📅 Week 1 · Monday
day-01
What is WA.Expert?.
Today you'll learn: what WA.Expert is and why businesses pay for it — explained so clearly you could teach it to your parents by tonight.
⏱ ~20 mins
📖 Read + Quiz + Submit
✅ Need 3/5 to unlock
🔒 Monday only
Week
Week 2 of 8
Day
1 of 14
Program
2-Week Program
📖 Read This First — About 8 Minutes
AI in a WhatsApp chatbot is powerful. AI without a well-written system prompt is a liability.
WA.Expert supports AI integration — connecting your chatbot to models like GPT-4 to enable natural language understanding and open-ended conversation. This opens up genuinely powerful use cases: AI-powered FAQs, product recommendation engines, personalised support at scale.
But AI without proper prompt engineering in customer-facing contexts creates real risks: hallucinated information, off-topic responses, inconsistent brand voice — and in regulated industries, potentially harmful advice. Writing the system prompt is the most important technical skill in AI chatbot implementation.
The system prompt is the director's brief. Without it, the actor improvises — sometimes brilliantly, sometimes disastrously. With a precise brief: "You are a customer service agent for Apollo Pharmacy. You help with prescription queries. You never give medical advice." — the actor knows exactly what to do and what not to do.
Role definition
"You are a customer support agent for [Brand]. You assist with [specific topics]. You do not discuss [off-limits topics]." Precision prevents drift.
Hard boundaries
Explicit instructions on what the AI must never do: "Never provide medical diagnosis. Never quote competitor prices. When asked about X, always say Y."
Tone + response length
"Respond in a friendly, professional tone. Keep replies under 80 words. No technical jargon. End every reply with a question or clear next step."
Escalation instructions
"If a customer expresses frustration, immediately acknowledge and offer human connection. If asked about [specific topics], do not answer — escalate immediately."
⚠️ High-caution industries for AI: Healthcare (never diagnose), Finance (never advise on investments), Legal (never provide legal opinions). In these industries, hard boundaries are not optional — they are risk management.
Read the reference page below before taking the quiz.
Explore: AI Integrationwa.expert/pages/ai.html · AI integration features · ~5 mins
→
🧠 Quiz — 5 Questions
Day 1 Quiz
Score 3 or more to unlock your submission. Retry as many times as you want — every wrong answer tells you why.
5 questions
Need 3/5
Unlimited tries
Instant feedback
Question 1 of 5
What is a system prompt in an AI-powered WhatsApp chatbot?
✅ The system prompt is the invisible director's brief. Every response flows from this definition. Users never see it — but every word they receive is shaped by it.
❌ The system prompt is the invisible instruction set for the AI. Role, behaviour, limits, tone, escalation — all defined here. Invisible to users, critical to every response.
Question 2 of 5
Pharmacy AI chatbot has no boundary against medical advice. User asks: 'Is it safe to take ibuprofen and paracetamol together?' The risk?
✅ Medical advice from AI without hard boundaries = liability. Plausible but wrong in medical contexts can mean dangerous. Hard boundaries are non-negotiable in healthcare AI.
❌ Real medical risk + real liability. AI produces statistically plausible answers — not contextually appropriate ones. Hard boundaries are mandatory in healthcare.
Question 3 of 5
Which system prompt instruction best addresses tone and response length?
✅ Specific and measurable. 'Under 80 words' is testable. 'End with question or next step' is a concrete structural instruction. Vague instructions like 'be helpful' are unreliable.
❌ Option B is the answer. Specific, measurable, structural instructions produce consistent behaviour. Vague instructions do not.
Question 4 of 5
Real estate AI chatbot is asked: 'What is the current market value of properties in this area?' Correct system prompt handling?
✅ Hard boundary + redirect + escalation. Property valuations from AI training data are unreliable — stale, potentially wildly inaccurate, creates trust issues.
❌ Hard boundary + redirect is the answer. Property valuations from AI training data are unreliable. Block explicitly, redirect to human experts.
Question 5 of 5
After AI chatbot deployment, how should performance be monitored?
✅ Active monitoring is essential. AI drift, new query types, edge cases — all require ongoing prompt refinement. A chatbot deployed and forgotten degrades in quality over time.
❌ Active monitoring + monthly prompt updates is the answer. AI chatbots are not set-and-forget. Regular review and updates are minimum maintenance requirements.
–of 5
Answer all 5 questions, then check your score.
✏️ Your Task
Score 3/5 to unlock this
Complete the quiz above first. The moment you score 3 or more, this section unlocks.
🎉 Day 1 — done!
Day 2 opens on your assigned Tuesday.
📝 Today's Task
Someone in your family runs a small business. In 3–4 sentences, explain WA.Expert to them like you're actually WhatsApp-ing them right now. Your own words — not copied from the page.
Start like this: "So there's this platform I was reading about — it's basically for businesses that get too many WhatsApp messages to handle manually. It lets them..."
0 / 800
From your registration confirmation email. Can't find it?
Submitting before 11 PM IST on your assigned Monday counts as Day 7 complete.