Customers
Customer journey map with AI: how to build and improve it
A practical guide to building a customer journey map with the help of AI. What the map contains, a quick first read to find where customers get stuck, the full method and where AI can improve each stage.
What a customer journey map is and why it helps
At an estate agency in Leicester, a buyer emails for the third time in a few days to ask how the sale is going. The negotiator replies, but cannot work out why someone who was delighted after the viewing now sounds anxious. That change of mood belongs on a customer journey map, and it almost never shows up in sales reports.
The customer journey is a customer's whole path with your company, from first hearing about you to buying again or recommending you. A customer journey map writes that path down, stage by stage, with what the customer does, thinks and feels at each point.
Many companies draw this map in a meeting room, based on what the team believes happens. AI helps you check it against what customers actually say in emails, messages, reviews and surveys. It can read and sort hundreds of comments quickly, something few people ever do by hand, so the map reflects real experience and shows where change is worth the effort.
The four layers of a customer journey map
Every useful map combines these four layers. Leave one out and the map describes internal processes instead of the customer's experience.
Stages
The phases a customer moves through. A common outline is discovery, consideration, purchase, onboarding, support and loyalty. At an estate agency they might be first enquiry, viewings, offer, conveyancing and completion.
Touchpoints
Every place a customer meets your company. A property portal listing, the website, a phone call, a WhatsApp message, a confirmation email or a face-to-face meeting.
Emotions
How the customer feels at each stage. Excitement while choosing, worry while waiting, relief when someone explains the next step. Negative emotions often signal a complaint or a lost customer ahead.
Pain points
The moments when something goes wrong or takes more effort than it should. A vague timeline, a long form, a reply that takes days. These are where every improvement starts.
A first map from your last 30 customer messages
Before the full method, this first read gives you a rough picture in about ten minutes. All you need is a spreadsheet and an AI assistant your company has approved, such as ChatGPT, Claude or Gemini.
- 01
Gather your last 30 customer messages
Emails, WhatsApp messages or web forms, from different customers. Remove names, phone numbers, addresses and reference numbers before doing anything else.
- 02
Sort them by stage
Ask the assistant to assign each message to a stage from your list and sum up in a few words what the customer wanted. Check the sorting yourself, because you know your business better than the model.
- 03
Count per stage
Where questions pile up, the journey does not explain itself. A stage full of "any news on my case?" messages points to missing information, even when the customer is as keen as ever.
- 04
Mark the handovers
Flag the messages that arrive just after the customer moves to another person, channel or tool, for example from sales to admin. That is usually where the first pain point worth tackling sits.
How to build a customer journey map with AI, step by step
The full method builds on that first read with more material and what your team knows. Everything starts from real material, and every conclusion gets checked.
- 01
Choose the customer and the journey
Pick one type of customer and one specific journey, such as a first-time buyer. A map for every customer at once ends up too general to decide anything.
- 02
Gather and anonymise the material
Collect reviews, emails, call notes, open survey answers and chat transcripts, plus CRM data such as cancellations or reasons for leaving. Strip out personal data and use only tools and accounts your company has approved.
- 03
Ask the AI to sort with evidence
Ask it to assign each comment to a stage and identify the touchpoint, tone and problem. Insist on verbatim quotes to back up every finding. Work in batches and paste the results into a spreadsheet so you can count and filter them.
- 04
Check with your team and with customers
Go through the results with the people who deal with customers every day. Then call a handful of customers, new and long-standing, including one who did not come back, to confirm what you found and hear what never gets written down.
- 05
Draw the map and set priorities
Lay out stages, touchpoints, emotions and pain points in one view. Score each problem on how often it happens and how much it affects sales or satisfaction, then pick two or three improvements to start with.
Where AI can improve each stage of the customer journey
Once you know your pain points, these are common improvements. Start with the handovers you marked, because that is where a happy customer turns into a worried one.
Discovery and consideration
AI turns the questions people ask before buying into useful content, such as guides or comparisons. An assistant on your website or WhatsApp answers questions using information your team has approved.
Purchase
Analysing the questions asked before signing or paying shows what holds people back, such as a cost that appears late or a document nobody explained. AI spots the pattern and your team decides the change.
