Mar 18, 2025 · 41m · a16z
Why AI Voice Feels More Human Than Ever
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the a16z Podcast, host Steph Smith and partners Olivia Moore and Anish Acharya analyze the rapid evolution of AI voice technology, exploring how recent technical breakthroughs in latency, tonality, and LLM intelligence are enabling transformative B2B vertical applications and deeply engaging consumer experiences.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 13% of the talking time here. How this is scored →
The host as informed peer 3.8 Guest teaching 4.9 Guest disagreement 1.3 The host pushing back 1.3
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 33:46 Anish dismisses incumbent AI voice efforts
Anish forcefully rejects the idea that tech incumbents can compete, declaring that products like Google Home utterly fail compared to modern LLMs and arguing big corporations are structurally incapable of shipping opinionated products.
Hardest push from the host ▶ 16:00 Steph pushes back on Gen Z location sharingSteph directly challenges Anish's assertion about consumer receptivity to tracking technology, stating she personally finds constant location-sharing incomprehensible.
Biggest teaching moment ▶ 14:17 Non-obvious candidate preference for AI interviewersOlivia educates the host on how candidates frequently prefer AI interviewers over tired human recruiters because the AI provides an unbiased, attentive evaluation.
The host holds their own ▶ 36:11 Steph introduces time-to-laugh metric for voiceSteph demonstrates sharp industry expertise by arguing that traditional search KPIs fail for voice platforms and proposing emotional engagement metrics like time-to-laugh.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Why Legacy Assistants Failed vs. Modern Engines | 3 | 4 | 1 | 1 | Steph opens by citing her personal habit of turning off Siri to ask why legacy voice assistants failed. Olivia and Anish explain how older architectures lacked underlying intelligence and personality compared to modern LLM engines. | |
| Phone Calls as Distribution Channels & Consumer Adoption | 2 | 4 | 2 | 1 | Anish reframes media narratives regarding consumer hesitancy toward AI voice, arguing that users adapt immediately once phone calls begin. Olivia highlights phone calls as a natural distribution channel for enterprise adoption. | |
| Technological Unlocks: Latency and Speech Interruption | 4 | 5 | 1 | 1 | Steph prompts a specific technical breakdown on latency benchmarks, asking what natural human speech latency is. Olivia educates on sub-300ms human thresholds and breakthroughs in interruption handling. | |
| Case Studies: NotebookLM and Sesame | 4 | 5 | 1 | 1 | Steph references data from the guests' report regarding YC founder activity. Olivia details how founders are shifting from horizontal voice engines toward vertical applications and workflow tools. | |
| Vertical SaaS Parallels & Disruption in Logistics | 3 | 6 | 1 | 1 | Anish and Olivia educate on vertical opportunities like high-value legal SKUs and logistics. Olivia highlights the counterintuitive finding that job applicants often prefer AI interviewers over tired human recruiters. | |
| Consumer Receptivity, AI Companions, and Passive Listening | 4 | 5 | 2 | 3 | When Anish uses Gen Z location-sharing to illustrate shifting privacy norms, Steph pushes back with personal disbelief. Olivia then details why AI companion apps provide consistent active listening that humans cannot match. | |
| Business Wedges: Overflow Calls and High-ROI Tasks | 4 | 5 | 1 | 1 | Steph frames the discussion around augmentation versus substitution strategies. Olivia outlines high-ROI wedges like overflow calls, credit card activation reminders, and administrative doctor office calls. | |
| Relentless Consistency and High Net Promoter Scores | 5 | 5 | 1 | 2 | Steph presses on failure modes and asks how AI pricing is evolving beyond basic per-minute models. Olivia explains shifts toward platform fees, per-seat SaaS, and outcome-based pricing models. | |
| Building Long-Term Competitive Moats in AI Voice | 4 | 5 | 1 | 2 | Steph questions whether the current AI voice land grab mirrors the cash-burning era of Uber. Olivia and Anish break down defensibility through vertical integrations, proprietary call data, and personal trust moats. | |
| The Consumer Voice Landscape and Incumbent Limitations | 4 | 6 | 3 | 1 | Steph asks if big tech incumbents will capture consumer AI voice opportunities. Anish strongly criticizes legacy incumbents like Google and Apple, arguing their corporate structures prevent them from launching opinionated products. | |
| Commodity Tasks vs. Independent Startup Opportunities | 6 | 4 | 1 | 1 | Steph introduces a novel framework that voice requires opinionated personalities and proposes metrics like time-to-laugh. The guests strongly agree, elaborating on how personality friction builds user trust. | |
| Advice for Founders: Execution Speed and High-Value SKUs | 3 | 5 | 1 | 1 | Steph prompts the guests for concluding founder advice. Olivia stresses execution speed as a primary moat, while Anish challenges founders to design extremely high-value, high-cost SKUs. |
Statements from this episode (34)
Acharya: AI voice models will reach full potential within 12 months
“Trust has to be earned, and if the models don't design for that, they're never going to get to their full potential. And I think we're going to see it in the next 12 months, not the next five years.”
