Sustainability Management: From ambition to impact
Are LLMs becoming the supermarkets of the internet?
How AI platforms are becoming a new distribution layer for brands: capturing demand, shaping preference and orchestrating transactions through app-in-app experiences.
From supermarkets to AI-mediated shelves
Supermarkets transformed retail by aggregating supply, simplifying choice and influencing the point of purchase. Later, marketplaces, app stores and super apps played a similar role online by becoming key intermediaries between users and brands.
Today, AI platforms may be extending this logic further.
With app-in-app experiences, third-party services can increasingly be discovered, compared and activated directly within AI interfaces, without users ever visiting a brand’s own environment. This is not just a new channel, but a new distribution layer: one that can capture intent, structure choice and may increasingly orchestrate transactions.
App-in-app refers to an AI-native interaction model in which users can access external services, inventory or actions directly within an AI platform. Several major brands have already moved in this direction. Sephora, TripAdvisor and Accor, for instance, have partnered with OpenAI to make parts of their offer accessible to ChatGPT users.
But this shift is no longer limited to ChatGPT: it is expanding across LLMs and conversational interfaces, as well as embedded assistants in vertical ecosystems. The app becomes portable, distributed and multi-interface.
The impact of AI on distribution is already visible. Salesforce reported that AI and agents influenced $262bn in global online holiday sales, representing 20% of all retail sales. Adobe analytics also found that traffic from generative AI sources to U.S. retail websites increased by 1,200% between July 2024 and February 2025. These numbers show that a conversational, intent-driven and increasingly actionable layer is emerging between customers and brands.
And while this phenomenon has often been framed through the lens of product e-commerce, the service dimension is equally structural. Insurance, banking, telecom, travel, healthcare, energy, hospitality: these are the categories where distribution remains the most frictional, opaque and poorly automated. Hidden quotes, long forms, non-standardized offers, mandatory sales calls. This is precisely where AI captures the most value: by turning complex journeys into comparable, actionable and orchestrated decisions.
Before looking at what this changes for brands, it is critical to understand what is changing in the user behavior itself. The shift is not primarily technological, but behavioral: users no longer search, they express intent; they no longer compare, they delegate; they no longer browse, they converse. This cognitive delegation is the real driver of the shift.
This also reflects a broader paradigm shift.
For three decades, digital marketing was an industry of placement: who pays the most, who ranks first on Google, who wins the slot on Amazon. With conversational AI, scarcity moves from placement to relevance. We're shifting from an industry of placement to an industry of intent, and very few brands have realized what that means yet for their marketing and commerce stack.
For three decades, digital marketing was structured as an industry of placement: top Google ranking, best Amazon slot, highest media bid. With conversational AI, scarcity no longer lies in placement but in relevance. The machine no longer selects a slot; it interprets an intention.
As this phenomenon scales, brands face a bigger question: what happens when AI becomes the interface where demand is captured and routed?
Three structural shifts are emerging:
The first shift is about customer access.
For years, brands have invested heavily in owning digital entry points: websites, apps, CRM programs, media activation and search visibility. But in an app-in-app model, the first interaction may no longer happen in a brand-controlled environment. It may happen inside ChatGPT, Claude or another AI assistant in the future.
The first interface often captures the most valuable signals: customer intent, context, urgency and preferences. If users ask an AI platform to find a hotel, compare beauty products or book a restaurant, the platform becomes the first gateway to demand. The brand may still deliver the service, but it no longer fully owns the starting point of the relationship.
This intermediation creates a structural data asymmetry. In an AI conversation, the platform captures the initial intent, context, objections and hesitation. The brand often only sees an API call, a conversion or a few attributes. Where traditional digital journeys left behind sessions, cookies and observable paths, AI introduces a form of structural blindness. Extracting usable signal from these interactions becomes a critical strategic challenge.
This creates a familiar risk of intermediation, but with a new level of intensity. Brands are not only competing for traffic anymore; they are competing for inclusion in AI-mediated journeys. The key question is no longer just, “How do customers find us?” but increasingly, “How do AI systems select us?”
That shift also raises strategic concerns. Who owns the customer data? Who captures the intent? How dependent does the brand become on a handful of AI platforms? As with marketplaces, the opportunity for reach comes with a risk of dependency.
The second shift concerns differentiation.
In traditional digital journeys, brands could influence customer choice through design, storytelling, navigation, media and branded experience. In AI environments, part of that work is delegated to the machine. The interface summarizes, filters, ranks and recommends on the user’s behalf.
In that sense, AI becomes a new kind of merchandiser.
But this does not mean branding disappears. It means branding is being redistributed. Part of the persuasive work may happen upstream through recommendation logic, but part of it can also be reclaimed downstream through the design of the app-in-app experience itself.
When a brand is activated within an AI platform, the question is no longer only whether it is visible. It is whether it can create a distinctive interaction once selected: a tone of voice, a service flow, a level of reassurance, a curated set of options or a signature way of helping customers decide. In other words, app-in-app can become a new surface for branded experience, not just a loss of it.
