Sensor, Meet Sensor: The Autonomous Future of Consumer Goods

A recent tweet from Bill Gates caught the media’s attention and helped shine the spotlight on an interesting milestone achieved by OpenAI, Elon Musk’s artificial intelligence company. 

OpenAI trained a group of five AI-powered bots to play a multiplayer online game by accelerating the equivalent of 180 years of play into a single day. The bots then played against a team of human opponents and defeated them easily.

Why is this a milestone? First, before they could begin playing, the bots used massively accelerated data from past games to learn how to play — both individually and collectively — at a pace and scale that exponentially exceeds human capacity. 

Next, bots and sensor-enabled "things" had already proven their capacity to automate repetitive tasks on their own, but these bots demonstrated they could collaborate and coordinate with each other. They leveraged that knowledge to act in concert, making decisions not only in response to their opponent’s moves but also optimized relative to their team members' actions. Finally, they continue to learn with a level of recall and an absence of bias well beyond human capacity.

While this example is specific to gaming, it’s easy to see how opportunities like these could extend to any number of areas in consumer products. 

For example, SAP is working with beverage companies to install sensors and machine learning-enabled cameras in branded coolers and vending machines. With these sensors installed, the companies can monitor a wide variety of conditions and signals in real time, including:

• Location, to ensure the coolers are placed and operating properly, mitigating the risk of vandalism and theft. 
• Internal ambient temperature, ensuring that products inside the cooler are adequately chilled and fresh, and issuing alerts if the temperature exceeds certain thresholds.
• Stock in the coolers, to monitor merchandising compliance and consumer demand.

Companies are leveraging these signals to integrate with other related processes to (for example) automate work orders for cooler repair or replacement based on status and lifecycle indicators, or automate replenishment orders when stock levels reach certain thresholds. 

The machine learning capabilities associated with these sensors further improve the value proposition by continuously learning and adjusting, predicting and creating optimal targets for processes like service and replenishment at levels of specificity down to the individual cooler that previously would have been impossible.

Given the example demonstrated by OpenAl, it won’t be long before sensors like those installed in the coolers today will begin communicating and collaborating with one another to create intelligent, autonomous processes that continuously adjust to optimize outcomes based on real-time market dynamics.

Expanding the example above, consider if the sensors in coolers within a given region could automatically generate replenishment orders when inventory levels reach a certain point, and that AI-enabled bots would monitor and aggregate those replenishment orders. 

On the one hand, the bots could communicate with nearby warehouse locations and delivery vehicles to optimize truck fill rates and routes, minimizing fuel consumption while ensuring the most at-risk of out-of-stock coolers are prioritized for replenishment. 

On the other hand, those same bots could also collaborate with bots and sensors associated with the company’s manufacturing processes, automatically adjusting short-term forecasts and capacity plans to ensure planned production aligns with current and predicted demand.

Taking it even further, as the AI-enabled bots continuously learn how to optimize forecasts and production, they could collaborate with other AI-enabled bots responsible for procurement and sourcing. All of these bots could work together autonomously to find optimal reorder points and negotiate pricing and contract terms with vendors for commodities and packaging materials based on increasingly accurate predictions of future beverage demand — all derived from observations at the individual cooler level.

While we may not see this level of AI-enabled automation and collaboration between sensors and things on the immediate horizon, rapid advances in artificial intelligence and machine learning — coupled with rapidly decreasing costs for sensors, data streaming and storage — point to intelligent, autonomous processes across a wide variety of end-to-end business processes. 

The same processes applied by OpenAI to rapidly train a group of bots to work together in a way that bested human capacity will soon be applied to business processes, opening the door to unprecedented new levels of productivity, efficiency, cost savings and new business opportunities.