AI Shopping Assistants: Is Your Customer Journey Ready?

Got a relative who’s hard to buy gifts for? The anxiety’s real when their birthday’s in a week, you have a rough budget and no real idea what they’d like, so you lose an evening to searching, clicking through product pages and reading reviews, hoping something jumps out. 

An AI shopping assistant gives you another way to start: describe the person, the budget and the problem, then ask for ideas. And… breathe.

AI now has a role in the customer journey. SAP’s Customer Loyalty Index 2026 surveyed 10,002 consumers in the US, UK, Germany and Hong Kong. Among them, 38% use AI to help make purchase decisions, and 33% have purchased AI-recommended products.

For marketers, that creates two priorities: helping shoppers through the AI assistant on your own site or app, and making useful product information available to the AI assistants customers use elsewhere.

Both depend on accurate information. An assistant needs to know more than whether a product sounds suitable. It needs the right price, availability and delivery details. If it acts for a customer, that customer also needs a way to check the action, change it where possible and get help.

An AI shopping assistant is software that uses AI to understand what a shopper wants, in the shopper’s own words, and help them find, compare and buy products. Some answer questions and make recommendations. Others can act for the shopper, such as adding items to a cart or placing an order, depending on their capabilities and permissions.

For brands, it’s useful to distinguish between two kinds:

• The brand’s own assistant. It sits on your website, in your app or in a messaging channel. What it knows depends on the catalog, customer records and business systems you connect to it. We’ve explored shopping assistance among the AI use cases in e-commerce.
• The assistant the customer brings. General-purpose assistants such as ChatGPT, Gemini and Perplexity can help shoppers research across brands. Your products may be compared with competitors’ products outside your own website.

This distinction changes what you, as a brand team, can influence. With your own assistant, you can decide which information it can access and which actions it can take. For an outside AI assistant, focus on the product information you publish or make available. Keep it accurate, current and easy to cite.

Take a request like “trail running shoes for wide feet, under $150, that’ll arrive by Friday.”

A useful response has to account for all three constraints. A shoe that fits and costs $120 still fails the request if it arrives next week. When designing or evaluating an assistant, look for a process that can:

• Interpret the request. Identify the fit, budget and deadline, and ask a follow-up question when something is unclear.
• Check product and operational information. Use attributes, prices, stock, delivery options, reviews and policies to assess the options.
• Use relevant customer context. Where the customer has agreed to share it, consider past purchases and preferences.
• Explain the shortlist. Show why each suggestion fits and respond to follow-ups such as “which of these is lighter?”
• Confirm any action. If the assistant can add to cart or place an order, make clear what the shopper is authorizing.

The customer’s Friday deadline shows why a good conversation sometimes isn’t enough. The best possible recommendation needs current stock and delivery information to establish whether the shoes can arrive in time.

The stages themselves haven’t changed. A shopper still works out what they need, weighs up the options, buys, and then lives with the purchase: the delivery, the questions, sometimes the return. Our digital customer journey roadmap covers those stages in depth.

What changes is who, or what, the shopper turns to along the way. Where they once had a search bar, a friend’s advice and a dozen open tabs, many now have an AI assistant too, and it helps in a different way at each stage. SAP’s research shows where shoppers already ask AI for help and what they want from it.

1. Discovery: “I have no idea what to buy”

Sometimes you don’t have a clue what to buy someone. In our Customer Loyalty Index 2026, that was the reason respondents gave most often for using an AI assistant. Their top reasons were:

• Unsure what to buy (23%)
• Too many options (22%)
• Short on time (20%)
• Buying gifts (17%)
• New to the brand or platform (16%)

Read that list again and it’s mostly a list of small frustrations: too much choice, too little time, a gift that has to be right.

A search bar is little help in those moments, because it expects you to know what you’re looking for. An assistant lets the shopper start a step earlier, with the problem instead of the product: “my dad’s turning 70, he loves gardening and he already has everything.”

For brands, these findings suggest that product content should explain what a product is for and who it suits, alongside its name and specifications. If a shopper asks for “something for a beginner who’s short on space,” information about experience level and space requirements gives the assistant a basis for judging whether your product fits.

2. Consideration: a second opinion shoppers trust

Once there’s a shortlist, the hard part starts: which one? Again, that used to mean phoning a friend, reading reviews or flicking between tabs until you gave up and picked one.

