What core capabilities appear in Wavesteam's demonstrated apps and systems?

Wavesteam's public work broadly covers consumer apps and mini-programs, enterprise systems, AI applications, and IoT platforms. Their core loops are respectively completing a customer service, moving enterprise work to an accountable outcome, assisting a defined task with a model, and monitoring and operating equipment. They are not one reusable feature catalogue.

The useful case question is not how many pages, memberships, payments, and reports exist. It is whether a real workflow closes: how a user completes learning or a booking, an order exception is resolved, employee data is restricted, an AI error is reviewed, or an offline device creates an actionable alert. Login, upload, messaging, and administration are foundations, not proof of business value.

Product type	Core task	Supporting capabilities	Demonstration focus	Public direction
Consumer app/mini-program	Discover, choose, order/learn, fulfil, after-sales	Account, content/product, payment, messaging, service, membership	First use, payment failure, refund, weak network, account deletion	QuanYuTong, community, events, lifestyle
Enterprise system	Enter, review, execute, reconcile, analyze	Organization, CRM/order/stock, approvals, reports, logs, integrations	Exceptions, field permissions, audit, export	Factory, park, renovation, quotation
AI application	Retrieve, extract, generate, classify, support decisions	Source preparation, knowledge, model, citations, evaluation, review	Failure, refusal, access, cost, version change	Learning, recruitment, documents
IoT platform	Connect, observe, alert, control, maintain	Protocol, telemetry, time series, maps, tickets, firmware, permissions	Offline data, duplicates, command confirmation, recovery	BMS, energy swap, drones

Consumer products differ by core loop. Education emphasizes learning paths and feedback; appointments use time inventory, confirmation, and cancellation; communities need publishing, moderation, interaction, reports, and account governance. The public QuanYuTong case shows one learning/AI direction but cannot be copied into another product without discovery.

For an enterprise case, inspect how data travels through roles and exceptions: submission, review, payment, warehouse, and consistent reporting, including rejection, amendment, partial payment, and employee handover. The factory management case establishes a manufacturing direction, not a mandatory module list for another client.

An AI demonstration must include failures. Representative sanitized questions or documents should test citations, unauthorized sources, appropriate refusal, and human corrections. “RAG,” “agent,” and “OCR” are implementation forms; relevant measures may be review time, first-contact resolution, or severe errors. The AI recruitment case proves related experience, not a promised result.

IoT demonstrations should show identity, protocol version, disconnection recovery, timestamps, alert acknowledgement, command authorization, and logs—not only charts and maps. Sent, received, and executed commands are separate states. The BMS and energy swap cases provide public context before a proof on the target protocol.

Also establish Wavesteam's actual responsibility for product, UI, frontend, backend, model, device protocol, deployment, and operations and whether the artifact is live, historic, or conceptual. Confidential data can be sanitized, but a static design should never be presented as a production system.

Choose reference cases by business structure rather than industry name. Classify their capabilities as reusable foundation, configurable rule, redesigned requirement, or unnecessary to derive the first scope. The case centre and app introduction are the complete public entry points.