Shelf Life | Vol. 57 - The Client Book: The Receipt Never Got the Full Story
Shelf Life | Vol. 57 - The Client Book: The Receipt Never Got the Full Story
🗓️ August 2026 | ✍️ Jackie Swanson
Conversational commerce is turning the messy, useful backstory behind a purchase into data that marketing, merchandising, product, planning and supply chain can actually use.
The Client Book
For nearly 50 years, one of the best customer intelligence systems in American retail worked on the third floor of Bergdorf Goodman.
Her name was Betty Halbreich.
Betty ran Bergdorf’s personal shopping service, Solutions, and became famous for knowing her clients well enough to tell them what they needed to hear, occasionally whether they wanted to hear it or not. Her value went well beyond knowing the inventory. She knew the occasion, the taste, the insecurities, what had already been worn to whose party and which client needed to be physically separated from another sequined jacket.
The register downstairs captured what the client bought. Betty carried around the rest of the story.
Digital retail has spent the last two decades getting extraordinarily good at the register part. Searches, clicks, filters, page views, carts, abandons, purchases, returns, loyalty activity. An impressive trail of evidence exists for what the customer did.
The backstory has been harder.
Consider a shopper who says: “I need a dress for an outdoor wedding in September. Formal, but nothing that looks bridesmaid. I need to dance. I hate strapless. And under $500 unless I love it.”
There is a lot going on in there. Occasion, timing, function, style boundaries, product rejection, price flexibility and the universally recognized retail budget of “$500 unless I love it.”
That is where conversational commerce gets interesting.
Agentic commerce has captured plenty of attention because AI is beginning to research, compare, select and increasingly transact for shoppers. That transformation is significant. Sitting right beside it is another opportunity with fewer robots-buying-toothpaste headlines and potentially enormous strategic value.
Customers are starting to explain themselves.
At scale.
In complete sentences.
For retailers, that could become a digital client book.
Top Shelf Insights
💬 Conversational commerce creates a richer customer signal. Search captures the query. Clickstream captures behavior. Conversation can surface the occasion, constraint, urgency, hesitation, tradeoff, acceptable substitute and reason behind the purchase.
🛍️ Conversational commerce spans more than AI chat. Salesforce includes messaging apps, chatbots and voice assistants in the category, while conversational AI is increasingly turning those channels into more capable shopping experiences.
🤖 Agentic commerce extends further into action. AI agents can shop on a customer’s behalf, researching choices and increasingly moving through the transaction. Conversational and agentic capabilities can coexist inside the same journey without meaning the same thing.
🧠 The bigger enterprise story may be the data created along the way. Marketing gets language and need states. Merchandising gets unmet demand. Product gets requests and frustrations. Planning and supply chain get timing and substitution signals. CX gets an extremely detailed list of everything customers apparently could not find on the website.
🔐 Who holds the client book matters. As more shopping conversations happen inside third party AI environments, retailers need to know whether they receive the customer context or simply the resulting order.
The Appointment
Conversational commerce has been around longer than generative AI. Messaging, chatbots and voice assistants have been part of the category for years.
Many early retail chatbots also provided a memorable lesson in managing expectations. A shopper could type a perfectly reasonable question and receive three FAQ links, a cheerful emoji and the longing for real human interaction.
Generative AI changes what the interaction can handle.
A shopper can describe a complicated mission in natural language, add another constraint halfway through, reject the first suggestion, ask why the second is better, change the budget and continue the conversation without learning the retailer’s internal taxonomy first.
Salesforce describes conversational AI for commerce as a personal-shopping-assistant experience that can deliver tailored recommendations, inventory information and interactive buying guidance.
Agentic commerce adds another capability: action. An AI agent can increasingly research products, evaluate alternatives and shop on the customer’s behalf.
Sephora shows where the two are beginning to meet. Its 2026 app in ChatGPT lets U.S. customers discover and shop for beauty products through natural-language conversation. With permission, Beauty Insider information can be connected to personalize the experience, and Sephora has said payment and checkout capabilities are planned as the experience evolves.
The consultation is conversational. The ability to carry that consultation into action makes the journey increasingly agentic.
Most of the industry attention has gone to the action. Can the agent find the right product? Compare it? Check availability? Pay? Apply loyalty? Reorder it?
Important questions.
The consultation is generating another asset at the same time.
The Client Notes
A search for “black wedding guest dress” provides useful intent.
Now compare it with: “I need a black dress for a wedding in Miami next month. The ceremony is outside. I want sleeves, nothing bodycon, and I’d like to stay under $500 unless it’s amazing. I also need shoes.”
