Last updated: November 2026 · By Khanh Nguyen, marketing lead at SMK Vietnam — a decade in SEO, content and AI-assisted marketing for Vietnamese and foreign brands

After a decade in SEO, content, and AI-assisted marketing in Vietnam, I have seen dozens of beautiful customer journey maps die in a slide deck. They were drawn in a workshop, based on what the team believed customers do, and never touched again. The real Vietnamese journey is messier: a TikTok video, a price check on Shopee, three questions in Zalo, a cousin’s opinion, a COD order, and — if you are lucky — a second purchase two months later. AI finally makes it practical to map that journey from evidence instead of assumptions, and to keep the map alive.

Quick answer: how do you use AI for customer journey mapping in Vietnam?

Use AI to build the Vietnamese customer journey map from real evidence: feed it Zalo and Messenger chat logs, Shopee and Google reviews, CRM order data and GA4 paths, and ask it to cluster the questions, objections and channels customers use at each stage — discovery, consideration, purchase, delivery and repeat. Then let AI flag the biggest drop-off points and draft the fix for each. Keep humans in charge of validating the map with 5–10 real customer interviews and of deciding which moments get budget.

In this playbook

Why is the Vietnamese customer journey different?

Western journey templates assume a funnel: search, website, cart, email. In Vietnam the journey is chat-first and platform-hopping. Discovery happens on TikTok or Facebook, validation happens on Shopee reviews and in Facebook groups, and the actual decision often happens in a Zalo or Messenger conversation where the customer asks about price, freeship, warranty and whether the product is “chính hãng”. Many orders are still cash on delivery, so the journey does not end at checkout — a failed delivery is a lost sale. Foreign brands that import a global journey map miss the moments that actually decide revenue here.

3–5
touchpoints before a typical first online order
2–3
platforms a shopper hops between in one journey
COD
still common, so delivery is part of the sale
60–90
days: the window where most repeat orders are won or lost

Figures are directional practitioner estimates from SMK Vietnam client work across e-commerce, F&B and services; validate against your own data.

What does the Vietnamese customer journey look like, stage by stage?

Here is the five-stage frame I start every client with. AI’s job is to fill each row with evidence for your category rather than generic assumptions.

StageTypical touchpointsWhat AI reads
1. DiscoveryTikTok, Facebook Reels, KOC videos, YouTube, GoogleComment themes, hook performance, search queries
2. ConsiderationShopee reviews, Facebook groups, comparison posts, websiteReview objections, competitor mentions, GA4 paths
3. Conversation & purchaseZalo OA, Messenger, TikTok DM, marketplace chat, checkoutChat logs: questions, hesitation, time-to-reply vs conversion
4. Delivery & first useShipping updates, COD call, unboxing, ZNS messagesFailed-delivery reasons, early reviews, support tickets
5. Repeat & advocacyZalo broadcasts, loyalty, community, referralsCRM repeat intervals, churn signals, referral sources

Where does AI actually help in journey mapping?

AI does not replace customer research. It removes the part that used to make journey mapping too expensive: reading thousands of messages, reviews and rows of data by hand. These are the four jobs where I see it pay off fastest.

Chat-log mining

Export 3 months of Zalo OA and Messenger conversations, anonymise them, and have Claude or ChatGPT tag each by stage, question type and outcome. You get the real objection list in an afternoon.

Review & comment clustering

Cluster Vietnamese reviews from Shopee, Google Maps and TikTok comments into consideration-stage themes: size, authenticity, delivery speed, after-sales. Each cluster becomes a content or FAQ task.

Path & drop-off analysis

Ask Gemini in Sheets or a GA4 export to compare converting and non-converting paths. AI is good at spotting the step where Vietnamese mobile users abandon — often a shipping-fee surprise.

Persona-journey variants

One map rarely fits all. AI can split the journey by segment — Hanoi office worker vs Mekong Delta household buyer vs Viet Kieu gifting — when the data supports it.

For tool choice, see our Claude playbook for long chat-log analysis, the ChatGPT playbook for spreadsheet work, and Perplexity for category and competitor desk research. The full picture is in The Complete AI Marketing Stack for Vietnam.

“The most valuable page in a Vietnamese journey map is usually not discovery. It is the Zalo conversation where a customer asks ‘freeship không shop?’ and waits 40 minutes for an answer.”

Workshop maps vs AI-evidence maps: what changes?