Now, many shoppers ask AI. When respondents named the value they get from AI assistants, fast advice (25%) and help finding the right products (24%) came top.

Some trust that advice a great deal. Among respondents, a surprising 35% prefer AI recommendations over friends and family, and 31% regularly rely on AI to choose what to buy.

SAP’s report explains how easier comparison can affect loyalty: “As AI makes comparisons, recommendations, and product discovery easier, customers no longer need to stay loyal simply to access the best deal.”

That gives brands a practical reason to make comparisons straightforward. Make sure you publish clear specifications, current prices, delivery times and returns terms. Missing details leave questions about whether your product meets the shopper’s needs, and the AI will be all over it.

Your brand’s own assistant matters here too. Among respondents, 30% are more likely to shop with brands that use AI assistants.

3. Purchase: handing over the checkout, not the controls

In the Customer Loyalty Index 2026, 33% of respondents say they trust AI to complete purchases. That’s a bold step. Letting software spend your money, at your address, on something you’ll have to live with, takes trust. So for any assistant that can place an order, the question is how much the shopper wants to hand over, and how easily they can take it back.

The survey also asked how much control people want over AI in general, rather than purchases specifically:

• Full human control: 20%
• Shared control: 60%
• AI-led with oversight: 16%
• Fully autonomous: 4%

Separately, 56% want to reverse AI decisions and 48% want human intervention to be available.

Trusting AI to complete a purchase doesn’t establish that someone wants it to act without review. The broader control findings suggest that brands should let shoppers:

• Review the items, total price and delivery details before an order is placed
• Change the proposed purchase and understand their options for canceling or returning it
• Reach a person when something goes wrong

These are design recommendations drawn from attitudes toward AI generally, rather than findings about checkout preferences.

SAP’s report puts it this way: “Loyalty lives in the shared-control zone.”

We explore these expectations further in our guide to AI personalization and customer trust.

4. After the purchase: an assistant that remembers you

In the Customer Loyalty Index 2026, 42% of customers would prefer an AI assistant that remembers them from previous interactions. Asked what AI assistants should remember, respondents said:

• Past purchases and order history (32%)
• Products regularly purchased (29%)
• Loyalty points and rewards (29%)
• Delivery preferences (27%)
• Customer service history (27%)

Those preferences point to practical uses after checkout: helping a customer find a previous order, reorder a regular product or pick up a service conversation without explaining the problem again.

Each of those depends on more than a silicon-grade memory. The AI assistant also needs the current records behind it. Remembering that someone placed an order won’t answer “where is it?” without current delivery information.

Similarly, remembering that they belong to a loyalty program won’t establish which rewards they can use without their current balance and eligibility.

Customers also expect these parts of the experience to work. In return for their loyalty, 30% expect priority delivery and free returns, and 20% expect real-time order and fulfillment updates.

Product information now needs to help both shoppers and the AI they consult. In the Customer Loyalty Index 2026, 15% of respondents are more loyal to brands because they’ve been suggested by AI chat responses, such as ChatGPT or Gemini.

Using AI doesn’t necessarily mean a shopper feels less loyal. Among all respondents, 69% consider themselves loyal to brands. Among AI shoppers, defined as those using AI to help make and inform purchase decisions, that figure rises to 78%. This comparison doesn’t show that AI causes loyalty, but it does show that AI use and reported brand loyalty can coexist.

Plan for both your own assistant and the general assistants customers may consult. The supplied research doesn’t establish which kind shoppers rely on more. Changes in how brands get found are also among the AI marketing trends marketers need to consider.

Start with the information you publish, the data you connect to your own AI assistant and the experience customers get when they act on a recommendation.

Product and operational data

Review the information an assistant would need to assess your products:

• Product attributes: Size and fit, materials, compatibility, care instructions, and who the product is for
• Pricing: Current prices and promotions, accurate for the channel where the customer will buy
• Availability: Current stock, by channel and location where relevant
• Delivery: Options, costs and realistic timings
• Returns and warranty: Clear, consistent terms

Picture a shopper asking whether a jacket will keep them dry on a wet walk to work. If your product information gives its waterproof rating and says it’s built for daily commuting, the AI has the receipts. Now the shopper can get to the questions that make it their jacket: ‘Do I like it? Does it go with what I wear? Is it worth the price?’

