According to a 2021 report, nearly half of U.S. internet users own a smart speaker. Smart speakers, like the market-leading Amazon Echo and Google Nest, allow consumers to use natural language to do anything from placing an online order to search for a recipe, all with just a simple, “Ok, Google” or “Hey, Alexa.” But beyond the functionalities provided directly by the manufacturer, these speakers also serve as a platform on which customers can connect with any third-party business that offers an app (known as a Skill on the Echo and an Action on the Nest) on that system. For example, customers can ask the Chipotle Skill, “Hey Alexa, reorder my most recent Chipotle order;” they can tell the Whirlpool Skill, “Hey Alexa, start the laundry cycle;” and they can ask CNN Action, “Hey Google, what’s my flash briefing?”

To explore the value-add of these apps, we gathered stock price data for all 112 firms that released a Voice assistant feature on the Amazon Echo or Google Nest between 2016 and 2020. We then excluded data from any firms whose launch of a voice assistant feature coincided with an earnings release (to isolate the effect of announcing the voice assistant alone), leaving us with a dataset of 96 firms.

When used for object-control features, voice assistants have minimal impact on company value.

In contrast to informational features, we found that when a firm introduced an object-control feature, there was no significant impact on stock price. Object-control features allow customers to control physical objects in their homes, for instance by asking Alexa to start a load of laundry or adjust the thermostat. While these features can be helpful in some cases, several inherent limitations have hindered their widespread adoption.

First, these tools generally only allow users to control a narrow set of features, meaning that customers still have to physically interact with the object in order to access the full array of available functionalities. For example, Whirlpool’s voice assistant can start a new load, but to change settings such as laundry cycle type, customers have to use the physical interface on the machine.

Second, as natural language processing technology is still far from perfect, it’s fairly common for voice commands to be misinterpreted — and for object-control features, this can lead to dangerous situations, such as setting the oven to 600 degrees.

Finally, many devices require interactions that are far more complex than would be possible via voice commands. A voice-activated stove, for example, still isn’t able to proactively let you know when water is boiling or food is burning, reducing its usefulness.

In addition, in contexts where voice software does add value, it often does so at the cost of cannibalizing or lowering demand for existing, revenue-generating products and features. For instance, when TiVo released a Skill that allowed users to control their smart TVs, some observers commented that the feature seemed to make TiVo’s existing voice-enabled remote obsolete, potentially causing Amazon Echo owners to be less likely to buy a TiVo remote. Between these factors and the substantial costs associated with developing a voice assistant, we found that any potential positive market reaction to launching an object-control feature was functionally cancelled out by its negative side effects.