Meta’s next gamble: shopping agents as the new surveillance engine

Meta is quietly preparing its biggest product shift since the News Feed. Not a new app, not another VR headset, but an invisible layer of AI agents that will sit between billions of users and almost every commercial interaction they have on Meta’s platforms. If Mark Zuckerberg is right, your next shopping assistant will not live on Amazon or Google, but inside Instagram, WhatsApp and Facebook – fuelled by the most intimate profile of you that has ever existed online. In this piece, we look at what Meta is really building, why commerce is the wedge, and why regulators should be paying close attention.

The news in brief

According to TechCrunch, Mark Zuckerberg told investors on a January 2026 earnings call that Meta will begin rolling out new AI models and products in the coming months, following a full rebuild of its AI program in 2025. While he did not commit to specific launch dates, he singled out AI driven commerce as a priority.

Zuckerberg said Meta is working on so called agentic shopping tools designed to help users find an optimal set of products from businesses in Meta’s commerce catalog. He argued that Meta’s access to rich personal context – history, interests, content and relationships – would allow it to deliver unusually personal AI agents.

The call followed Meta’s December acquisition of Manus, a developer of general purpose AI agents, which Meta plans to continue offering while also integrating into its own products. In its latest filing, Meta projected capital expenditures of 115 to 135 billion dollars in 2026, up from 72 billion in 2025, largely to support its Meta Superintelligence Labs and core business.

Why this matters

Meta is not just adding a new feature; it is trying to move up the value chain from advertising to transaction orchestration. If AI shopping agents live inside WhatsApp chats, Instagram messages or Facebook groups, Meta can do more than show ads. It can guide the decision, broker the purchase and potentially control payment rails or at least the flow of referrals.

The winners, at least in the short term, are clear. Meta gains a fresh monetisation surface at a time when classic targeted advertising faces regulatory and platform headwinds, particularly from Apple’s privacy changes. Merchants that bet on Meta Shops and WhatsApp Business could suddenly gain a powerful digital sales assistant that understands the customer far better than any CRM they can afford.

The losers are less obvious but more worrying. Independent ecommerce sites risk becoming invisible layers behind Meta’s conversational interface, reduced to low margin fulfilment nodes. Competing discovery channels, from Google Search to Amazon’s search bar, may see high intent product queries siphoned away into closed agent ecosystems. And users will pay in the most valuable currency of all: even deeper behavioural and relational surveillance, wrapped in the promise of convenience.

Crucially, agentic commerce gives Meta a new form of lock in. If your AI assistant knows your style, budget, recurring needs and social context, switching away from Meta is far more painful than closing a Facebook tab. This is not just about selling more shoes; it is about owning the context graph that future AI systems will depend on.

The bigger picture

Zuckerberg’s bet slots neatly into a broader industry race around AI agents that can take actions, not just answer questions. Google is pushing its own transaction capable agents, integrated into search and Android. OpenAI is experimenting with agent enabled plugins and partnerships with platforms such as Stripe and Uber, effectively turning ChatGPT into a meta interface for services.

What Meta brings to this fight is distribution and intimacy. Billions of people already spend hours a day in its apps. Those apps are not primarily search or productivity tools; they are social environments, full of implicit trust signals, peer pressure and aspiration. Inject a shopping agent into that loop and you end up with something more persuasive than a standard recommendation engine.

There is a historical echo here. Facebook once used its social graph to reinvent online advertising, moving from crude demographic targeting to finely grained interest based campaigns. Reels and Stories were Meta’s response to losing feed attention to Snapchat and TikTok. Now, agentic commerce is the defensive and offensive response to the rise of foundation model companies that sit outside Meta’s walled gardens.

The long term trend is clear: consumer tech giants are trying to own the decision layer. Whether you book a trip, buy a gadget or choose a bank, they want the initial intent to be captured by their AI, not a search box or a physical store. Commerce becomes conversational, cross platform and increasingly intermediated by systems whose incentives users barely understand.

The European and regional angle

For European users and merchants, Meta’s plan runs straight into the thicket of EU regulation. An AI shopping agent that leverages your content and relationships is textbook profiling under GDPR. That means explicit consent, clear purpose limitation and a genuine option to say no. The Digital Services Act and Digital Markets Act further restrict how gatekeeper platforms can combine data across services and self prefer their own commerce offerings.

The upcoming EU AI Act, with its focus on transparency and high risk systems, adds another layer of complexity. An agent that autonomously recommends financial products, health related items or potentially manipulative in app offers could end up in a stricter compliance category. Dark patterns in how such agents are presented or nudged would not play well with European regulators.

At the same time, there is opportunity. European retailers from Zalando to smaller Shopify style stores could tap Meta’s agents as a new acquisition channel, especially in markets where Instagram and WhatsApp essentially function as discovery engines. But dependence comes at a price: algorithmic opacity, platform fees and sudden policy shifts.

European AI startups, from French foundation model players to German enterprise AI specialists, may take a different path. Rather than building consumer facing agents that monetise via ads and data, they can focus on white label, privacy conscious assistants that retailers control directly. The question is whether that quieter, more regulated approach can compete with the gravitational pull of Meta’s social graph.

Looking ahead

If Meta follows its usual playbook, agentic commerce will roll out gradually, starting with opt in experiments in high value verticals such as fashion, beauty and local services. Expect early tests inside Instagram DMs and WhatsApp Business chats where users already converse with brands. Initially, the agents may look like smarter chatbots that can browse a catalog, check stock and propose bundles.

Over time, the line between chatting with a human rep and a Meta controlled agent will blur. The real power move will be when the agent can act across merchants: not just recommend products from one shop, but weigh options across the entire Meta catalog, apply promo codes, schedule deliveries and keep track of your recurring needs.

Regulators, meanwhile, will watch three things. First, how Meta updates its privacy policies and consent flows to justify using relationship and content data for commercial AI decisions. Second, whether merchants feel coerced into using Meta’s tools to maintain reach, a potential antitrust concern. Third, how transparent and contestable the agent’s recommendations are, especially in sensitive categories.

For users and businesses, the next 12 to 24 months will be an experiment in how much agency we are ready to hand over to invisible systems. The value is obvious: time saved, better deals, fewer clicks. The risks are subtler: enshrined bias, over personalisation that narrows choice, and a commercial layer that becomes ever harder to opt out of.

The bottom line

Meta’s push into AI shopping agents is both strategically brilliant and deeply unsettling. It offers a credible path to monetise massive AI investment and defend attention against new entrants, but it doubles down on the surveillance driven model that has already reshaped the web once. Whether this next layer of personal superintelligence serves people or platforms will depend on how loudly users, merchants and regulators insist on transparency, control and real alternatives.