The world has been infatuated with artificial intelligence (AI) ever since OpenAI released ChatGPT to the public in November 2022. The onslaught of tech companies releasing their own AI models since then has further validated AI’s growing presence in our work and personal lives. While the focus of this new age of advanced AI has largely been catering to the business world, the technology is slowly beginning to make its way to the smart home industry.

The likes of Microsoft, Google, Amazon, Apple, Meta, and a host of other major brands have thrown their weight behind conversational and generative AI tools, and now, once-conceptual ideas of a hyper-intelligent home experience that anticipates a homeowners’ preferences and adjusts automations accordingly aren’t really that far away.

Early Applications of Conversational AI

This new age of advanced AI largely began with generative AI chatbots like ChatGPT that were at first able to create written content from nearly any prompt. Now, this is evolving to include other communication modes, such as voice, which has long been a smart home control option.

Companies like Google, Amazon and Apple are beginning to bring their advanced AI models to the consumer space, allowing for more conversational and natural language voice control of devices.

Josh.ai is one of the early adopters of this technology, using large language models with its JoshGPT feature that not only allows for control of lights and other devices, but also as an AI companion that can assist with trivia questions, cooking requests, bedtime stories, and more.

2024 Lighting Controls and Fixtures Report

Lightapalooza took place in late February, and the growth of the event has mirrored the rapid ascension lighting fixtures and controls.

Download your copy now!

“I think AI will make the single biggest impact on the home automation industry compared to any other technology in my lifetime,” says Alex Capacelatro, CEO of Josh.ai and CEDIA Board of Directors chairperson. “That said, currently very few are doing much with it.”

According to Capacelatro, the custom smart home industry has historically been a hardware-focused industry. A walk of the show floor at CEDIA Expo will make that apparent, as hardware such as displays, speakers, tablets, switches, shades, and more dominate the show floor.

For the last two decades, software has become a larger part of the industry, allowing manufacturers to build some intelligence and connectivity into their interfaces and devices. The same thing is now happening with AI, the Josh.ai chief says.

“The reality is, AI is going to completely change the way we listen and the way we control our environments, even smart lighting and making sure the color temperature and everything is just right,” Capacelatro says. “We won’t even have to think about it.”

Where AI Can Excel in Smart Home Applications

Aside from voice control, which should only enhance the user experience with its natural language programming and proactive suggestions, AI really excels in areas that end customers don’t really ever see.

Configuring advanced systems and personalizing automations for the client becomes much easier with AI. According to Capacelatro, one key feature of Josh.ai’s platform is the ability to create automations by simply telling the system what they want to happen using voice. Similar to how individuals can ask ChatGPT or Google’s Gemini to summarize difficult topics in a concise manner, complex automations can be created simply by telling the system in natural language.

In addition, AI does a lot behind the scenes with monitoring systems, flagging alerts when devices need attention, and keeping homes secure.

However, companies need to exercise caution when developing AI systems and incorporating them into people’s homes, as too many automatic changes to automations and scheduling could negatively impact the customer experience.

According to Capacelatro, the Josh.ai system can learn habits, make suggestions, and even automate all aspects of the home as it better understands and learns a client’s lifestyle, but manufacturers and integrators need to tread lightly to respect both the experience and privacy.

“When done right, the home should suggest or automatically play the right music, at the right volume, in the right locations,” he says. “Or adjust the thermostat down if people enter a hot room, or close the shades if the sun is glaring in as they try to watch a movie. There are so many ways AI will enable the home to help clients live more comfortably in their homes.”

Just Remember: AI is Still in Its Infancy

While super-intelligent systems at first glance seem like a great thing that any customer would want, the industry needs to recognize that these new AI applications are in their infancy, especially as it relates to the home and using AI to predict a homeowners’ activities.

According to Brad Hintze, executive vice president of global marketing at Crestron, moving too quickly with AI in the home can ruin a homeowner’s experience with smart home technology.

“There’s a lot of promise, and it’s super interesting, but we have to make sure we don’t mess it up,” Hintze says. “This is something our market has to be really careful about.”

Hintze says there is a “determinism” around the design of most home automation systems with a device or app interface like Crestron Home. Users typically control systems and turn on lights by pushing a button. The idea of using AI to collect that data and then automate systems based on a historical snapshot of user activity risks turning those manual controls into a guessing game for the AI.

“At the end of the day, a homeowner wants what they want done, and there’s no way for an AI to accurately predict, every time, what you want,” Hintze says. “Maybe they could 60% or 70%, but the first time it’s wrong, the homeowner isn’t going to rely on it.”

Essentially, homeowners value predictability, and reliability — attributes that this form of predictive AI cannot always guarantee.

Where AI Has Applicable Effects Today in Smart Home Tech

However, AI has immediate applicability in improving existing products and services. For example, AI can help adjust settings in an audio system to balance speakers according to a room’s layout. In fact, that technology exists in a number of audio systems. Crestron is also working on this technology for its own, Hintze says.

