E-commerce

AI chatbot for customer referral: explain rules and track rewards

AI chatbot for customer referral: explain rules and track rewards

August 14, 2026

"Where is my referral link?" "My referral has placed an order, when do I receive my €10?" "Why isn't my code working?" Three referral questions repeated in chat even though the program is configured in Smile or ReferralCandy.

ReferralCandy estimates that referral programs with transparent tracking on the customer side convert 3 to 5 times more referrals than those where the referrer can never see the status of their invitations (ReferralCandy, referral statistics 2026). Smile.io documents the five-step journey, from the unique link to the reward credit post-validation (Smile.io, referral program).

This guide #377 covers the AI chatbot for e-commerce customer referral: explaining rules, tracking rewards, REF-BOT intents, and API integration. Distinct from referral support (#376) (agent REF-INV investigation, fraud, manual credit): here, we focus on AI self-service automation for referrers and referrals.

Summary

Why automate referral programs via AI chatbot?

An e-commerce referral chatbot resolves sponsor links, program rules, referee status, and reward pending times without agent tickets, while handing off incidents to REF-INV (#376) if the credit is missing post-delay.

Five frustrations without a referral bot

  • Link not found : sponsor cannot see the dashboard code

  • Opaque reward pending time : "when do I get my €10?"

  • Checkout code rejected : referee blocked without explanation

  • Referee status invisible : sponsor does not know if their friend has ordered

  • Vague program rules : minimum cart, 1st order

Antavo lists lack of reward transparency among the major frustrations in loyalty programs, referral included (Antavo, 2026 Loyalty Report).

Angle #377

The referral support guide (#376) covers agent REF-INV, fraud, and manual crediting. The #375 LOY-BOT handles "my referral" handoffs. #377 deploys AI bot intents referral_link, referral_status, referral_rules.

DTC Example

Smile fashion brand referral, 64 referral tickets/month before REF-BOT. After link + status + delay bot: referral_bot_resolution 69%, referral_ticket_deflection -42%, referees converted via bot-shared link +18%.

Referral bot vs loyalty bot

LOY-BOT = general program points. REF-BOT = sponsor link, referee, specific referral reward. Separate intents, distinct RAG corpus.

Journey Touchpoint

Customer account (sponsor link), referee checkout (code failure), sponsor post-order (reward pending). Contextual bot per page.

REF-INV Handoff

Reward missing post-delay policy or fraud deny contest → #376 agent. Bot never applies credits.

Referral Bot ROI

Referral FAQ ticket = 10 min agent. Bot status + rules = 40 s. Out of 50 deflected tickets/month: 475 min saved.

Ambassador Activation

Referral_link bot with 1-click sharing increases sharing rate compared to the /referral page alone. An active sponsor = the cheapest DTC acquisition channel.

Frequent Error

Mixing LOY-BOT and REF-BOT intents: customer asking for "my referral points" receives general loyalty balance. Route with distinct NLP.

How does it differ from support #376 and the LOY-BOT?

Seven neighboring contents, seven roles on referral and AI automation.

Referral support (#376)

REF-INV (#376): investigation agents, fraud, manual credit. The #377 = upstream self-service bot, handoff if failed.

Loyalty bot (#375)

LOY-BOT (#375): points balance. Handoff referral intents → REF-BOT #377.

Pre-launch loyalty questions

Loyalty questions: referral go-live FAQ. The #377 = live post-launch bot.

Missing points (#374)

PTS-REC (#374): if referral reward = Smile points and missing post-credit.

AI Governance (#142)

Governance (#142): bot explains, does not credit referral reward.

Order fraud

Order fraud: high-risk referee. Bot cites pending, handoff #376 if denied.

Promise #377

REF-BOT framework, 10 intents, referral RAG corpus, referrer/referee flows, referral_bot_resolution KPI.

Which REF-BOT intents should be prioritized for mapping?

Map the referral bot intents before corpus and API integrations.

