.5 (claude-sonnet-4-5) on 2026-04-27. Re-runs may differ slightly.
What the model picked when given freedom
Read the output and ask: who is "speaking"? The voice is corporate and generic — the kind of reply you'd get from a Fortune 500 helpdesk. Compare it to what Bayt Coffee actually wants:
| Bayt Coffee wants | The default reply gave |
|---|---|
| Warm, plainspoken, never corporate | "Sincerely apologize", "fallen short of our commitment", "valuable" |
| Acknowledge the specific issue first | "I sincerely apologize..." (apology-first, problem-second) |
| Offer one concrete next step with a date | Three vague action bullets, no date |
| Avoid "unfortunately", "kindly", "valued customer" | Uses "Dear [Customer Name]", "I'm truly sorry" |
| Sign off "— Bayt Coffee team" | "Sincerely, [Your Name] [Title]" |
The model didn't do anything wrong. With no system prompt and no constraints, it gave you the average professional support reply. The problem is "average" doesn't sound like Bayt Coffee.
Why this matters
Whenever you reuse a model across many similar tasks (customer support, internal docs, product copy), the default voice will be wrong. Not bad-wrong — just generic-wrong. Generic kills brand. Generic also burns context window because Hagar would have to add the same constraints to every single user message.
This is the moment a system prompt earns its keep. A system prompt is a one-time piece of instruction that defines who the model is — its persona, its rules, its banned phrases — and stays active across every message in the conversation.
Three rough rules for when you need one:
- You're shipping more than one message of the same type. Reuse → system prompt.
- The default voice is wrong for your audience. Tone correction → system prompt.
- You have rules the model must always honour. Always-on constraints → system prompt.
All three apply to Bayt Coffee.
What's coming next
In the very next lesson we run the same user message — "A customer is angry their order is 3 days late. Reply." — but with a tight system prompt. The result is dramatically different. That's the whole point of this two-lesson pair: see the same input produce two completely different outputs based purely on the system prompt.
Next: the same task with a tight Bayt Coffee system prompt — and the moment everything clicks. :::
Sign in to rate