Other use cases include Crestron’s Solar Sync, a feature that will read the color temperature of the sun, factor in the time of day, and adjust lighting systems appropriately. Going further, potential use cases would include an individual’s own biometrics being used to adjust lighting systems to adjust to a circadian rhythm, Hintze says.

One important thing to consider is that not everything needs to be fully automated based on an AI model’s predictions of user habits. Essentially, home automation is not a zero-sum game, says Capacelatro.

AI models in the home should tread lightly, and at first, simply offer suggestions based on what it thinks the user wants to do.

“When we start to learn patterns, we want to ask the user, ‘Is this right? Should we continue doing this? It’s really hard to screw up when you take that approach,” Capacelatro says.

How Manufacturers Are Approaching AI in Their Products

One of the key strategies many manufactures are taking to integrate AI into its systems is maintaining an open ecosystem and allowing customers to choose their preferred AI services, whether it’s voice control systems like Josh AI, Siri, Google Assistant, or Alexa.

“We think that open ecosystem is super important to be able to integrate a wide variety of products and take advantage of those AI investments,” says Crestron’s Hintze.

The same is true in the commercial sector, where Crestron hardware helps support Microsoft Teams AI features that identify individual speakers. Essentially, Crestron wants its hardware to be the endpoint that enables and enhances these integrations with AI services.

Similarly, Savant, which has closely aligned itself with Apple with its main home controller running on the Mac OS platform, is primed to reap the benefits of the device maker’s recent investments in AI, such as Apple Intelligence, a new ChatGPT-like feature that will be embedded into iPhones, iPads, and Macs.

Savant’s Apple-based automation allows them to tap into Apple’s M series chips, vast datasets and powerful AI APIs, which Hall sees as a game-changer for smart home technology. According to Chris Hall, product manager for Savant, the company is now exploring using those APIs for security purposes and AI-based object detection.

Hall identifies two primary AI-driven products currently shaping the smart home landscape. The first focuses on processing intents, which he describes as “those dog tricks, right? You can ask it anything… like who’s on TV, what movies are playing, what’s popular.” This capability is seen most prominently in voice control systems, such as those powered by Google Home, Alexa, and Apple’s Siri.

The second area Hall notes involves automating routine tasks for users, such as suggesting activities based on habitual behavior. However, like his counterpart at Crestron, Hall says this is dangerous territory.

Instead, Hall says the industry should shift its focus to leveraging AI to make intelligent home systems that function as a home’s support staff. Home systems would be able to process basic, natural language requests and automate mundane tasks.

“I kind of always think of ‘The Great Gatsby’ where there are hundreds of support staff that are managing these properties,” Hall says. “I think AI will actually be able to take the place of that.”

Where Might AI Take the Smart Home?

Looking ahead, Hintze shared an intriguing concept that Crestron is exploring: AIoT, or Artificial Intelligence of Things. This concept involves leveraging the data generated by various devices in the home to create an individualized AI stack. This AI stack could infer intelligence from the data, offering personalized insights to enhance daily life.

For example, a smart home system could monitor sleep patterns, lighting preferences, and other behaviors to optimize the living environment continuously. While Crestron may not build the AI stack itself, it aims to contribute to it by ensuring that its devices can feed valuable data into these systems.

“The core foundation of that is you have to have the data to start,” Hintze emphasizes, pointing to the potential of AIoT to revolutionize how we interact with our homes.

However, as they push the boundaries of innovation, these companies are also keenly aware of the significant challenges posed by data privacy and security. Balancing AI’s capabilities with the responsibility to protect user data is a central concern shaping their strategies.

According to Hintze, Crestron is committed to keeping data local rather than relying on cloud-based storage.

This approach is particularly important in the high-end custom installation market, where clients often include prominent individuals with heightened privacy concerns.

“They view it as essential, and that’s why they chose Crestron, because of that level of security,” Hintze says.

This Is Where We Need to Talk About Data Privacy in AI

Sophisticated AI-driven automations require significant data, and one way to get that data in a home is by leveraging sensors that monitor users and systems through the home. However, Savant’s Hall cautions the market may be hesitant to accept that level of monitoring.

“I wouldn’t want that level of surveillance in my own home, and I know many consumers feel the same way,” Hall says. However, Savant is able to rely upon Apple’s enormous dataset, and thus isn’t responsible for maintaining the security and privacy of its customers’ data.

Perhaps similar to how consumer products and services are increasingly making users aware of how their data is being used and providing options to stop that data from being shared, the smart home industry could allow users to access a local model that helps an AI system learn even more about a user’s habits.

Capacelatro says AI models require a lot of data, but user’s don’t need to give up their data to get access to those models. At Josh.ai, models are designed to offer users the benefits of advanced technology without the need to share personal data, ensuring that privacy remains intact. However, users can choose to open up access to a local model to enhance their experience.

“If you said, ‘Play some music,’ we can play music that we know you’ll like if you allow that, but that’s up to the customer whether or not they want to open that,” he says. This approach allows the AI to become smarter and more personalized, but Capacelatro stresses that the decision to share data is entirely up to the user, ensuring that privacy and control are prioritized.

• Page 1