The second shopper has practically filled out her own client card.
The retailer now knows the occasion, location, timing, preferred attributes, rejected silhouette, approximate price ceiling, willingness to stretch and an adjacent category need. Another question or two could uncover fit considerations, favorite brands, delivery requirements and exactly how much financial damage the word “amazing” is permitted to cause.
Retailers have always wanted this context. Pieces of it live today in surveys, reviews, service transcripts, store-associate knowledge, panels, social listening, loyalty data and behavioral analytics.
Conversation moves some of that information into the actual decision. That distinction matters.
A category tree is designed around how a retailer sells. A conversation is much closer to how a customer lives. Someone rarely wakes up yearning to “shop Women > Apparel > Occasion > Midi.” She has a wedding in Miami, hates her arms today, forgot about shoes and has four weeks to solve all of it. The client notes contain that context.
The Edit
Marketing gets the words. For years, marketing teams have worked to understand how customers naturally describe their needs. Search terms, social listening, surveys, reviews and focus groups all provide useful pieces.
Conversational commerce adds language volunteered during an active decision.
“I need skincare for my teenager because she suddenly wants a ten-step routine and preserving her skin barrier would be nice.”
“I need luggage that looks polished enough for a client trip and durable enough for whatever the airline has planned for it.”
“I forgot the hostess gift and need something that suggests this was absolutely planned in advance.”
Those are more useful briefs than “teen skincare,” “carry-on luggage” and “hostess gift.”
At scale, conversational data can expose the jobs customers are trying to accomplish, the benefits that matter in context, the objections slowing down the purchase and the phrases people actually use to describe the problem.
That has implications for creative, segmentation, product content, FAQs, SEO and AEO. Google is already adapting commerce tools to more conversational discovery, including richer Merchant Center attributes and AI performance insights for AI shopping surfaces.
It also provides a brutally efficient content audit.
If customers ask 6,000 times whether the white sofa can survive toddlers, dogs and Cabernet, the chatbot has learned something useful. The product detail page should probably be invited to the hindsight debrief.
Need-state segmentation could get more useful too. “Female, 35 to 44” may describe the customer. “Forgot the birthday, needs delivery tomorrow, willing to pay for convenience and still wants credit for being thoughtful” describes the mission.
Yes, Marketing has spent years personalizing around who the customer is. But conversation, adds why she showed up today.
The Pull
Merchandising gets the wish list that never made it to the register. Merchants have always wanted to know what customers wanted and could not find. Today those signals are scattered across zero-result searches, abandoned journeys, returns, reviews, service transcripts, store feedback and the associate who has been saying for six months that everyone keeps asking for the same thing. Conversational commerce can make those gaps harder to ignore. “I love this dress. Does it come with sleeves?” “Anything like this in navy?” “I want this exact shoe with a lower heel.” “Do these pants come in petite?” “Is there a version in cashmere?” “Show me something similar that can actually get here before Friday.”
A sales report captures converted demand.
The client book can begin to expose the demand that nearly converted, then died quietly somewhere between “perfect” and “doesn’t come in my size.”
That can inform assortment in ways the transaction alone cannot. Which attributes repeatedly appear together? Which sizes and fits are missing? Which substitutions work? Where will customers trade color for delivery speed? Which features support a higher willingness to pay? Which products naturally belong together when the shopper explains an occasion rather than navigates a department?
A customer who wanted navy and left empty-handed does not appear in unit sales. Neither does the woman who needed petite, the parent who required delivery before visiting day or the shopper who would have purchased the outfit if the right shoe had surfaced.
The absence of a transaction is easy to read as the absence of demand.
Merchants know better.
Conversational commerce could give them receipts for the misses too.
And if 8,000 customers ask for sleeves, sleeves have officially moved beyond anecdotal feedback.
The Fitting
Product and CX teams get everything that almost worked.
Anyone who has ever stood outside a fitting room knows how much useful information follows the words “I like it, except…”
Except it wrinkles. Except it has no pockets. Except the fabric feels cheap. Except it cannot be washed. Except the instructions make no sense. Except somebody decided women apparently no longer require functional pockets.
Conversational commerce can capture those edges around the product at scale.
Product teams can look for recurring feature requests, missing configurations, confusing specifications and compromises customers keep making. One shopper asking whether something comes fragrance-free is feedback. Thousands asking the same question begin to resemble a product-development brief.
The same logic applies to digital experience.
Which questions keep appearing because customers cannot find the answer? Which comparisons are confusing? Which policies require a translator? Where does confidence disappear? What finally resolves the hesitation?