Ten REF-BOT intents

  1. referral_link: referrer link + sharing

  2. referral_rules: general program rules

  3. referral_reward_amount: reward amount for referrer and referee

  4. referral_status: status of invited referees pending/completed

  5. referral_delay: credit delay post-referee order

  6. referral_code_fail: invalid checkout code, explanation

  7. referral_first_order: referee first order eligibility

  8. referral_min_cart: minimum cart threshold

  9. referral_promo_stack: logic for stacking coupon code + referral

  10. referral_handoff_ref_inv: reward missing post-delay → #376

MVP Prioritization

Week 1: referral_link, referral_delay, referral_code_fail, referral_rules (75% of FAQ tickets). Week 2: referral_status, referral_reward_amount.

Routing referrer vs referee

Logged-in active referrer email → referral_link, referral_status. Guest referee checkout → referral_code_fail, referral_first_order.

Page context

  • /referral: link, rules, status

  • Checkout: code_fail, min_cart, promo_stack

  • Customer account: link, delay, referee status

90-day ticket mining

Cross-reference referral #376 tags with bot-eligible volume. Validate intent deployment order. Prioritize referral_delay if 40% of tickets = "when will I get my reward".

How do you structure the REF-BOT framework into six blocks?

The REF-BOT framework structures referral bot deployment into six technical blocks.

Six REF-BOT blocks

  1. RB-1 Auth: Shopify logged-in referrer email or referee guest verification

  2. RB-2 API referral: Smile/ReferralCandy referral link + status + pending rewards

  3. RB-3 Corpus RAG: program rules, deadlines, indexed exclusions

  4. RB-4 Intent router: NLP → 10 intents section 3

  5. RB-5 Response template: live API status + RAG rule

  6. RB-6 Handoff: referral_handoff_ref_inv → agent #376

Example response referral_link

"Here is your referral link: [url]. Share it: your referee receives €10, you receive €10 after their first eligible order (min €50)." Buttons Copy | WhatsApp | Email.

Example response referral_delay

"Your referee [masked email] ordered on August 8. Referrer reward credited within 14 days post-delivery, around August 22. Current status: pending."

Example response referral_code_fail

"Referral code valid only for 1st order and min cart €50. Your cart: €42. Add €8 or remove the incompatible promo code."

Fallback API down

RAG rules only + "status temporarily unavailable, contact us if urgent". Do not invent referee status.

Privacy referee

Referrer sees status "referee has ordered" without referee's full email if GDPR policy is active: "Referee ***@gmail.com".

Sandbox test

Before prod: test referrer A, test referee B, order €1 min staging override. Validate RB-2 pending status → completed bot flow.

Rate limit

Cache referral status 5 min. Bot indicates "status as of [time]" if cache is recent.

Which RAG corpus should be used to feed the referral rules?

The RAG referral corpus feeds stable supplementary rule answers to the live API.

Eight REF-RAG corpus blocks

  1. Rewards: referrer amount €, referee amount € or %

  2. Referee eligibility: 1st order, new account

  3. Min basket: € threshold before referral discount

  4. Referrer credit delay: typical D+14 post-referee delivery

  5. Cookie attribution: 30-day window click → purchase

  6. Exclusions: gift card, wholesale, marketplace

  7. Promo stacking: referral vs promo code policy

  8. Return clawback: referee return = referrer reward withdrawn

Marketing sync

Reward change 10→15 €: RAG update + test 5 bot questions same day. Link FAQ bot cleanup.

Anti-hallucination

Reward amount from RAG or API, never LLM guess. Referee status RB-2 API only.

Multilingual version

FR/EN corpus if multiple markets. Referrer link is locale-aware.

Which referral and referee conversational flows?

Referral conversational flows guide referrer and referee without an agent ticket.

Referrer flow: share link

"How do I refer?" → referral_link + short referral_rules + copy button. Proactive customer account widget.

