How to redesign discovery, choice, purchase, post-purchase care, and loyalty around customer intent rather than website navigation.


1. Why the Customer Journey Is Changing

For two decades, digital commerce taught customers how to behave like site operators. They learned the menu structure, searched by keywords, opened filters, compared product pages, and eventually reached checkout. That model still works-but it puts the work of translation on the customer.

The new opportunity is to design commerce around intent. Instead of requiring a shopper to know the vocabulary of the catalog, the experience can start with a human request: “I need something comfortable for an outdoor wedding in humid weather,” “I need a gift for someone who just started a new job,” or “I bought these pants before; what is similar but more relaxed?”

This guide develops your original five-moment framework: Discover, Choose, Purchase, Post-Purchase, and Loyalty. The important idea is continuity. These are not five separate AI features. They are five connected moments in one relationship.

Research note: Your original blueprint defines the five connected moments and states that every interaction should help the customer discover, decide, or resolve a real need. your original source, “The Modern Customer Journey Blueprint”

 
The central design principle: Do not ask, “Where can we insert AI?” Ask, “Where is the customer doing unnecessary work-and what would helpful intelligence look like there?”\

What you will learn

• How conversational discovery changes the first step of shopping.
• Why more choice can create more anxiety-and how contextual comparison solves it.
• What must remain consistent when the customer moves from an AI assistant to a commerce site or human associate.
• How post-purchase service becomes a trust engine rather than a support queue.
• How loyalty can evolve from discounts and points into recognition and useful continuity.

CONTEXT & DATA

2. The Evidence: This Is Becoming a Different Commerce Model

The strategic shift is not happening in a vacuum. Recent industry research shows that shoppers are increasingly using AI during discovery and research, while retailers are beginning to prepare for agents that can assemble and eventually execute more of the journey.

Signal	What it suggests	Why it matters
39% of U.S. consumers surveyed by Adobe said they had used generative AI for online shopping in early 2025.	AI-assisted shopping is moving beyond experimentation.	Discovery experiences need to work when the shopper starts outside the retailer website.
53% said they planned to use generative AI for shopping that year.	Intent is likely to become increasingly conversational.	Brands should prepare before usage becomes a default behavior.
Adobe observed a 4,700% year-over-year increase in traffic from generative-AI sources to U.S. retail sites in July 2025.	AI can become a meaningful traffic and referral layer even when it is still smaller than traditional channels.	The handoff from AI discovery to retailer experience needs to be strong.
McKinsey reported that 44% of users who had tried AI-powered search said it had become their primary and preferred source for internet searching, versus 31% preferring traditional search.	Some customers are beginning to prefer synthesized answers over link lists.	The competitive question becomes: can an AI system accurately understand and recommend your products?
McKinsey estimates agentic commerce could mediate $3T–$5T of global consumer commerce by 2030 under moderate scenarios.	The long-term opportunity extends beyond answering questions into delegated shopping.	Product data, availability, policies, and transaction rails become strategic infrastructure.

Research note: Adobe consumer/traffic data: 2025 survey and Analytics reporting.  McKinsey agentic-commerce estimates and search-preference finding: October 2025 report.

Important nuance: These numbers do not mean every shopper wants an autonomous buying agent today. They mean the direction of travel is changing-and retailers have to design for a wider range of customer intent and delegation.


 

EDUCATION

3. The Mental Model: From Website to Intelligent Journey

The biggest mistake teams make is to think the “agentic customer journey” means replacing a website with a chatbot. It is much broader. The interface may be chat, voice, search, an external assistant, a product page, a store associate, or a blend of all of them.

Traditional question	Modern question
How do we get the shopper to our homepage?	How do we become useful wherever the shopper expresses intent?
How do we improve search relevance?	How do we understand what the shopper is actually trying to accomplish?
How do we show more products?	How do we reduce the work required to choose the right product?
How do we increase checkout conversion?	How do we remove mechanical friction while preserving trust and control?
How do we reduce support contacts?	How do we resolve issues with less effort and better continuity?
How do we increase repeat purchases?	How do we earn the next interaction by proving we remember what matters?

A useful way to think about agency

Customers will likely occupy different points on a delegation spectrum. One person may want inspiration but prefer to click through products themselves. Another may ask an assistant to narrow the options. A third may ask the assistant to build a basket but still approve the final purchase. A fourth may delegate the whole transaction within explicit rules.

Design for variable delegation	Do not force autonomy. Let customers choose how much work they want the system to perform, and make the boundary visible.