Sometimes the assistant needs a better answer. Sometimes the site needs better content. Sometimes the product needs fixing.
That distinction is useful before the next cross-functional meeting concludes that the solution is another homepage redesign.
Back of House
Planning and supply chain get the customer’s preferred forecasting methodology:
“I need it Thursday.”
Customers talk about operational constraints constantly. They simply have the good manners not to call them operational constraints.
“I’ll take the other color if it arrives tomorrow.”
“Can I pick this up near my office?”
“I need twelve.”
“I need all three pieces to arrive together.”
“I’ll pay more if it comes assembled.”
“Only show me what is actually in stock.”
Those statements contain information about urgency, availability, substitution tolerance, local inventory, service expectations and fulfillment flexibility.
Aggregated carefully, conversational signals could help identify where delivery speed changes the purchase decision, which substitutions preserve demand during stockouts, where local pickup matters and which occasions create hard deadlines.
They can also show where a customer will trade one attribute for another. The preferred color may suddenly become less preferred when one arrives Wednesday and the other arrives after the birthday party.
This data should remain a signal rather than becoming the forecast by executive enthusiasm alone. People say things they never buy, conversational users may not represent the total customer base and expressed intent needs to be connected with actual transactions, inventory, returns and other demand indicators.
A chatbot should not inherit demand planning simply because it learned to use punctuation.
Still, planning already combines imperfect signals into a view of future demand.
“I wanted it, but it could not arrive before camp visiting day” is a useful signal that abandoned-cart data cannot explain on its own.
Who Holds the Client Book?
Betty’s knowledge benefited Bergdorf Goodman because Betty worked at Bergdorf Goodman.
The modern client book may live somewhere else.
A shopper can tell an AI assistant that she is furnishing her first apartment, has a dog, hates beige, needs the room finished before Labor Day, prefers a deep sofa and is willing to spend more there if the dining table stays reasonable.
That is a useful amount of customer intelligence.
The retailer may eventually receive SKU 48217, quantity one.
Google is already embedding more conversational and agentic commerce capabilities across Search, AI Mode and Gemini, with tools such as Business Agent, Merchant Center AI insights and Universal Commerce Protocol infrastructure. Shopify is similarly connecting merchant catalogs and transactions into AI shopping environments.
That makes data-sharing terms strategically important.
Which conversational signals come back to the retailer? At what level of aggregation? With what customer permission? Can the information connect to an existing customer relationship? Can aggregated signals inform merchandising and product development? How quickly are they available? What does the platform retain?
These questions belong beside conversion rates, fees and checkout ownership.
A retailer can fulfill the order perfectly while somebody else develops the deeper understanding of why the customer placed it.
There is a certain indignity in becoming the world’s most efficient stockroom for somebody else’s client book.
The Invisible Appointment
The measurement problem compounds the issue because conversational influence can happen long before a shopper arrives on a retailer’s site.
Adobe reported that AI-referred visitors during Prime Day 2026 converted 40 percent better than visitors from non-AI channels once they reached retail sites.
Then there is the influence that never arrives with an AI referrer attached.
A 2026 study linking opt-in AI conversations with subsequent browsing found that when an assistant recommended a brand to someone with no recent observed engagement, same-name Google searches rose by 4.3 percentage points and visits to the brand’s own site rose by 2.4 percentage points. The authors did not observe transactions, so the result is purchase-adjacent rather than proof of sales impact, but the attribution problem is clear: standard referral and last-click models miss the upstream AI exposure.
The assistant makes the introduction.
The attribution dashboard sends Google flowers.
Conversational commerce therefore needs a broader measurement model. Conversion still matters. So do needs clarified, recommendations accepted, objections resolved, brands introduced, journeys resumed and purchases completed elsewhere.
The client book contains influence that may never appear on the receipt.
Handle With Care
A digital client book comes with a few complications Betty never had to discuss with the privacy office.
Customers can reveal personal information quickly when explaining what they need. Health considerations, family circumstances, body concerns, financial limits and other sensitive context can easily enter a shopping conversation because, from the customer’s perspective, it is simply relevant to getting a better answer.
Retailers need clear rules around what can be remembered, what becomes a persistent attribute, what stays within one conversation, how long information is retained and what customers can review, correct or delete.
Commercial steering also deserves attention.
A 2026 preregistered study with 2,012 participants found that conversational AI designed to promote sponsored products drove sponsored selections at a much higher rate than traditional search placement. Most participants did not detect the promotional steering, and explicit sponsorship labeling did not significantly eliminate the effect in the experiment.
