The Rise of Agentic Commerce: How Physical AI Is Redefining Retail

by Jamie Hall, Co-Founder & Chief Marketing Officer, Pentatonic

Commerce infrastructure is broken. Not visibly, not catastrophically, but structurally. For twenty-five years, the industry has poured billions into optimizing a single moment: the first sale. Everything that happens after checkout - the return, the resale, the trade-in, the end-of-life - has been treated as an afterthought, a cost center, or someone else's problem.

That someone else has been paying attention. Third-party platforms capture the lion's share of post-consumer value, dictating pricing, positioning, and customer experience while brands watch from the sidelines. Post-consumer commerce is growing more than five times faster than primary retail, yet the companies with the deepest product expertise, the strongest customer relationships, and the most at stake have largely ceded the territory.

This is beginning to change. A new category of ecosystems is emerging, one that treats commerce as continuous rather than transactional, and products as assets that can be understood, tracked, and optimized to move through multiple use-cycles. At the bleeding edge of this shift is Physical AI.

From Language Models to Product Intelligence

The past three years have demonstrated what happens when machines learn to understand vast amounts of data. Large Language Models trained on text have transformed search, customer service, content creation, and software development. But language in text-form, is only one domain of human intelligence. Commerce - the exchange of physical goods - has its own data foundation, its own semantics, and its own intelligence requirements.

What language models did for text, custom-trained Product Intelligence Models are now doing for commerce. These are purpose-built systems that understand products: not just their names and SKUs, but their materials, conditions, provenance, usage patterns, market values, and optimal next destinations. Where an LLM predicts the next word in a sentence, a Product Intelligence Model predicts the next state and value of a product - and increasingly, acts on that prediction autonomously.

This is the foundation of agentic commerce: AI workflows that don't simply recommend or inform, but orchestrate the transaction, routing, pricing, and settling on behalf of consumers and brands. The shift from assistive to agentic AI represents a step-change in what retail infrastructure can accomplish.

Physical AI and the Intelligence Layer

During my years at Nike, running NikeLab, their global innovation platform at the time, I watched the gap between digital sophistication and physical retail capability widen year after year. Online experiences became personalized, predictive, and seamless. In-store experiences remained largely static, dependent on human knowledge that varied wildly by location and shifted with every staff change.

The missing piece was always an intelligence layer that could operate in the physical world with the same fluency it operated online. That layer is now possible.

At Pentatonic, we've developed Physical Intelligence Nodes (PINs), compact, modular, deployable edge-compute devices that bring Physical AI and agentic capabilities into retail and logistics environments. A single PIN can power personal shopping, process returns, run trade-in programmes, authenticate products, assess condition, feedback insight to Product and Brand teams and even train retail staff. And it's all connected to the same underlying intelligence ecosystem.

The technical architecture matters here. By running Physical AI and specialised, compact models at the edge - on small, power-efficient, embedded compute in the hardware itself - we eliminate the latency and connectivity dependencies that have historically made in-store AI impractical. Computation happens locally, in real time, with results delivered as instantly as a lightbulb illuminates - but at a fraction of the energy consumption.

What Agentic Commerce Unlocks

The implications extend across both sides of the transaction lifecycle.

Pre-consumer, Product Intelligence Models enable conversational product discovery that replaces the keyword search and tick-box filters that have dominated e-commerce for two decades. AI personal shoppers can understand intent, not just input, guiding customers to the right product rather than the most keyword-relevant one. Staff training becomes continuous and contextual, keeping associates current on materials, innovations, and product details without pulling them from the floor.

Post-consumer, the same intelligence layer automates returns processing with instant product identification and routing. Trade-in and buyback programmes gain real-time pricing and settlement capabilities. Authentication, condition grading, and wear assessment - historically manual, inconsistent, and expensive - become automated, objective, and scalable. Products can be routed to their highest-value outcome: resale, refurbishment or recycling.

The critical insight is that these aren't separate systems. The same AI ecosystem that helps a customer find the right product can later authenticate it, price it, and route it to its next owner. When the intelligence layer is unified, commerce becomes continuous, an unstoppable growth engine.

Use-Cycle Intelligence: A New Era of Brands Understanding Their Products

Every interaction with Pentatonic's agentic commerce stack generates data: condition assessments, wear patterns, failure signals, demand indicators, behavioral insights. Aggregated across millions of products and thousands of touchpoints, this creates what we call use-cycle intelligence - a continuous feedback loop for product, brand, and merchandising teams to inform their own increasingly intelligent product creation systems.

Physical products begin to behave like software: observable, continuously updated, and optimizable over time - and AI models absorb these signals as units of information, tokens to generate their next insight. Brands gain visibility into how products actually perform in the real world, not just how they sell. Design decisions can be informed by actual usage data. Pricing strategies can reflect true residual value. Customer relationships can extend across multiple ownership cycles rather than ending at the first checkout.

In my former life within product companies, where we were essentially blind to precisely how consumers were using our products, this depth of performance and usage insight would have been unthinkable, the stuff of dreams.

The Market Moment

The convergence of enabling technologies - edge compute, Physical AI, foundation models, real-time payment infrastructure - has made this possible now in ways it wasn't five years ago. Pentatonic is now supporting over 150 brands to reduce consumer friction and unlock significantly improved value from their products, through AI pricing and authentication, with on-the-spot settlement via rewards and incentives - at home and in stores.

The opportunity is substantial. An estimated $30 trillion in products sit idle in homes and warehouses globally, value that current infrastructure is not set up to access, price, or move. Post-consumer commerce alone represents a multi-hundred-billion-dollar market growing at multiples of traditional retail every year. The brands that build or adopt agentic commerce infrastructure will capture that value. Those that don't will watch it flow to platforms that have no particular interest in protecting brand equity or customer relationships.

The Next Era of Commerce

Today's commerce stack was designed for a world where transactions were discrete events and products disappeared from view after purchase. That world is ending. Consumers increasingly expect brands to take responsibility for products across their entire lifecycle. Sustainability pressures demand it. And now, technology makes it commercially viable.

The next era of commerce will be built on product truth - structured, scalable, machine-readable understanding of what products are, where they are, what condition they're in, and where they should go next. Agentic AI will deepen, and act upon, that truth, autonomously and at scale. This represents the first step towards foundation models for commerce - and the Product Intelligence Model is that foundation.

The infrastructure is ready. The leading brands already know it - and they're moving now. The success stories of 2026 won't be written by those who waited for certainty. They'll be written by those who recognized the shift and built ahead of it. The only question left is which side of that line you'll be on.

---

About the Author

Jamie Hall

Co-Founder & Chief Marketing Officer

Pentatonic

As Co-Founder and Chief Marketing Officer, Jamie directs the integration of customer experience with multi-commerce principles, ensuring that multi-commerce is embedded across the platform, partnerships, and programs. He brings invaluable insights from his years in senior leadership roles at Nike and Levi’s, where he developed extensive knowledge of product design, brand communications, marketing operations, and customer service innovation.

Connect with Jamie on LinkedIn