SAP’s Global Engagement Index Report 2026 advises enterprises: “Create a comprehensive view of your product data to drive consistent omnichannel experiences. Use commerce AI agents to organize and enhance product catalogs so customers can easily discover, research, and buy while commerce teams are freed from routine tasks.”

However, access to that data remains a challenge for many businesses. Of the 4,800 enterprise senior decision-makers surveyed, 54% can’t access and use real-time data. Asked what’s holding their AI efforts back, 32% point to data quality and availability, as well as integration with existing systems.

For the shopper looking for trail running shoes, those gaps have a direct consequence: an assistant may lack the information needed to assess the shopper’s urgent Friday deadline. Bringing product, order and inventory data together helps address the information gap that a conversational interface alone cannot solve.

Customer data

Customer records provide the context for the memory respondents want: past orders, regular purchases, loyalty points, delivery preferences and service history.

That use of data needs to earn trust. Respondents named data privacy and security assurances (35%) among the things that build trust in AI, and 23% are more loyal to brands they trust with their data.

Ask clearly for consent and explain what remembering information will do for the customer. “Use my order history to help me reorder” gives someone a concrete benefit to assess. Connect your approach to customer data and AI-driven personalization to those useful, understandable purposes.

Respondents in the Customer Loyalty Index 2026 named these reasons for breaking loyalty:

• Declining product quality (45%)
• Rising prices (40%)
• Poor customer service (36%)
• Irresponsible data use (30%)
• Misleading advertising (29%)
• Limited access to human support (23%)

These expectations apply when AI helps with the purchase too. A recommendation may bring a customer to your brand, but your business remains responsible for the product and the service that follow.

Check that the promises made during shopping match what happens afterward:

• The item shown as available is in stock
• The quoted delivery date is met
• The checkout price matches the price presented to the shopper
• Returns work as the policy describes
• A person is available when the customer needs help

That last one matters more than it might seem. When an order goes wrong, the moment a customer most needs reassurance is the moment a chat window can feel least helpful. Respondents agree: when they named what builds trust in AI, human support if something goes wrong came top (42%), ahead of the ability to review or override decisions (36%).

• Plan for both kinds of AI assistant. Consider the experience on your own site and the information available to assistants customers use to compare brands.
• Make product information useful for a decision. Keep attributes, prices, availability, delivery and returns accurate. Describe the needs a product meets as well as its specifications.
• Connect operational data to engagement. Use orders, inventory, deliveries and service records to inform what your assistant says and what your messages promise.
• Choose what to remember, with consent. Start with the records respondents identified, including past purchases, delivery preferences and the loyalty experience.
• Make control visible. Show customers what an assistant proposes to do, how they can change it and where to get human help.
• Check the experience after the recommendation. Product quality, pricing, service and responsible data use remain reasons customers give for staying or leaving.

Shopping assistance is also one part of a wider AI marketing strategy. Keep the customer’s experience connected to the engagement work your marketing team already manages.

The shared-control findings suggest preparing for journeys in which shoppers choose how much to delegate. They don’t establish where each shopper will draw that line.

Think about the dog food or toilet paper you buy every month: same brand, same packet, the same moment of realizing you’re nearly out. Few people would mind an AI assistant reordering it within limits they’ve set. It’s one less thing to remember.

Now think about an engagement ring, a wedding outfit or a gift for a parent’s 70th birthday. Those purchases carry nerves, hope and a lot of meaning. Getting them right matters, and so does the feeling of having chosen them yourself. The same shopper might welcome help researching the options and still want the final choice to be theirs alone.

These are illustrations, not survey findings about those products. But they show why it makes sense to offer control at each decision rather than assume one level of automation suits every purchase.

For the related question of how brands can use agents in their own marketing work, see our guide to agentic marketing.

Preparing for AI-assisted journeys means connecting customer context with what’s happening in the business.

SAP Engagement Cloud describes that connection: “Intelligent automated workflows activate based on real-time business events, such as orders, inventory, service tickets, and loyalty milestones.” Its personalization reaches customers “whether through conversational experiences, AI-driven dialogues, or channels like email, SMS, mobile, and web.”

That gives engagement workflows business context: an order, a change in inventory or a service ticket can trigger a response. It’s relevant to the same requirement seen throughout the shopping journey – what you say to a customer needs to reflect what your business can deliver.