For years, ecommerce teams have worked around the idea that customers rarely move neatly from awareness to purchase.
They search, compare, leave, come back, ask friends, read reviews, check delivery costs and open far too many browser tabs. Google gave this behaviour a useful name: the messy middle.
Agentic commerce will not remove that behaviour. It will change who carries it out.
Instead of the customer manually navigating every stage of exploration and evaluation, an AI agent can increasingly do much of the work on their behalf. It can interpret a need, gather options, compare products, check constraints and present a shortlist before the customer reaches a retailer’s website.
For shoppers, the journey may feel much shorter.
For brands, it becomes more difficult to see and influence.
From a search query to a customer brief.
Traditional search has always required customers to translate their needs into keywords.
Someone looking for a coat might search for:
- Men’s waterproof commuter coat
- Best waterproof coat under £200
- Sustainable UK rain jackets
- Lightweight waterproof coat reviews
They may repeat and refine those searches several times before finding something that feels right.
An AI conversation can begin with a much richer request:
“I need a waterproof jacket for commuting that does not look too outdoorsy. My budget is £200, I would prefer a responsible UK brand, and I need it delivered before next Thursday.”
That is not simply a longer search query. It is a brief.
It contains the customer’s need, intended use, budget, preferences, concerns and deadline. The AI can ask follow-up questions and refine its recommendations without requiring the shopper to start again each time.
This matters because it brings much of the traditional consideration journey into one interaction.
What happens to the messy middle?
Google’s messy middle describes the loop between two behaviours: exploration and evaluation.
Customers explore the options available to them, then evaluate which one best meets their needs. New information often sends them back into exploration, creating a loop that can continue until something gives them enough confidence to purchase.
Agentic commerce does not remove that loop. It moves it behind the interface.
The AI may still:
- Explore multiple brands and retailers.
- Compare similar products.
- Check reviews and ratings.
- Evaluate price and value.
- Consider delivery and returns.
- Check size, fit or compatibility.
- Remove products that fail the customer’s criteria.
- Revisit the search after receiving new instructions.
The customer does not necessarily see every step.
Instead, they see a recommendation, a shortlist or a comparison that reflects work previously carried out across several searches, pages and websites.
The messy middle has not become less complex. It has become less visible.
The customer journey becomes compressed.
The traditional ecommerce journey often looks something like this:
Need recognition, Google search, editorial content, retailer website, collection page, filters, product pages, reviews, basket and checkout.
An AI-influenced journey could look considerably shorter:
Need recognition, AI conversation, product recommendation, product page or checkout.
This is already beginning to affect the type of traffic ecommerce websites receive.
Shopify has reported that more than half of AI-referred sessions begin on product detail pages. It has also found that AI-referred visitors can convert at a higher rate and spend more than visitors arriving through traditional organic search.
This makes sense.
A user arriving from Google may still be researching a broad category. Someone arriving after a detailed AI conversation may already have narrowed the market, compared alternatives and decided that a particular product appears to meet their needs.
The website receives fewer of the early exploratory interactions, but potentially gains a more informed visitor.
That creates a very different challenge for ecommerce teams.
The journey may be shorter, but expectations are higher.
A customer arriving directly on a product page after receiving an AI recommendation is unlikely to browse in the same way as someone arriving on the homepage.
They may be looking for immediate confirmation that the recommendation is accurate.
They want to know:
- Is this the exact product I was told about?
- Is the correct size, colour or specification available?
- Can I trust this retailer?
- When will it arrive?
- Can I return it?
- Does the product genuinely provide the features the agent described?
- Is there anything important the recommendation missed?
If the product page contradicts the information supplied by the agent, confidence can disappear quickly.
A different price, unavailable variant, vague delivery information or unclear product description introduces a new type of friction. The customer may not blame the AI. They may simply decide that the brand is unreliable.
The job of the product page changes from helping someone begin their research to helping them validate a decision that has already partly been made.
Ecommerce friction is changing.
Much of conversion rate optimisation has traditionally focused on visible interaction problems:
- Confusing navigation.
- Poor collection filters.
- Hidden calls to action.
- Complicated forms.
- Weak product imagery.
- Unexpected checkout costs.
These issues still matter. Customers will continue to browse websites, and not every journey will involve an AI agent.
However, agentic commerce creates an additional layer of friction before the customer arrives.
