Key Concepts
- 1Agents must reason about effective price, not list price — stacking coupons, tier discounts, loyalty points, and shipping thresholds change what the user actually pays
- 2Promotion logic is combinatorial and rule-heavy; the agent needs a reliable pricing engine to call, not an LLM guessing which discounts stack
- 3Threshold nudges are high-value: 'add $8 to hit free shipping' is a genuine service that also lifts AOV — the agent should surface these proactively
- 4Never let the LLM invent or approximate prices; pricing must be authoritative and real-time, because a wrong quoted price is a trust and potentially legal failure
- 5Promo eligibility depends on context (first-time buyer, region, cart contents), so the agent must pass full cart state to the pricing service to get a correct answer
- 6Transparency in savings framing builds trust: showing the real before/after and the reason for a discount beats vague 'special price' language
- 7Time sensitivity: expiring promos and price changes mid-session must be re-validated at checkout so the quoted price and the charged price always match
Why It Matters for PMs
AOV lever: Proactive threshold and bundle nudges convert directly into larger baskets — one of the clearest revenue impacts an agent can have when pricing reasoning is correct.
Trust and legal integrity: A quoted price that differs from the charged price destroys trust and can violate consumer-protection law. Authoritative, real-time pricing is non-negotiable.
Margin control: Reasoning about stacked discounts prevents accidental over-discounting, protecting merchant margin while still delivering the savings the customer expects.
Test Your Understanding
Question 1 (Multiple Choice)
Why should an agent call a pricing engine rather than let the LLM compute discounts?
Question 2 (Multiple Choice)
A user's cart is $42 and free shipping is at $50. What is the highest-value agent behavior?
Question 3 (Multiple Choice)
Why must a quoted promo price be re-validated at checkout?
Question 4 (Short Answer)
How would you design the agent-to-pricing-engine boundary so the LLM never quotes an incorrect price?
Think about your answer before revealing the explanation.
Question 5 (Short Answer)
What's the tradeoff between aggressive promo nudging (higher AOV) and coming across as pushy?
Think about your answer before revealing the explanation.