Referrer flow: reward tracking

"Has my referee ordered?" → referral_status list of last 3 referees + referral_delay credit date. If completed: "€10 credited on August 15".

Referee flow: checkout code

Checkout error → bot referral_code_fail diagnoses: min cart, first order, promo conflict. Deep link to cart if simple fix.

Referee flow: pre-purchase

"My friend sent me a link" → referral_rules + referee referral_reward_amount + "click the link before adding to cart for cookie attribution".

Handoff flow reward missing

Referrer "didn't receive €10" + API status pending past delay → referral_handoff_ref_inv pre-filled ticket referee order, referral_id.

Quick replies

My link | Referee status | When is my reward | Code is not working.

Proactive post-share

Referrer email D+7 without referee conversion: "Your friend has not ordered yet. Reminder: [link]. Questions? Ask the bot."

Flow contest deny

Referrer "my reward was denied" → bot quotes general fraud policy + handoff #376 appeal with referral_id.

Checkout widget

Referral discount error at payment: bubble bot referral_code_fail context cart amount realtime.

How to integrate Smile, ReferralCandy, and Shopify?

The REF-BOT technical integration connects Smile, ReferralCandy, and Shopify customer data.

Smile.io referral API

GET customer referral link, GET referrals list status pending/completed/cancelled. Cache 5 min per customer.

ReferralCandy

Endpoints invite status, reward history. Adapt RB-2 depending on the app.

Shopify customer

orders_count = 0 for referral_first_order referred friend check. Tags ambassador if sponsor VIP program.

Checkout discount validation

Bot reads cart total via Storefront API if session: diagnostic min_cart realtime.

Webhook referral completed

Trigger sponsor email + optional proactive bot "Your €10 is available".

Widget /referral embed

Chat trigger "Questions about referral?" with referral page context.

Smile vs ReferralCandy mapping

Document equivalent RB-2 API fields per installed app. Switch app = reconfig intents, not rewrite flows.

Mobile app wrapper

Deep link referral_link opens native iOS/Android share sheet from bot button.

Which governance: follow without crediting the reward?

The referral bot governance defines limits: explain and track, do not credit or override fraud.

Bot can say

  • Referral link, rules, referee status, expected credit delay

  • Reason for code failure (min cart, first order, promo)

  • Referrer and referee policy reward amount

Bot cannot do

  • Manual reward credit (→ #376 agent)

  • Override fraud deny

  • Promise reward before policy delay

  • Generate referral code outside of API

Align governance (#142) and limits (#124).

GDPR

Partially mask referee email. Referrer only sees their own referrals.

Audit logs

Intent, API status returned, handoff yes/no. Monthly sample review.

Which referral_bot KPIs should be measured?

Referral bot KPIs measure ticket deflection and referral conversion.

Eight key metrics

  • referral_bot_resolution: resolved without agent / referral intent sessions

  • referral_ticket_deflection: decrease in tickets #376 post-bot

  • referral_bot_csat: post-referral conversation survey

  • referral_link_copy_from_bot: clicks copy bot link

  • filleul_convert_post_bot: referrals converted 7 days post bot referral_link

  • referral_handoff_rate: handoff #376 / referral sessions

  • referral_code_fail_bot_fix: completed checkout post referral_code_fail bot

  • rag_miss_referral: unanswered corpus questions

DTC Benchmark

Target referral_bot_resolution > 65%, referral_ticket_deflection -35%, referral_handoff_rate < 20%.

Weekly Dashboard

Volume intents, resolution, handoff, link copies. Share #marketing + #support.

Acquisition Correlation

referral_link_copy_from_bot vs new referrals/month: bot ROI beyond customer support.

Handoff Spike Alert

referral_handoff_rate +30% week: check change in app delay or API status bug before blaming bot scripts.

What edge cases and handoffs should be anticipated?

Six bot referral edge cases requiring handoff or specific rules.

Referrer with no referee yet

referral_status empty: bot referral_link + sharing tips, no handoff.