Moore: People interact with modern AI voice as human or better
“It's in no way like a true conversational partner in a way that people are interacting with AI voice now like it is a human or in some ways even better than a human.”
Moore: Businesses will broadly adopt AI to replace traditional phone calls
“But also I think many businesses in a great way will impose it on them because you can now use AI to replace phone calls, which is so much more efficient and cost effective for them.”
Moore: Businesses are making tens of thousands of daily AI calls
“We see a lot of businesses that are already doing 1010 of thousands of phone calls with AI every day.”
Acharya: Consumers quickly accept voice AI when conversations feel human
“It's interesting because I think that talking heads want to tell you that people don't want to talk to an AI, but in all the cases where people do interact with an AI that starts a call by announcing, I'm an AI, people are like, oh, cool, let's just get into i…”
Moore: One-second AI voice latency is now considered too slow
“So this time last year, two to three seconds of latency was pretty good. Now, a second of latency is too long. Maybe even half of a second of latency is too long in many cases.”
Moore: Human conversational latency is under 300 milliseconds
“I mean, it's sub, definitely sub, like, 300 milliseconds. Sometimes even less than that if you have humans interrupting humans.”
Moore: Leading AI voice agents support bidirectional interruptions
“And you can have some of the most human-like voice agents that I've seen are capable of being interrupted by humans and also capable of interrupting humans, too.”
Moore: Natural AI voice interaction requires artificial pauses and vocal tics
“To an AI model, they will know exactly what words that they want to say back to you, right? So there's no reason for them to put in any pauses, any gaps, any little vocal tics. But to a human listener, very few humans just speak perfectly with no interruptions…”
Moore: NotebookLM sounds human by incorporating speech errors and filler
“Notebook LM is one example where that sounded so human because they put in all of these things that to an AI might feel like an error, but to a human, it sounds like another human talking.”
Acharya: Voice assistants should prioritize emotionality over intelligence to win consumers
“And I would argue even for the Alexis and series, like even if they didn't invest a lot more in intelligence and capabilities, if they over invested in emotionality, They might actually get a lot of the way there in terms of consumer experience. And yet I have…”
Moore: Up to 25% of recent YC startups build with AI voice
“So in recent YC cohorts, upwards of 20, 25% of companies are building with AI voice, which is really exciting.”
Moore: Building basic AI voice agents has commoditized
“And the next wave that we're starting to see is a lot more verticalized, and I think it makes sense because the ability to build a voice agent has kind of commoditized.”
Acharya: Businesses paying $100k for phone answering are prime AI targets
“Any business that pays a person a hundred, a 150 K a year to answer phone calls is a potential customer of voice AI and, you know, can lead to a really interesting vertical opportunity.”
Moore: Voice AI startups displace human labor costs rather than existing software
“When we talk to most voice agent companies, they aren't necessarily replacing existing software, but they're probably actually allowing businesses to either cut down on human labor or reallocate their human labor to kind of more effective things for the busine…”
Acharya: High-value AI voice agents will emerge within 12 months
“Like, what is the AI skew that gets paid thousands of dollars an hour to make a phone call? And I think we're gonna see it in the next 12 months, not the next five years.”
Moore: 45 publicly traded staffing companies conduct job candidate interviews
“Recruiting is one, so there's like, 45 publicly traded staffing companies that do interviews for, yes, blue collar jobs, but also engineering jobs.”
