Challenge

Wyze is a fast-growing smart home brand shipping millions of devices across multiple channels including direct-to-consumer, Amazon, and retail partners. The company faced a cascading set of challenges that rules-based automation and manual processes could not address.

First, there was no established playbook for processing orders placed by AI agents. When a consumer tells an AI assistant to find a good smart home camera under $40, the agent needs to discover products, evaluate options, and transact. The resulting order then needs to flow through validation, fulfilment routing, shipping, and post-purchase support. No traditional ecommerce workflow was designed for this.

Second, Wyze's multi-channel, multi-fulfilment operation spanning Amazon Multi-Channel Fulfilment, direct warehouse, and 3PL partners already generated significant complexity. Before the implementation, managing order routing across these fulfilment paths required dedicated full-time staff making manual decisions about which orders go where, based on inventory levels, geography, shipping costs, and carrier availability. These decisions needed to be made rapidly for each order, a fundamentally unscalable approach as order volumes grew. Rules-based routing tables could handle basic scenarios, but could not adapt to real-time inventory fluctuations, carrier disruptions, or the dynamic cost optimisation that Wyze's operation demanded.

Third, adding AI agents as a new sales channel could not mean adding yet another operational silo. If every new channel requires its own manual processes and monitoring, the operational overhead scales linearly, defeating the purpose of composable architecture.

The core insight was that agentic commerce introduces decision-making requirements at a speed, volume, and complexity that humans and rules-based systems cannot match. On the pre-purchase side, AI shopping agents need to make autonomous product evaluation and purchasing decisions because they are acting on behalf of consumers in real time. A rules-based chatbot could present a static product list, but cannot interpret nuanced consumer preferences, compare products on multiple dimensions, and make a judgment call, all within seconds. On the post-purchase side, autonomous fulfilment orchestration addresses a scale challenge that rules-based approaches cannot solve.

Strategy

Wyze deployed a two-layer agentic commerce pipeline that spans the entire customer journey from product discovery through to post-purchase management.

On the pre-purchase side, AI agents including ChatGPT, Claude, Gemini, and Perplexity autonomously browse Wyze's product catalogue. These agents interpret consumer intent, evaluate product options, compare specifications, and complete purchases, making autonomous buying decisions rather than following scripted purchase flows. The agent selects products, initiates checkout, and processes payment without any human in the loop.

On the post-purchase side, once payment is processed, an AI-native order operations platform takes over autonomously. It validates the order, makes real-time fulfilment routing decisions (selecting between Amazon Multi-Channel Fulfilment, direct warehouse shipping, or 3PL partners based on inventory availability, buyer geography, carrier performance, and cost), syncs inventory across all channels, and pushes order lifecycle events to Wyze's ERP and logistics partners.

A critical piece of the implementation is an industry-first Model Context Protocol (MCP) server purpose-built for order management. This enables AI agents to continue operating autonomously after purchase: querying order status, tracking shipments, and initiating returns through natural language, adapting their actions based on each order's real-time state. This transforms post-purchase operations from a black box into an agentic interface. Any MCP-enabled AI agent can now query order status, track shipments, check inventory, and initiate returns through natural language, capabilities that previously required human customer service agents or custom integrations for each AI platform.

The architectural innovation is that adding AI agents as a new buyer channel required zero changes to Wyze's existing fulfilment infrastructure. Traditional ecommerce treats each new channel (marketplace, direct-to-consumer, wholesale) as a new integration project. By building on composable, API-first architecture, the agentic channel simply plugs into the existing orchestration layer.

The system has been operational since January 2026, with Wyze running on the order orchestration platform since 2023 and the MCP server for order management launching in September 2025.

Impact

The agentic commerce pipeline is delivering measurable outcomes across operational efficiency, customer experience, and business performance.

Order processing speed has transformed from manual fulfilment routing decisions that required hours per batch to autonomous orchestration that processes and routes orders in real time, reducing the order-to-fulfilment handoff from hours to seconds. Click-to-delivery times have been reduced by more than 50%, with intelligent, autonomous fulfilment path selection consistently routing each order to the optimal fulfilment location based on real-time inventory, geography, and carrier performance.

Manual intervention has been eliminated for routine order processing. Autonomous orchestration removed the need for dedicated manual order processing staff, with humans retaining oversight at the strategic level (fulfilment partner selection, pricing strategy) while tactical execution operates autonomously. The same infrastructure handles growing order volumes across direct-to-consumer, marketplace, retail, and now agentic channels with no incremental operational headcount. Adding AI agent buyers as a new channel required zero additional manual processes.

For customers, the agentic pipeline reduces purchase effort to its minimum: express a need to an AI assistant, and the agent handles everything from discovery through delivery tracking. Post-purchase, the MCP server enables AI agents to query order status and initiate returns through natural language, replacing the traditional friction of logging into accounts, navigating order history, and contacting customer service. The agentic interface is also inherently more accessible than traditional ecommerce, enabling consumers who find website navigation difficult, who are visually impaired, or who simply prefer conversational interaction to purchase and manage products through natural language.

For Wyze as a business, agentic commerce opens an entirely new acquisition channel. AI agents across platforms can now recommend and sell Wyze products to consumers who may never have visited the website directly. This represents incremental revenue with near-zero customer acquisition cost. Autonomous fulfilment routing based on real-time inventory data also reduces scenarios where orders are routed to locations with insufficient stock, reducing waste in shipping costs and customer-facing delays.

The near-term roadmap includes proactive agentic reordering where AI agents autonomously identify when customers need consumable replacements based on purchase history and product lifecycle data, multi-brand agentic bundles, and agentic customer service handling returns, exchanges, and troubleshooting end-to-end.

Composable architecture in action

The agentic commerce pipeline between Wyze and its technology partners would not be possible without a composable, API-first architecture. Every system in the pipeline exposes its capabilities through well-defined APIs, which is what allows AI agents to coordinate across vendor boundaries and execute decisions autonomously.

In practice, an AI shopping agent discovers products through an agentic commerce API that exposes product catalogue data, pricing, and real-time inventory. The agent processes payment through a payment API. Upon successful transaction, order data flows to the order operations platform. The orchestration engine, operating as an independent microservice, receives the order, accesses real-time inventory data from multiple fulfilment partners via their respective APIs, evaluates fulfilment options, and autonomously routes the order to the optimal location. Order lifecycle events flow back through APIs and via the MCP server to AI agents for post-purchase operations. At every stage, independent systems communicate through APIs. No system needs to know the internal workings of any other.

The MCP server is where composable architecture and agentic capability converge most powerfully. It exposes order operations through a standardised protocol that any MCP-enabled AI agent can use natively. One MCP server serves all AI platforms. This is the same principle that makes MACH architecture powerful: standardised, open interfaces that decouple providers from consumers.

The modularity of this architecture was critical in three specific ways. Adding the agentic channel required no rebuilding. Wyze did not need to change its fulfilment setup, its ERP integration, or its 3PL relationships. The agentic channel was, from an architecture perspective, simply another API-connected order source feeding into the same orchestration engine. Swapping and scaling fulfilment partners is straightforward: a new API connection, not a rebuild of the agentic pipeline. And because the MCP server uses an open protocol standard, the system is not locked into any single AI platform. As new AI agents emerge or existing ones evolve, they can access order operations capabilities through the same standardised interface.

This implementation demonstrates that MACH architecture and agentic capability are mutually reinforcing. The composable foundation made it possible to deploy agentic commerce with minimal effort and no replatforming. And the agentic capability, in turn, validates the investment in composable architecture by delivering a use case that monolithic systems simply cannot support.