Due to the huge amount of data involved, the high cost and complexity may limit use to larger corporations…
The proliferation of IoT-enabled devices and sensors has filtered down from warehousing and supply chains to the retail sector. Now, to leverage the data from IoT to improve AI, retailers will likely focus on generative AI (GenAI) and Large Language Models (LLMs).
The biggest benefits from the convergence of AI and IoT in retail will be realized by retail organizations identifying intelligent use-cases to deliver benefits to customers, staff members, and the business as a whole.
However, one of the biggest issues with today’s LLM-based AI is that it is relatively expensive and slow. Simply fire-hosing IoT data to an LLM for processing will quickly become unwieldy and very expensive.
Event-driven thinking in retail IoT-AI
One way to reduce the complexity and cost of IoT data streaming in retail AI is to use fine-grained routing via “event streaming”. This allows systems to be more selective in what is analyzed by AI, thereby reducing usage cost while making the response more reactive.
In this event-driven architecture (EDA), the trigger event in retail is a change in state of, say, an item being placed in a shopping cart, or a loyalty-card application being submitted, or an order becoming ready to ship.
When any event occurs, an announcement is “published” to the rest of the EDA system. AI systems will process events to produce real-time results that allow for real-time solutions/actions to be automatically triggered.
This data feed also provides a stream for constant learning, through either ingestion into a vector database or for fine-tuning of the AI/ML model itself.
Three use-cases for EDA
Event-enabling IoT streams can provide benefits to retail customers and employees in-store, via customer service channels and even in warehousing.
Here are three use cases where the convergence of AI and IoT in retail, underpinned by event streaming, can make a real difference.
- In the retail store
By using AI to analyze customer data from IoT devices, retailers can tailor product recommendations, offers, and even in-store experiences to individual preferences.
For example, a customer could tell the store app that they are looking to build a fence. They would no longer have to wait for the hardware-section representative to advise them on which products are best suited to their specific needs, or where to find the products in-store.
Instead, an AI assistant would use store-specific information to provide a response tailored to each customer’s needs.
- Ensuring that stock information is “event-enabled” for AI processing in real-time from the IoT data supplied by sensors and other triggers. Sensors in the store could also be used by the AI to direct customers to the area where the goods are located.
- Integrating both device data and AI processing into an “event mesh” — a network of interconnected “event brokers” that manage the distribution of events among applications, cloud services, and devices. This event mesh facilitates real-time processing of data to produce predictive insights about customer engagement.
- Using EDA to improve post-purchase customer care, such as offering helpful documentation and instructions that explain to customers how to complete their projects when they get home.
- In the call center
GenAI can process recorded or real-time calls to highlight critical issues that need human follow-up, improving customer experience beyond non-AI setups.
At the back end, GenAI bots can be “event-enabled” by connecting them to numerous data points across the customer service process. AI agents can be subscribed to a narrow set of events, programmed to provide a “prompt template” specific to that subscription, and then use an LLM to enhance the event with additional information.
For example, EDA can be used to perform sentiment analysis on user interactions: identifying customers with issues that need expert attention, shortlisting customers ripe for an upsell, or synthesizing new events based on accumulated data.
- In the warehouse
Most retailers now use mobile or tablet devices in warehousing operations, supported by IoT devices for stock monitoring and other inventory-related tasks.
These IoT devices provide valuable data that AI can analyze to glean insights and address potential issues.
For example, a GenAI solution could provide workers with an easy way to report issues, incidents, near misses, or thoughts on improving efficiency. Although qualitative, an LLM-based AI can review, sort, group, and provide curated advice to management.
In an emergency, AI and event-driven systems can greatly increase response times in the warehouse or factory floor, improving safety and operational efficiency.
An event mesh can link many AI agents, each tailored to a specific set of events. For example, subscribing to “all events containing raw audio” allows an AI to use speech-to-text models to create transcriptions, which are then published back into the mesh. All components communicate asynchronously via the event mesh using guaranteed messaging, ensuring that no events are lost in transit and that they are delivered to the appropriate person or device for a prompt response.
The future of retail
The convergence of AI and IoT in retail is not just a fad, it is already achievable with the technologies and data available to retailers today.
The key to making these benefits more accessible lies in an event-driven approach: by selectively feeding relevant data to AI systems, retailers can implement real-time solutions to elevate customer experience, empower employees, and optimize operations.
