Differentiating User Experience via Generative AI

While Matter has been incubated within the smart home industry, an even more significant disruptor has hailed from beyond it. The launch of ChatGPT by OpenAI has created a tsunami of heightened interest around generative AI and how we will interact with a digital world. It has also posed a challenge to what had once been hailed as the state of the art in human-AI interaction: voice-driven assistants such as Alexa, Google Assistant, and Siri. Following decades of sci-fi portrayals of natural voice communications with computers from the mindful (the Star Trek ship computer) to the murderous (HAL 9001), these assistants raised consumer expectations of device interactions. However, their limited knowledge bases often led to answers reflecting misunderstanding or ignorance.

Controlling home devices, however, emerged as a proficiency as it is convenient and natural to ask an assistant to turn on a light or raise the thermostat. Voice assistants have thus become one of the primary ways that consumers interact with smart devices.

While they have improved over time, the core assistant experience has not evolved much since their debut a decade ago. Smart speaker and display adoption has plateaued with roughly half of US internet households owning such a device, but has seen little expansion of the user base since 2020. This stands to change with the integration of generative AI. Google, for one, has been all-in on the technology since ChatGPT emerged as a competitive challenge, particularly as Microsoft integrated it with rival search engine Bing. And while what Google has shared of its AI-integration roadmap hasn’t included the Google Assistant—in part because of the challenges of integrating two approaches to AI—ultimately a voice agent infused with generative AI capabilities delivers the best of both worlds and will pen new applications.

Rather than type prompts into a rectangle, consumers will be able to make requests via voice and touch as they do today. Generative AI-infused agents will continue to push capabilities far beyond what these assistants can do today, including presenting different scenarios and personalizing answers based on the level of a consumer’s understanding. For example, a resident could respond to an alert that water pressure is dropping with follow-up questions asking about possible causes and remedies, questions that today’s assistants would likely hand off to a basic website citation. At its September 2023 device launch, Amazon showed off a generative AI-infused version of Alexa that moves the assistant several rungs up on the evolutionary ladder, greatly improving its knowledge base and conversational style. The company even showed a “chat with Alexa” feature that dispenses with wake words and may prove helpful to those suffering from loneliness and social anxiety.

These capabilities will help restore the air of omniscience shattered once an assistant says it can’t answer a question and will provide a more literal definition of a “smart” home, improving the user experience without forcing any interface changes. At the September 2023 launch event, Amazon showed how the improved Alexa will be better at identifying smart devices in the home from context, e.g., “the bedroom light” instead of having to remember and invoke a specific name for each device. From the consumer perspective, Alexa infused with generative AI will be just a smarter Alexa.

Josh.ai, startup partnered with Amazon, offers a voice-driven agent aimed at outfitting luxury home automation capabilities while providing enhanced privacy. In a recent interview, company CEO Alex Capecelatro praised how integrating ChatGPT has allowed his system to now offer answers to a wide range of questions beyond the system’s optimized understanding of a home’s structure while opening it to general questions and even creative applications.

One challenge, however, has been deciding which requests to handle directly and which to hand off. He notes how a guest sought to use Josh.ai to close shades covering a skylight. However, as that command was not yet in the system, it handed the request to ChatGPT, which provided instructions that included getting on a ladder. This highlights a broader issue that device makers will face in their AI development efforts. However, given the high customizability of generative AI, vendors should begin exploring ways to integrate as soon as possible to differentiate and avoid falling behind competitors.

Another challenge will prove a more significant barrier for the ecosystem stewards: generative AI’s susceptibility to misinformation and “hallucinations,” untrue statements or invented phenomena presented as facts. Today’s voice-driven assistants may be clunky in how they share web-based information, but their references are typically transparent and well-sourced. Microsoft and Google can place prominent warnings and caveats in the context of a web search. But an interactive dialogue striving for natural interaction leaves less opportunity for explanation or tolerance for distrust-engendering misinformation.

This is an excerpt from Parks Associates new report, Interoperability & Generative AI: Next Generation Smart Home UI, authored by Ross Rubin , sr contributing analyst. A unified smart home experience is the holy grail of convenience and value for consumers, but fragmentation and interoperability issues plague the smart home ecosystem. This report evaluates consumer preferences for smart home control and UI, trends in interoperability and standards, such as Matter, and the expected impact of generative AI on smart home controls. It includes a 5-year forecast of the installed base of smart home devices in the United States.?

Thanks for reading. If you have a product in the market and would like to do a briefing, please contact me or any of our team.

?