This includes:
- Missing or inconsistent product attributes.
- Incorrect variant relationships.
- Out-of-date pricing.
- Stock information that differs across channels.
- Unclear returns and delivery policies.
- Product descriptions that rely on vague brand language.
- Important information stored in images rather than text.
- Conflicting answers across the website, product feed and marketplace listings.
These are not always thought of as customer experience issues because the customer may never encounter them directly.
An AI system will.
If an agent cannot confidently determine whether a product meets a customer’s needs, it may exclude it from the recommendation. The customer journey can end before the brand knows it began.
Brands must influence a journey they cannot fully see.
This is one of the most important changes agentic commerce brings.
Brands have become used to measuring digital journeys through page views, sessions, clicks and conversion events. Even where attribution is imperfect, there is usually some evidence that a customer visited the website.
When discovery and comparison happen inside an AI platform, many of those interactions take place outside the merchant’s analytics.
The brand may only see the final referral, the product page visit or the order. It may see nothing at all if the AI does not recommend its products.
This means ecommerce teams need to think beyond website traffic.
The new questions include:
- Does the brand appear when customers describe relevant needs?
- Which products are agents recommending?
- Is the product information being represented accurately?
- Which questions can agents answer confidently?
- Where are missing attributes preventing inclusion?
- Are customers arriving with different expectations?
- Do AI-referred visitors behave differently from search or paid traffic?
The early customer journey becomes a visibility and representation challenge as much as a traffic acquisition challenge.
The role of search is changing, not disappearing.
It would be easy to see agentic commerce as the replacement for search. That is too simplistic.
AI systems still depend on information from websites, product feeds, catalogues, reviews and trusted sources. Search engines are also adding conversational and generative experiences within their own platforms.
The distinction between search, recommendation and guided shopping is becoming less clear.
SEO remains important because brands still need accessible, authoritative and well-structured content. Product feeds remain important because agents need current commercial information. Reviews matter because they provide evidence. Digital PR and brand authority matter because an agent needs signals that a source can be trusted.
The change is that these activities no longer exist only to earn a click.
They also help AI systems understand when a brand should be included in an answer.
Journey mapping needs to start earlier.
At blubolt, we use Journey-Led Growth to understand how customers move across channels, where friction appears and which improvements are most likely to create value.
Agentic commerce expands that journey.
It is no longer enough to map a route that begins at the homepage or even at the Google search result.
The journey may begin with a conversation in ChatGPT, Google AI Mode, Copilot or another assistant. It may move through a generated comparison, a product recommendation and an external checkout before the customer visits the main website.
The agent becomes another participant in the journey.
That means journey mapping needs to consider:
- What the customer is trying to achieve.
- How they may describe that need to an AI.
- Which information the agent needs to make a recommendation.
- Which sources shape the agent’s response.
- Where the customer enters the brand experience.
- What reassurance they need when they arrive.
- Which actions might happen away from the website.
- How post-purchase support continues the relationship.
This is not the end of customer journey optimisation. It is a broader version of it.
The brands that make decisions easier will win.
Agentic commerce should not be treated as a reason to chase every new platform or add a chatbot to the website.
The immediate opportunity is more practical.
Brands need to make it easier for customers and agents to understand:
- What they sell.
- Who each product is for.
- Why it is different.
- What it costs.
- Whether it is available.
- How quickly it can be delivered.
- What happens if it is unsuitable.
- Why the brand can be trusted.
These have always been important ecommerce fundamentals.
The difference is that incomplete information may now prevent a brand from entering the customer’s consideration set at all.
The messy middle is not disappearing. Customers will still explore, compare, hesitate and look for reassurance.
They just may not do all of it themselves.
Ready to navigate the new messy middle?
Agentic commerce is changing the ecommerce landscape, but the fundamentals of a seamless, high-converting customer journey remain. At blubolt, we help ambitious brands eliminate friction, structure their data, and implement Journey-Led Growth so your products stand out to both AI agents and human shoppers. To see how our strategic, design, and technical teams can optimise your end-to-end digital journey, explore our services, or browse our work to see how we’ve driven growth and solved complex ecommerce challenges for leading brands. If you’re looking to upgrade your technical foundation, discover how we leverage Shopify Plus to build scalable, high-performance websites ready for the future of search. Don’t let your brand get left out of the AI conversation; contact us today to discuss how we can help!