Referee existing account

orders_count > 0: referral_first_order fail. Explain policy, no bot exception.

Referral denied fraud

API status cancelled fraud: bot quotes general policy, handoff #376 appeal if contested.

Multi-click attribution

Bot quotes RAG first-click policy, does not decide attribution.

Reward store credit vs points

Bot explains reward format. Points missing post-credit → #374.

B2B wholesale referrer

DTC program only. Redirect pro policy.

Bot referral anti-patterns

Inventing referee status without API, promising reward before the delay, ignoring multiple referral_code_fail causes, obsolete RAG corpus reward amount post-campaign.

How does Qstomy deploy REF-BOT on Shopify?

Qstomy deploys REF-BOT: live referrer link, API referee status and REF-INV handoff if reward missing.

Referral bot capabilities

  • referral_link_live: Smile API referrer URL + share buttons

  • referral_status_api: pending/completed referee list

  • referral_rag_rules: 8 indexed REF-RAG blocks

  • referral_code_diagnose: min cart, first order, promo conflict

  • referral_delay_calc: scheduled credit date policy + referee order

  • referral_handoff_376: pre-filled REF-INV agent ticket

Encrypted DTC Scenario

Skincare brand ReferralCandy, 58 referral tickets/month, referral_bot_resolution 0% before Qstomy.

After REF-BOT MVP: referral_bot_resolution 67%, referral_ticket_deflection -38%, referral_link_copy_from_bot 124/month, referee_convert_post_bot +16%.

Explore AI support, Shopify, request a demo.

Complement #376

#377 = self-service bot. #376 = agent fraud investigation and manual credit.

What is the checklist for launching the referral bot?

REF-BOT Checklist (10 steps)

  1. Audit referral tickets for last 90 days and prioritize MVP intents

  2. Write 8 corpus blocks for REF-RAG

  3. Connect Smile/ReferralCandy API for referral link + status

  4. Deploy link, delay, code_fail, rules intents

  5. Live status response templates + RAG rules

  6. Handoff referral_handoff_ref_inv → #376 configured

  7. Customer account widget + /referral bot trigger

  8. Anti-hallucination tests on status and amounts with 20 questions

  9. Weekly referral_bot_resolution dashboard

  10. Quarterly corpus review synced with marketing rewards

At a glance

  • #377 = self-service referral AI bot, not REF-INV #376

  • REF-BOT: 10 intents, API + RAG

  • MVP: link, delay, code_fail, rules

  • Bot tracks status, does not credit reward

  • KPI referral_bot_resolution: target > 65%

FAQ

Difference with referral support #376?
#377 = self-service bot for rules and tracking. #376 = agent investigation for fraud and credit.

Can the bot credit €10 to the referrer?
No. Handoff to #376 REF-INV if reward is missing after the delay period.

Relationship with LOY-BOT #375?
LOY-BOT = loyalty points. REF-BOT = distinct referral bot, separate intents.

Blocked referee checkout code?
referral_code_fail intent: min cart, first order, promo.

Does the referee see the referrer's email?
No. Referee bot = rules and code, no referrer data.

Going further

Test the complete referral mystery shop: bot link → referee eligible order → bot delay → completed status without REF-INV agent ticket.

Share this guide #377 with product and marketing: a connected and well-governed REF-BOT transforms opaque referral into self-service that converts referees and deflects support tickets.

Enzo

August 14, 2026

Convert over 2,000 customers on average per month with Qstomy.

The world’s 1st Shopify AI dedicated to customer conversion

Empowering 200+ e-commerce merchants

Subscribe to the newsletter and get a personalized e-book!

No-code solution, no technical knowledge required. AI trained on your e-shop and non-intrusive.

*Unsubscribe at any time. We do not send spam.

Subscribe to the newsletter and get a personalized e-book!

No-code solution, no technical knowledge required. AI trained on your e-shop and non-intrusive.

*Unsubscribe at any time. We do not send spam.