The Smart Speaker Economy: How Much of It Reaches Human Voice Actors

Voice assistants are one of the largest, most habitual uses of synthetic speech in daily life, and one of the least examined from a working voice actor's perspective. This article covers what is actually documented: how many people use these assistants, who the human voices behind Siri, Alexa, and Cortana were reported to be, how synthetic those voices have become since, and the one clearly documented corner of this industry where human voice actors still get hired directly.

How Many People Actually Use These Assistants

eMarketer's 2025 forecast puts total US voice assistant users at 154.3 million, projected to reach 168.2 million by 2029, broken out by platform as Google Assistant leading at 91.9 million users, Apple's Siri at 86.5 million, and Amazon's Alexa at 77.2 million; these figures overlap since a large share of people use more than one assistant across different devices. The hardware side of the market tells a more mixed story: IDC's Worldwide Quarterly Smart Home Device Tracker reported global smart speaker unit shipments fell 8.8 percent in 2024, described by IDC as the steepest single-year drop in the category's history, even as the number of people actively using a voice assistant on some device kept growing.

Who Actually Voiced These Assistants, Originally

Voice actor Susan Bennett recorded four hours a day, five days a week for a month in July 2005, for a company called ScanSoft, later acquired by Nuance, without being told what the recordings were for. She only learned in 2011, after the iPhone 4S launched, that her voice had become Siri, when a fellow voice actor called to tell her. Apple has never officially confirmed this, but CNN had forensic audio expert Ed Primeau independently verify the match with what he described as 100 percent certainty, and Bennett has spoken publicly about the discovery, and about receiving only her standard 2005 session fee with no residuals from Siri's global use, in interviews with CNN and NBC News.

Amazon's Alexa has a less confirmed but widely reported origin: journalist Brad Stone, in his 2021 book *Amazon Unbound*, identified Colorado-based voice actor Nina Rolle as the likely source voice, after matching her voice on old radio ads to Alexa's speech patterns. When Stone reached her, Rolle said she was not permitted to discuss it, and Amazon has never officially confirmed or denied her identity, so this remains a widely reported, never-confirmed match rather than a settled fact. Microsoft's Cortana is the one clearly confirmed case: the company itself publicly credited actor Jen Taylor, already known for voicing Cortana in the Halo game franchise, as the voice of the Windows Phone digital assistant, a rare instance of a tech company naming its assistant's voice actor on the record.

Even a 'Synthetic' Voice Is Trained on a Real Actor

All three major assistants have moved to fully synthetic, neural text-to-speech delivery rather than playing back recorded human audio clips. What is less widely understood is that this does not remove the human voice actor from the process; it moves them earlier in it. Apple's own machine learning research team has published that Siri's current voices are built from a neural text-to-speech system trained specifically on studio recordings of professional voice actors, meaning every new synthetic Siri voice still starts with a real actor recording hours of varied material before any model generates a single word of live speech. Apple has also continued to record new voice actors for new Siri voice options over time, including voices added specifically for accessibility and diversity in recent iOS updates.

The Clearest Documented Case of Paid, Direct Voice Actor Work

Amazon's Polly "Brand Voice" program lets a company work directly with Amazon's team to build a custom neural text-to-speech voice modeled on a voice actor the company hires, for use in its own branded Alexa Skill. The best-documented real example is KFC Canada, which in February 2020 hired a voice actor across multiple sessions to build a Colonel Sanders voice model in a Southern US accent, then used that model to power an Alexa Skill letting customers chat with a digital "Colonel" and reorder food by voice, confirmed by KFC's own global press release and covered independently by Voicebot.ai, Marketing Dive, MediaPost, and Campaign Canada.

This is the clearest documented instance found of a brand hiring a human voice actor specifically for smart-speaker work in the way this article set out to check, and it points to a real, ongoing niche: companies building a distinct branded voice persona for their own Alexa Skill or Google Action, rather than using a platform's generic default assistant voice. No neutral market-research firm appears to publish a figure sizing how much of this specific work exists industry-wide, and this article will not invent one; the honest answer is that the niche is real and documented, but not currently measured.

The Case for Keeping a Human Voice in the Loop

Deborah Harrison, who led the team that wrote Microsoft's Cortana personality, has spoken publicly about why the industry designing these voices should not treat the human element as a solved, disposable problem:

"It's critical to us and any product we work on that there's never any ambiguity about whether you are talking to a person or a device."

That transparency principle sits alongside a separate, practical point: voice actors already working in short, recombinable phrase-based recording for IVR and phone systems or automotive and navigation prompts are doing structurally similar work to what a Brand Voice project requires, just without the assistant-platform branding attached, which makes this a more reachable niche for an experienced voice actor than it might first appear.

A Huge, Habitual Market Built on a Shrinking Sliver of Direct Human Work

More than 150 million Americans talk to a voice assistant regularly, and the voice they hear started, in every documented case in this article, as a real person recording in a studio. What has changed is not whether a human voice actor was ever involved, but how far upstream that involvement now sits: fewer live recordings, more training data for a model, and one clearly documented, narrow niche, custom branded assistant voices, where direct, paid human work still visibly exists today.

Sources and Further Reading

• eMarketer, Voice Assistant User Forecast 2025: the US user figures by platform.
• IDC, Worldwide Quarterly Smart Home Device Tracker: the 8.8 percent 2024 shipment decline.
• CNN Business, Original Siri Voice Susan Bennett and NBC News, There's a Real Human Behind the Voice of Siri: the Susan Bennett story.
• Seattle Times, Boulder, Colorado Woman Identified as Person Who Voiced Amazon's Alexa: the Nina Rolle reporting, sourced to Brad Stone's *Amazon Unbound*.
• Seattle Times, Seattle Actor Jen Taylor Reprises Role of the AI Cortana: the confirmed Jen Taylor and Cortana connection.
• Apple Machine Learning Research, Siri Voices: Apple's own confirmation that Siri's neural voices are trained on professional voice actor recordings.
• KFC Global, KFC Canada Re-Creates the Voice of Colonel Sanders Using AWS AI Technology and Voicebot.ai, coverage of the Amazon Polly Brand Voice launch: the KFC Canada Brand Voice case.
• Microsoft News, Building Personalities Into AI: the Deborah Harrison quote.