Moore: Job candidates often prefer AI interviewers over human recruiters
“And what we find is that a lot of candidates would actually prefer talking to an AI interviewer than talking to a human recruiter that maybe has to take 10 calls that day, is tired, is in a bad mood, doesn't really have the technical... Hasn't eaten lunch.”
Acharya: AI companions enhance human-to-human interaction rather than replace it
“Analog to this in AI is companionship and friendship, you know, which is a much broader concept than voice, though voice really brings it to life, and people say, hey, do people really want to be friends with an AI, and is that good for our society? And I thin…”
Acharya: People want AI friends and AI companions benefit society
“Do people really want to be friends with an AI and is that good for our society? And I think like yes and yes.”
Moore: AI voice companions offer superior availability and empathy to human friends
“Like, in many cases, the AI is more human than the human. For sure. Even your best friend, if you give them a call, they might be busy. They're at work. They're having a bad day. Are they actually gonna listen to every single word that you're saying and respon…”
Moore: Startups find success automating credit card activation calls
“We see, I've seen a couple voice agents that are really successful now with that use case alone.”
Moore: AI voice agents outperform humans on awkward sales upsells
“Maybe they have to make an upsell, and it's awkward, but they are not getting an extra commission for doing that, so they're gonna skip it 80% of the time. An AI will just do it every time, and will kind of do it proudly, and if they get turned down, you know,…”
Acharya: AI voice agents deliver lower costs and higher NPS
“The sort of intuitive thinking of a lot of the customers is like, well, it's lower price, but probably a lower NPS experience, and it's not. It's actually lower price and a higher NPS experience in many cases.”
Moore: Businesses expect 70% discount when replacing human agents with AI
“And then I would say this gets back to the constrained point, but even though the voice agent is still probably doing better than your human agents, most businesses don't want to pay that much for it because it is AI and they see it as a way to cut costs. So i…”
Acharya: Voice plus reasoning models will rapidly eliminate unwanted hallucinations
“This is where I think the capability is just going to get better and better faster than we appreciate. You know, with the language models, they're prone to hallucination, and there are certain conversations like the therapy one that benefit from the hallucinat…”
Moore: AI voice startups shift from per-minute pricing to platform fees
“The other thing about the price per minute model is it really just puts your value as a platform solely on the phone calls, which again are kind of commoditizing versus like the other software that you're building around the phone calls. So, I would say as a r…”
Moore: OpenAI won't build long-tail vertical software integrations
“It's not gonna make sense for OpenAI to go integrate with every long tail, you know, transportation management software that a fleet company is gonna, or a freight company is gonna be able to need to run their, you know, fleet of trucks on a voice agent produc…”
Moore: Enterprise AI voice providers build moats through onboarding call data
“It's gonna take months and months of training calls to make that better. And so you, as a voice agent provider, if you get in early, ah, benefit from having all that special proprietary data that just gives you months of a head start for anyone else who has to…”
Acharya: Winning AI voice markets will cost less than Uber's landgrab
“Yeah, I mean, it's certainly gonna be less expensive than Uber to go win the market, but yes, I mean, as Ben said many times, you have to both make a product people want, and then you have to go take the market, get from zero market share to all the market sha…”
Acharya: Tech incumbents are vastly behind LLM startups in voice AI
“I think that the incumbents, it's just such a daily demonstration of how far behind they are when you both have Google Home in your home and you've got ChatGPT in your pocket. You know, my children try to ask Google Home to tell them stories in the same way th…”
Acharya: Corporate risk aversion stops big tech from shipping opinionated voice AI
“Corporations, sort of committees, lawyers, like, these big companies have a hard time shipping opinionated products, at least opinionated in the way that many of these voice models may need to be, and startups have no problem doing that.”
Moore: Tech incumbents won't build the first AI-native personal assistant
“Are they gonna build the first AI native personal assistant that works across all of your products and all of your information sources? Probably not, I would say. I think that any and all of the calls that the incumbents end up doing, which will be some volume…”
Moore: Consumer AI voice products require conversational friction to retain users
“This is actually, this gets at what we're looking for in voice companion products, or, but even any consumer voice agent, like, there has to be some friction. If it's, like, too easy to build the relationship, if they're always saying yes to you, if they're no…”