Workshop-only map
  • Built from team opinions and global templates
  • Misses chat, COD and delivery moments
  • One persona, one path
  • Updated once a year, if ever
  • No link to budget or KPIs
AI-evidence map
  • Built from chat logs, reviews, CRM and GA4
  • Captures Zalo, Messenger and delivery steps
  • Segment variants where data supports them
  • Refreshed monthly with new exports
  • Each drop-off tied to an owner and a metric

How do you build an AI-assisted journey map in 5 steps?

This is the workflow we run for clients. A first version takes about two weeks of part-time work for one marketer plus one analyst.

1
Collect and anonymise the evidence (days 1–3).

Export 90 days of Zalo OA and Messenger chats, marketplace reviews, CRM orders and GA4 paths. Strip names and phone numbers before anything touches an AI tool — see our stack guide on handling personal data under Vietnam’s PDPL.

2
Tag every interaction by stage (days 4–6).

Give the AI your five stages and a tag list (question, objection, channel, outcome). Ask it to output a table, then spot-check 50 rows by hand. Fix the prompt until accuracy is above roughly 90%.

3
Draft the map and the drop-offs (days 7–8).

Have AI summarise each stage: top 5 questions, top 3 objections, main channels, and where volume falls most sharply. This becomes your draft map.

4
Validate with real customers (days 9–12).

Run 5–10 short Zalo calls or interviews with recent buyers and non-buyers. AI can draft the Vietnamese interview guide; a human runs it and listens for what the data cannot show — feelings, family influence, trust.

5
Turn drop-offs into owned fixes (days 13–14).

For the top three drop-offs, let AI draft fixes — a Zalo auto-reply for shipping questions, a review-backed comparison block, a 30-day ZNS re-order reminder — then assign an owner, a metric and a review date to each.

What are the most common journey-mapping mistakes?

MistakeFix
Mapping only the website and adsInclude chat, marketplace and delivery steps; they carry most of the decisions
Pasting raw customer chats into public AI toolsAnonymise first; use business plans with data controls
Trusting AI tags without checkingHand-check a sample; Vietnamese slang and teencode trip models up
Translating a HQ journey map from EnglishRebuild from local evidence, then compare with HQ’s version
Stopping at the first purchaseMap delivery, first use and the 60–90 day repeat window
A map with no ownersEvery drop-off gets a person, a KPI and a monthly review
Related reading

Key takeaways

  • The Vietnamese customer journey is chat-first and platform-hopping: TikTok or Facebook, then Shopee reviews, then Zalo or Messenger, then COD delivery.
  • AI makes evidence-based journey mapping affordable by tagging chats, reviews and paths at scale.
  • Always anonymise customer data and hand-check AI tags, especially for Vietnamese slang.
  • Validate the AI draft with 5–10 real customer conversations before acting on it.
  • A map is only useful if every drop-off has an owner, a metric and a monthly refresh.

About SMK Vietnam: SMK Vietnam (smkvietnam.com) is a marketing hub helping Vietnamese companies and foreign brands from the US, Japan, Korea, Thailand and Europe plan, run and measure AI-assisted marketing in Vietnam.

FAQ: AI customer journey mapping in Vietnam

What is AI customer journey mapping?

It is building a map of how customers discover, evaluate, buy and rebuy from you using AI to analyse real data — chat logs, reviews, CRM orders and web analytics — instead of relying only on workshop assumptions.

Which data sources matter most in Vietnam?

Zalo OA and Messenger chat logs, Shopee, Lazada and TikTok Shop reviews, Google Maps reviews, CRM or order data including COD delivery outcomes, and GA4 paths. Chat logs usually reveal the most about why people buy or drop off.

Is it legal to analyse customer chats with AI in Vietnam?

It can be, if you follow Vietnam’s personal data protection rules: have a lawful basis, remove names and phone numbers before analysis, use AI tools with business data controls, and document the processing. Check specifics with legal counsel.

Which AI tools work best for journey mapping?

Claude handles long Vietnamese chat exports well, ChatGPT and Gemini are strong for spreadsheet tagging and charts, and Perplexity helps with category research. Most teams combine two of them rather than buying a dedicated journey platform at first.

How long does a first AI-assisted journey map take?

About two weeks of part-time work for one marketer and one analyst: three days collecting data, a week tagging and drafting, and a few days of customer validation and fix planning.

How often should the journey map be updated?

Refresh the data monthly and do a deeper review each quarter or before major sale seasons such as Tet and the double-day sales, because Vietnamese channel behaviour shifts quickly.

Want an evidence-based journey map for your Vietnam business?

SMK Vietnam builds AI-assisted customer journey maps from your real Zalo, marketplace and CRM data — and turns the drop-offs into a 90-day fix plan.

Talk to SMK Vietnam

Leave a Reply

Your email address will not be published. Required fields are marked *