How does Cycle use OpenAI today?
Cycle is an AI operating system for physical-product businesses. Codex is part of the engineering workflow used to build and maintain Cycle’s Vue web platform, native iOS application, embedded Shopify app and Python/FastAPI services. In the product itself, the OpenAI API supports consumer-facing product-Twin conversations, product and document enrichment, transcription, text-to-speech and image workflows through a secure server-side gateway.
Cycle uses a model-routed architecture rather than claiming that one provider powers every reasoning path. OpenAI handles real production workloads where its coding, multimodal, audio and conversational capabilities fit the task.
What keeps the AI grounded?
The model does not invent Cycle’s core business numbers. Governed product records and deterministic engines calculate sell-through, stock cover, demand, margin, supplier concentration, product footprint and regulatory readiness. The AI receives those verified outputs together with the relevant catalogue, policy and tenant context, then explains the result or selects an authorized tool.
This separation matters in product operations: a useful agent needs fluent reasoning, but consequential recommendations must remain traceable to data, calculations and permissions.
Where is the OpenAI API used in the product?
- Conversational Digital Twins: a consumer can ask about a product, its care, provenance, Digital Product Passport and available services.
- Catalogue and document intelligence: imported product pages and documents can be normalized, classified and enriched into governed product fields.
- Voice: speech can be transcribed and responses can be returned as audio for mobile and hands-free product interactions.
- Visual workflows: image generation or editing can support governed creative and product-experience tasks.
API credentials remain server-side; the application calls an allowlisted gateway instead of exposing provider keys in the browser or mobile client.
Why do Digital Twins create a large AI surface?
Cycle gives a physical item a persistent digital identity that can be reached from QR, NFC or a commerce experience. A Twin can expose product story, care, authenticity, regulation, repair, returns, resale and other services. The next step is a persistent, conversational concierge attached to that identity.
This turns AI usage from an occasional back-office prompt into potentially millions of small, contextual interactions between companies, products and consumers. OpenAI’s conversational, audio and tool-use capabilities are therefore strategically relevant to Cycle’s unit-level architecture.
What is implemented versus product direction?
Implemented: Codex-assisted engineering; a server-side OpenAI gateway; OpenAI-powered consumer Twin conversation; catalogue and document enrichment; transcription, speech and image workflows. Direction: broader use of the Responses API and tools for durable multi-step Twin workflows, more real-time voice interaction and permissioned agents that can initiate services such as care, return, repair or resale. Cycle states that direction separately from what is live today.