 

Research note: McKinsey describes agentic commerce as a curve of delegation rather than a single leap from human shopping to full autonomy.

CORE FRAMEWORK

4. The Five-Moment Agentic Experience Framework

Your original framework organizes the modern journey into five connected moments. The power of the model comes from the progression: every moment should reduce a different kind of customer uncertainty or effort.

Moment	Customer question	Agentic capability	Human value preserved
1. Discover	“What might work for me?”	Interpret natural-language, voice, and visual intent.	Being understood without learning retailer vocabulary.
2. Choose	“Which is right for my situation?”	Compare trade-offs, fit, suitability, and context.	Confidence, taste, and personal judgment.
3. Purchase	“Can I complete this easily and safely?”	Use context, identity, loyalty, and transaction capabilities.	Trust, transparency, and control.
4. Post-Purchase	“What happens now, especially if something goes wrong?”	Proactive updates, conversational resolution, contextual handoffs.	Empathy, discretion, and accountability.
5. Loyalty	“Do you actually know me?”	Useful recognition and personalized continuity.	Relationship, meaning, and consent.

 A simple test: If the experience cannot answer “What customer uncertainty are we reducing?” it is probably a feature looking for a problem.

5. Moment One - Discover: Design for Human Intent

Discovery is where the old commerce model asks the shopper to translate themselves into a database. The modern model tries to perform that translation on the customer’s behalf.

What discovery looks like in the real world

Imagine a shopper preparing for a three-day work trip. They do not necessarily think, “I need category X, material Y, color Z.” They may think: “I need two outfits that can survive a flight, meetings, and dinner without taking up much suitcase space.” That statement contains functional constraints, occasion, duration, and implicit style needs.

A useful discovery experience should recognize the intent hidden inside the sentence. It should identify missing information-perhaps whether the trip is business formal or business casual-then narrow the field without turning the interaction into an interrogation.

Three layers of discovery

Intent: What is the customer trying to accomplish?

Context: What circumstances change what “good” means-weather, occasion, budget, timing, fit, destination, or recipient?

Evidence: Which products, policies, reviews, and operational signals prove that a recommendation is credible?

Example prompt

Customer	“I’m going to an outdoor summer wedding in Atlanta. I want something polished but breathable, and I do not want to spend more than $250.”

 

Weak response	“Here are 48 products in women’s dresses sorted by popularity.”

 

Helpful response	“For hot, humid weather, I’d prioritize breathable fabrics and silhouettes that do not feel restrictive. I can narrow this to options under $250 and compare the best three by comfort, dressiness, and ease of care.”

 

Research note: Your source recommends replacing rigid descriptions with occasion and lifestyle context and enabling discovery through voice, natural language, and image reference. your original source, Moment 1: Discover

Try This

• Collect the 25 most common natural-language discovery questions.
• Identify which questions cannot currently be answered from catalog data.
• Rewrite a sample of product descriptions around real customer situations.

6. Moment Two - Choose: Replace Choice Overload With Confidence

Discovery gets the customer to a set of plausible options. Choice is where the experience earns trust. The goal is not to prove the AI has considered everything. The goal is to help the customer understand the trade-offs.

Why more choice is not always better

Traditional commerce often treats breadth as value. But when two or three products are all plausible, more options can increase hesitation. The assistant should therefore explain differences in the dimensions that matter for this customer-not produce an exhaustive encyclopedia of features.

Customer constraint	What the assistant should compare	Useful explanation
Hot + humid destination	Breathability, weight, airflow, care	“Option A is lighter and easier to pack; Option B is more structured but may feel warmer.”
Frequent travel	Wrinkle resistance, packability, durability	“Option A is the easier travel choice because it is less demanding to maintain.”
Uncertain fit	Silhouette, stretch, size guidance, reviews	“If you prefer more room through the hip, Option B is the safer choice.”
Limited budget	Price, versatility, long-term usefulness	“Option A costs more, but it covers both your work and weekend use cases.”

The best recommendation is explainable

A recommendation should be traceable to stated customer priorities. “Because you said X, I would choose Y” is more trustworthy than “Y is our top pick.” It also gives the customer an opportunity to disagree.

The disagreement test: A good recommendation gives the customer enough reasoning to say, “Actually, comfort matters more to me than versatility.” That creates a better next turn instead of a dead end.

 

Try This

• Document the top five causes of choice paralysis.
• Create comparison templates tied to real customer constraints.
• Require recommendations to include a rationale.
• Test whether customers can easily override a recommendation.

7. Moment Three - Purchase: Make Speed Feel Safe

Once a customer has decided, the experience should not suddenly become administrative. The purchase moment is where context should travel forward rather than disappear.