This matters because AI may standardize comparison, but brands can still differentiate through execution. A well-designed app-in-app experience can make the interaction feel more trusted, more premium, more convenient or more expert than a generic recommendation layer alone.
As a result, differentiation depends not only on brand equity or agentic readability, but also on a brand’s ability to design AI-native experiences that express its value within the constraints of the platform. The next shelf war may not happen only on websites, and the first battle may not be traffic. But the winners will not be the most visible brands alone; they will be the ones that know how to turn selection into branded interaction.
The third shift is more structural. AI platforms do not operate like traditional websites. They rely on structured data, APIs, plugins, feeds and standardized service layers to connect users with third-party offers.
But beyond the technical layer, something more important is happening, AI agents are starting to act as orchestrators. They do not simply display options. They can interpret intent, assemble relevant services, compare providers, trigger actions and potentially manage parts of the customer journey across multiple brands.
In other words, commerce is not only becoming more protocolized, it is becoming more agent-orchestrated.
For brands, this means visibility is no longer enough. Offers must be machine-readable, comparable and executable. Product attributes need to be structured. Availability must be updated in real time. Booking, payment or service activation flows must be accessible through interoperable systems.
The strategic implication is significant. If orchestrator agents become the layer that routes demand and sequences actions, they may increasingly define which brands are included, how they are evaluated and where value is captured. The power no longer sits only with the interface that attracts traffic, but with the intelligence layer that decides what happens next.
This opens the door to new monetization models. As in marketplaces or app stores, brands may face commissions, revenue-sharing mechanisms, paid placement or sponsored recommendations inside AI environments. The familiar trade-off between reach and margin will likely intensify.
The deeper issue, however, is governance. Brands may find themselves participating in journeys they do not fully control, according to protocols and decision rules they do not define. The battle is no longer only about owning a digital storefront; it is also about being compatible with the systems that orchestrate AI-native commerce.
What brands stand to gain
Despite these risks, app-in-app represents a major growth opportunity, and likely one of the few windows of the decade where competitive positions can truly be reshuffled, thereby opening significant prospects:
First, it offers access to new audiences through AI-native distribution including ones that have historically been hard to reach. Conversational AI is drawing in younger users who have already moved away from traditional search toward TikTok, Instagram and now LLMs. In many categories, ChatGPT and Perplexity have already become primary entry points for information.
More importantly, this is not just an audience channel: it is a full sales channel, closing the entire journey end to end: discovery, conversion, payment, after-sales. Apps sit closer to a marketplace than to a billboard. They reduce friction in journeys where convenience and speed matter, and can improve conversion by shortening the path from intent to action. The model is especially valuable in categories where users want rapid support rather than prolonged browsing: travel planning, local services, beauty, hospitality, ticketing or second-hand marketplaces, for instance.
It also delivers on a long-disappointed promise: personalization. Arguably the most over-promised and under-delivered topic of the past decade, personalization has fallen short for lack of a sufficiently rich intent signal. Conversational AI captures explicit, contextualized and nuanced intents that no click stream has ever surfaced. For the first time, brands can genuinely tune their offer, tone, price and service to a high-quality signal, provided they are present on these interfaces and can retrieve the data.
App-in-app also creates room for new use cases. AI can help customers refine needs, compare options, narrow down choices and move directly into action. For brands, this can translate into better qualified traffic, more seamless journeys and a stronger presence in moments of high intent.
Getting ready for the next distribution shift
AI platforms are becoming demand orchestrators. For brands, the challenge is therefore not simply to be present, but to be selected, interpreted and activated inside these new environments. That requires rethinking customer access, technical architecture, monetization choices and differentiation strategies.
The landscape described here points to a reshuffling of competitive positions. Platform shifts are historically the rare moments when incumbents can be challenged and new leaders can emerge. Brands that move early can build structural advantages that will later be costly to displace. Waiting for the market to mature means accepting the risk of arriving too late.
At the same time, this market remains early. Standards are still forming. User behaviors are still evolving. Commercial models are still unstable. That is precisely why brands should avoid waiting for a mature playbook to emerge. As Lemrock puts it:
“No one has the playbook for AI distribution yet. Standards are still forming, user behaviors are still evolving. The right response is to start now: test partnerships, identify the journeys where app-in-app makes sense, and learn from what works. The brands that move first will be the ones writing the playbook everyone else will follow.”
For brands, the implication is clear: the right response today is test-and-learn.
Brands should start experimenting now: testing AI-native partnerships, identifying which journeys are most relevant for app-in-app, auditing whether their offers are machine-readable and exploring what makes them more recommendable in AI-mediated environments. The point is not to have all the answers already. It is to build the capability to learn faster than the market shifts.
Because if supermarkets reshaped retail by controlling the shelf, AI platforms may reshape digital commerce by controlling both access to demand and the orchestration of action. In that world, the brands that win will not only be agent-ready. They will be the ones that know how to turn AI selection into branded experience.
This article was co-written with Lemrock.
Curious to know more? Contact us:
-
Charles Denis
Managing Partner, Sia Experience | Paris
Charles is Managing Partner of Sia Experience. He has been working for 15 years on digital transformation projects, particularly for energy companies.