A recommendation has a different feel after a shopper has spent several minutes explaining what she needs. The system knows the mission, the objections and the compromises she is willing to make.
That makes for an excellent salesperson.
It also makes “we needed to move some inventory” a considerably less charming recommendation strategy.
The rules around memory, consent, sensitive data, commercial incentives, disclosure and human escalation should be decided while the client book is still manageable.
Deleting a few billion extremely personal Post-it notes later sounds less fun.
On the House
Here’s my take.
Agentic commerce remains one of the biggest transformations facing retail. Product data, customer identity, payments, inventory, loyalty, architecture, governance and operating models all need to evolve as AI takes on more of the shopping journey.
The client book deserves a place on that agenda.
For twenty years, digital retail has tried to infer intent from breadcrumbs. A shopper searched, filtered, clicked, abandoned, returned, purchased and possibly sent the whole thing back. Then teams tried to reconstruct the movie from seven screenshots.
Conversational commerce gives the customer an opportunity to narrate some of the missing scenes herself.
That is what feels genuinely new.
The flashy story is AI becoming the shopper. The strategic story may be AI becoming the personal shopper, learning something useful during every interaction and creating intelligence that can improve the next one.
Betty did that client by client.
Now there is a path to doing it across millions.
The transcript is not the asset I would obsess over. The learning is.
Made to Measure
Turning conversation into enterprise intelligence requires more than saving a lot of chat logs and declaring victory.
📝 Start with the client notes. Define a common taxonomy for signals worth understanding, including mission, occasion, urgency, preference, constraint, rejection, tradeoff, substitution, willingness to pay, unmet need and outcome. Transcripts are raw material. The structured signal is what teams can actually use.
🔗 Connect words to behavior. Where appropriate and consented, analyze conversational signals alongside search, product, transaction, inventory, fulfillment and return data. The useful question is what customers said, followed by what actually happened.
🛍️ Put the misses into merchandising. Create recurring views of unmet demand, rejected attributes, failed substitutions, fit gaps and repeated product requests. The line review should include some evidence from customers who tried to buy and could not, rather than exclusively celebrating the customers who found something on the rack.
📣 Give marketing the language. Feed recurring need states, customer phrasing, objections and unanswered questions into content, creative, segmentation, SEO and AEO. Internal taxonomy has many strengths. Sounding like an actual human shopper is rarely one of them.
📦 Give planning and supply chain the clock. Identify where availability, delivery timing, pickup and substitution meaningfully change customer intent. Treat the information as an additional demand signal and validate it against actual behavior.
🧵 Bring product and CX into the fitting room. Repeated feature requests, points of confusion and “almost right” feedback should have a clear path into product development and experience improvement. Ten thousand requests for pockets should not die in the chatbot archive.
🔐 Decide what stays out of the book. Establish clear governance for consent, memory, retention, sensitive information, inference, commercial steering and customer control before conversational data becomes deeply embedded in personalization and decision making.
The organizational implication may be the most important one. Conversational commerce cannot live as an ecommerce feature or customer-service project. Marketing, merchandising, product, planning, supply chain, CX, technology, data, privacy and legal all have something to do with what the customer is now saying.
The Last Look
Purchase data remains incredibly valuable because someone actually spent the money.
The client book explains everything around that moment: what the customer wanted, what she could not find, what almost stopped her, what she would have accepted instead, what she valued enough to pay more for and what ultimately changed the decision.
So this week’s debate is simple:
If the receipt tells retailers what customers bought and the client book explains why, which becomes the more strategically valuable asset as AI reshapes the shopping journey?
Betty spent nearly 50 years proving the value of knowing both.
More to come in the Shelf Life series.
Related editions
Browse the full Shelf Life Editions archive · Vol. 54 — The Recoupling · Vol. 51 — Make It Work
New editions publish weekly. Subscribe or book time with Jackie.
Jackie Swanson is a Managing Partner at Gartner Consulting, where she advises retailers and consumer brands on AI-readiness, agentic commerce strategy, and large-scale transformation. She lives in New York with her husband and three children, which is either great preparation for managing complex client engagements, or the other way around.
📩 Ready to talk about what this means for your organization?
Book a 1:1 with Jackie → jackie.swanson@gartner.com
Follow Shelf Life on LinkedIn | Substack | Instagram @ShelfLifebyJKS.
#ShelfLife #ConversationalCommerce #AgenticCommerce #RetailTrends #ConsumerGoods #AIinRetail #CustomerExperience #Merchandising #SupplyChainStrategy #RetailData #GartnerConsulting