What should carry into checkout?

·        The products the customer selected and why they selected them.

·        Relevant identity and account information that the customer has permitted the system to use.

·        Shipping preferences and delivery expectations.

·        Loyalty status or benefits that genuinely apply.

·        Any important constraints or decisions made during the conversation.

Remove friction, not visibility

There is an important distinction between reducing steps and hiding information. The experience can automate repetitive entry while still showing the customer the final items, price, delivery expectation, and action being taken. That preserves agency.

Friction today	Better pattern	Trust safeguard
Searching for loyalty benefits	Surface eligible benefits automatically.	Explain what was applied.
Repeating shipping preferences	Reuse permitted preferences.	Show the selected destination and timing.
Switching between chat and website	Deep-link or hand off with context.	Preserve the conversation summary.
Unclear automated action	Ask for confirmation before consequential action.	Provide an easy correction path.

 

8. Moment Four - Post-Purchase: Design the Care Seam

The post-purchase moment is often treated as an operational afterthought. It should be the opposite. A delivery delay, damaged product, wrong size, or refund question is when customers learn whether the brand is genuinely on their side.

Proactive beats reactive

The source framework recommends notifying customers before they are forced to discover a problem themselves. This sounds simple, but it requires connected operational signals: order status, shipment events, inventory, delivery expectations, and customer communication.

The handoff is part of the experience

When an automated assistant cannot resolve an issue, a human should receive enough context to continue the conversation. A handoff that says “Please explain your issue to the next representative” is technically a transfer but functionally a restart.

Bad handoff	Designed handoff
“A representative will assist you.”	“I’m connecting you with a specialist and sending them your order details plus the steps we already tried.”
Customer repeats order number, issue, and history.	Conversation summary, order, and relevant status travel automatically.
Agent must discover the problem from scratch.	Human receives the customer’s goal, prior attempts, and current emotional context where relevant.

Try This

• Define the top five post-purchase failure scenarios.
• Identify the operational signal that could detect each scenario early.
• Design the customer message before building the automation.
• Specify exactly what context a human receives during escalation.

9. Moment Five - Loyalty: Earn the Relationship

Loyalty is often measured through points, frequency, and campaigns. The deeper opportunity is recognition: helping a customer feel that the brand remembers what they have already told it and uses that knowledge to be helpful.

Recognition is more than personalization

Personalization can mean “show this customer more of the same products.” Recognition means understanding the relationship: what the customer buys repeatedly, what they return, what they prefer, what they have asked for, and what they explicitly want the brand to remember.

A useful loyalty interaction

Instead of	“20% off new arrivals. Shop now.”

 

Consider	“You’ve bought this relaxed-fit trouser twice and kept both. A new version is available in a lighter fabric that may work better for summer travel.”

 

The second example is more valuable because it provides a reason. It also leaves room for the customer to say no.

Consent matters

Deep recognition should never become a license to infer everything about a person. Use the context the customer has knowingly provided and give customers meaningful control over what is remembered, where it is used, and how it affects recommendations.

Research note: Your source recommends transforming points into service, using conversational memory based on what customers actually keep, and honoring consented preferences across interfaces. your original source, Moment 5: Loyalty

OPERATING FOUNDATION

10. What Has to Exist Behind the Experience

The visible experience is only the tip of the system. Agentic customer journeys depend on reliable underlying capabilities. A beautiful conversational interface cannot compensate for stale inventory, fragmented identity, incomplete product data, or disconnected order systems.

Foundation	Why it matters	Typical failure mode
Product truth	Answers “What is this product really like?”	Descriptions and structured attributes contradict each other.
Customer identity	Answers “Who is this customer and what context can we use?”	The customer appears anonymous after switching channels.
Inventory & availability	Answers “Can the promise actually be fulfilled?”	AI recommends something unavailable or unavailable in the needed size.
Pricing & promotions	Answers “What will this actually cost?”	Recommendation and checkout show different economics.
Order & fulfillment status	Answers “What happened after purchase?”	Support sees less context than the customer does.
Conversation memory	Answers “What have we already discussed?”	Human or system asks the customer to repeat themselves.
Escalation controls	Answers “When should a person take over?”	Automation continues even when uncertainty or sensitivity is high.

Think in capabilities, not channels

A customer may move from search to an external AI assistant, to your website, to a mobile app, and then to a store. The capability should travel even when the channel changes. That is the real meaning of “customer is the channel.”

Architecture question: For each moment in the journey, ask: “What information must be true, current, and available for the system to be useful?”