Onboarding
After the sale, automatic messages explain the next step, the documents needed and who to contact if something goes wrong. At a service company, collecting documents can be guided step by step with reminders.
Support
AI sorts incoming messages, prepares draft replies and answers routine questions such as order status. Complaints and decisions involving money go to a person.
Loyalty
Customer history lets you send relevant reminders, such as an annual service review. Regular review analysis tells you whether your changes are working.
Example
A Leicester estate agency finds where its buyers start to worry
Back to the agency from the start, a typical case with four negotiators and one sales progressor who handles the paperwork. The manager starts with the last 30 messages, then moves on to the full method using anonymised emails, WhatsApp messages and recent Google reviews. The AI sorts each comment by stage and supplies verbatim quotes.
The result surprises the team. Viewings and negotiation generate few complaints. Questions pile up between offer accepted and completion, exactly when the sale passes from the negotiator to the sales progressor and the solicitors take over. Buyers ask what is missing, when exchange will happen and what the mortgage lender is waiting for. Several lukewarm reviews mention "hearing nothing" during those weeks.
The agency makes two changes. When an offer is accepted, the buyer receives a clear list of documents and steps, with the name of the person who will keep them informed. Each week, an automated workflow uses AI to draft a short status update from the CRM, which the negotiator checks and sends.
To measure it, the manager counts each month the "any news?" messages per sale, the time it takes to gather the paperwork and the reviews that mention the process. Every quarter she sorts 30 recent messages again, in case the bottleneck has moved to another stage.
What AI cannot see in your reviews and emails
Written feedback is skewed. Reviews tend to come from very happy or very unhappy customers, and support emails are mostly about problems. Customers who left without a word do not show up anywhere. AI summarises what it is given, biases included, and can generalise from a handful of cases. That is why you ask for verbatim quotes and check a sample by hand. Reviews, emails and call notes also contain personal data covered by UK and EU data protection law, so decide who can access the material and how long it is kept.
To start, sort your last 30 customer messages and mark the handovers. You will then know which stage to look at first and which customers to call.
Obinova is an AI agency based in Barcelona and Paris. In our AI audit and consulting, we review your customer-facing processes, the material you have and the tools you already use with your team, and choose with you the improvements with the most impact. A mission starts with a launch kick-off, followed by follow-ups several times a year. If the map reveals repetitive tasks, such as status updates, our automation service builds them in n8n, connected to your CRM, email or WhatsApp, and documented so your team can improve them.
Frequently asked questions
What is a customer journey map?
It is an outline of the path a customer takes with your company, from first discovering you to buying again or recommending you. It records the stages, touchpoints, emotions and pain points at each moment. It shows where customers get stuck and helps the whole team agree on what to improve first.
What are the stages of the customer journey?
A typical outline runs from discovery through consideration, purchase, onboarding and support to loyalty, but there is no fixed list. An online shop might treat delivery as a stage of its own, and an estate agency might use viewings, offer, conveyancing and completion. What matters is that the stages reflect how your real customers buy and use what you offer.
Customer journey map vs sales funnel: what is the difference?
A sales funnel looks from the company's side and counts how many prospects move from one phase to the next until the sale. A customer journey map looks from the customer's side and includes what happens after buying, their emotions and every touchpoint. They work together. The funnel shows where you lose people and the map helps explain why.
Which AI tool do I need, and what does it cost?
A general assistant such as ChatGPT, Claude, Gemini or Microsoft Copilot, on an approved business account, is enough to sort and summarise emails, reviews and surveys. If the volume is high or the analysis needs repeating monthly, it is worth automating with n8n. The cost of outside support depends on the material and the scope, and we quote after a first meeting.
Can I put customer emails and reviews into an AI tool?
Yes, with care. First remove names, phone numbers, addresses, email addresses and reference numbers. Use only the tools and account types your company has approved, and check the provider's terms on how your data is used. Write these rules into a short usage policy that everyone on the team knows and follows.
Is there a customer journey map example for a small business?
At an estate agency, the stages might be first enquiry, viewings, offer, conveyancing and completion. For each one you note the touchpoints, such as the property portal, WhatsApp or the solicitor, how the buyer feels and what goes wrong. The case in this guide shows a common pattern: few complaints about viewings and a pile of questions between offer accepted and completion, when the sale changes hands.