 

MEASUREMENT

11. How to Measure an Agentic Customer Journey

A common mistake is to measure an AI experience by activity: number of conversations, prompts, recommendations, or automated actions. Those numbers can grow while the customer experience gets worse. Measure outcomes instead.

Dimension	Example metric	What it tells you
Discovery	Time to relevant shortlist	Whether the system reduces navigation work.
Understanding	Successful intent capture rate	Whether the assistant understands what customers mean.
Choice	Recommendation acceptance or confident selection	Whether explanations improve decision-making.
Purchase	Completion with fewer unnecessary steps	Whether context actually reduces friction.
Service	Time to resolution + repeat contact	Whether automation solves rather than deflects.
Handoff	Context completeness at escalation	Whether the human can continue without restarting.
Loyalty	Useful re-engagement / retention	Whether recognition creates ongoing value.

A practical measurement hierarchy

1. Customer outcome
2. Customer effort
3. Quality/accuracy
4. Business outcome
5. Automation efficiency

The ordering matters. Automation efficiency should be the last question, not the first. If the system is efficient at producing the wrong answer faster, the metric has become a trap.

Research note: McKinsey’s recent agentic-commerce work emphasizes delivered value, trust, and operational reliability-not simply interaction volume-as the basis for competition in agent-mediated commerce.

FROM FRAMEWORK TO ACTION

12. The 90-Day Implementation Roadmap

Do not attempt to transform the entire journey at once. A practical rollout starts with one high-value customer problem, proves value, and then expands the underlying capabilities.

Period	Focus	Outputs
Days 1–30: Understand	Journey audit + customer questions + data gaps	Top friction map, priority intents, foundation gaps, success measures
Days 31–60: Prove	Prototype one moment end-to-end	Working experience, human fallback, measurement, customer feedback
Days 61–90: Operationalize	Test + refine + prepare to scale	Exception playbook, governance, architecture backlog, scale decision

Choose the first use case carefully

A good first use case has a clear customer problem, enough underlying data to be useful, manageable risk, and a measurable outcome. It should not require solving the entire commerce stack before a customer can experience value.

Try This

• The customer problem is specific.
• A baseline exists.
• The underlying data is trustworthy enough.
• A human fallback is designed.
• The team can measure success.
• The experiment has a clear decision date.

Workshop exercise: map one journey

6. Write the customer goal in one sentence.
7. List the current steps from intent to outcome.
8. Circle every step where the customer must translate, search, repeat, compare, or wait.
9. Mark which steps are mechanical and which require judgment.
10. Design the minimum helpful intervention.
11. Define the human fallback.
12. Choose three outcome measures.

PRINTABLE WORKING SESSION

13. The Customer Journey Design Canvas

Use this page in a workshop with product, design, engineering, merchandising, operations, customer care, and analytics. The goal is to make the customer’s work visible before deciding what technology to build.

Canvas field	Write your answer
Customer goal
Customer’s natural-language request
Current journey
Biggest friction / uncertainty
Moment: Discover / Choose / Purchase / Post-Purchase / Loyalty
Data required
System capability required
What the agent may do
What the human must decide
Escalation trigger
Customer-facing explanation
Success metric

Five questions to pressure-test the idea

• Would the customer care about this if the word “AI” disappeared?
• Does it remove a real step, uncertainty, or delay?
• Can the system explain why it is making the recommendation or taking the action?
• What happens when it is wrong or uncertain?
• Can the customer move to a human without starting over?

FINAL PERSPECTIVE

14. Closing: Build for the Customer, Not the Technology

The future customer journey will not be defined by whether a brand has a chatbot, a shopping agent, or a new AI search experience. It will be defined by whether customers can accomplish what they came to do with less unnecessary effort and more confidence.

The five moments provide a durable way to think about that work. Discover the customer’s intent. Help them choose by explaining trade-offs. Remove mechanical friction from purchase without removing control. Stay present when the order becomes a problem. And earn loyalty by turning memory into genuinely useful recognition.

The homepage is not disappearing. It is becoming one interface among many. The strategic asset is the capability underneath it: product truth, identity, operational context, intelligent retrieval, transaction infrastructure, and well-designed human seams.

The question to take into your next planning meeting “What unnecessary work are we asking the customer to do today-and what would it look like for us to own more of that work?”

I hope this guide points you in the right direction! It’s not meant to be an exhaustive manual, just a solid baseline you can tweak based on where you are. If you want to chat more, dig into the specifics, or bounce some ideas around, you know where to find me.

Cheers,

Darko