Jun 6, 2025 · 43m · a16z
The State of Consumer Tech in the Age of AI
Justine Moore · 13m spoken Bryan Kim · 9m spoken Olivia Moore · 6m spoken Erik Torenberg · 5m spoken Anish Acharya · 5m spoken
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, venture partners Justine Moore, Bryan Kim, Anish Acharya, and Olivia Moore analyze how generative AI is transforming consumer technology, monetization models, social networks, and human relationships. Hosted by Erik Torenberg, the panel highlights why distribution velocity, high user willingness to pay, and new interaction modalities like voice and companions define the current AI software landscape.
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 14.5% of the talking time here. How this is scored →
The host as informed peer 3.7 Guest teaching 4.9 Guest disagreement 1.9 The host pushing back 2.6
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 30:51 Anish rejects the human-preference argument for AI music
Anish forcefully rejects the idea that AI music fails due to human bias, arguing that AI generates mid music because models are fundamentally averaging machines while culture is built on the edge.
Hardest push from the host ▶ 18:45 Erik challenges velocity as a moat using the Gingerbread StrategyErik directly challenges Bryan's assertion that product velocity equals a moat by citing Ben Thompson's Snapchat Gingerbread Strategy framework to ask if incumbents will simply copy features.
Biggest teaching moment ▶ 30:51 Anish outlines the theoretical limits of generative models in cultureAnish provides a fundamental theoretical critique of AI model limitations, explaining that models trained on historical data could never infer new cultural shifts like hip-hop because culture requires inputs outside training data.
The host holds their own ▶ 18:45 Erik introduces tech strategy framework on feature velocityErik demonstrates deep domain knowledge of tech strategy history by countering the guest's thesis with Ben Thompson's Snapchat Gingerbread Strategy.
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 |
|---|---|---|---|---|---|---|
| Has Consumer Tech Innovation Stopped or Shifted? | 3 | 5 | 2 | 3 | Erik asks if consumer innovation stopped after the era of mobile giants. Justine and Bryan reframe the question, explaining that AI innovation is driven by research teams rather than consumer product teams, and that underlying model updates differ from stable platform cycles. | |
| Business Models and Revenue Retention in AI | 4 | 6 | 3 | 4 | Anish counters traditional assumptions by pointing out that AI models command high monthly prices upfront. When Erik pushes back on product retention versus business model quality, Bryan educates on how AI subscription metrics split unique user retention from revenue retention. | |
| High Willingness to Pay for Value-Driven AI | 2 | 5 | 1 | 2 | Erik asks how to explain consumers paying $200 per month for software. Olivia and Justine clarify that AI products replace actual human labor hours rather than providing passive entertainment. | |
| Searching for the First AI-Native Social Network | 3 | 5 | 2 | 2 | Erik asks what will drive a net-new social connection platform. The panel breaks down why current AI social attempts feel skeuomorphic and lack the real emotional stakes required for social networks. | |
| Synthetic Profiles and Knowledge-Based Networks | 5 | 4 | 1 | 3 | Erik offers a thesis on people recommendation and brings up the counterintuitive trend of enterprise adoption preceding consumer adoption. Justine and Bryan illustrate this dynamic using 11Labs' consumer virality leading into enterprise contracts. | |
| Capability Frontiers and Market Segmentation | 6 | 4 | 2 | 4 | Erik tests Bryan's claim that velocity is the new moat by bringing up Ben Thompson's Snapchat Gingerbread Strategy. Bryan and Justine defend their view by highlighting quality frontiers, model iteration speed, and workflow lock-in. | |
| Voice as an AI Primitive and High-Stakes Workflows | 3 | 6 | 3 | 2 | Anish reframes the prevailing enterprise voice narrative, arguing that AI voice will handle high-stakes negotiations and sales rather than just low-stakes customer support. Olivia highlights how fast enterprises adopted voice to replace high-turnover call centers. | |
| Digital Clones, AI Personas, and Customized Learning | 3 | 5 | 1 | 2 | Erik asks why users would regularly talk to synthetic clones. Olivia and Justine illustrate how interactive voice agents turn passive linear content like Masterclass videos into custom, bite-sized learning conversations. | |
| AI Influencers, Content Creation, and Fictional Stars | 4 | 4 | 1 | 2 | Erik prompts the panel on whether future top stars will be real humans or synthetic creations. Justine provides a clear framework distinguishing human-experience celebrities from interest-based AI creators. | |
| Cultural Edge, AI Music, and Model Training Limits | 4 | 6 | 3 | 3 | Anish firmly asserts that AI music feels mid because models average existing data whereas true culture exists at the edge outside training data. Justine refutes negative companion stereotypes with an example of Character.ai helping a user develop real-world social skills. | |
| Next-Gen Hardware, Ambient AI, and Wearables | 4 | 4 | 2 | 2 | Erik steers the discussion toward hardware form factors and shifting recording norms. The panel explores ambient AI hardware options like AirPods and smart pins alongside emerging cultural etiquette. |
Statements from this episode (36)
Bryan Kim: AI has demolished traditional product velocity constraints
“We're living in this early era of AI where velocity is demolished.”
Anish Acharya: Consumer tech spend will expand to food and rent
“In the future, you're gonna see consumer span to be like food rent software.”
Justine Moore: AI research teams historically struggle with consumer product design
“AI is still relatively early, and so much of the new products and innovation has been driven by research teams who are, like, so good at training models, but historically have not been amazing at creating the consumer product layer around them.”
Bryan Kim: An AI-native social graph has not yet been built
“What I think is missing potentially is connection. Like this social graph, this thing hasn't rebuilt on AI yet, and That may be just a white space or something that we just continue to see what develops there.”
Anish Acharya: Google's top consumer subscription tier costs $250 per month
“The top Google consumer SKU is 250 dollars a month.”
Anish Acharya: AI foundation model companies are raising prices, not lowering them
“But different people use them for different things, and it seems like they're raising prices, not lowering them.”
Bryan Kim: AI consumer apps see higher revenue retention than user retention
“So you actually see revenue retention being meaningfully higher than unique user retention, which again, like, I haven't seen that before, so.”
Olivia Moore: Pre-AI consumer app subscriptions averaged around $50 per year
“I think before, the average consumer subscription was maybe 50 dollars a year, if that. Like, that was kind of a lot. Like, the best-in-class consumer products would charge that.”
Olivia Moore: Consumers happily pay $200 monthly for AI applications
“And now we have people very happily paying 200 dollars a month and even saying in some cases that they feel like they're being undercharged for that, or they would pay more.”
Olivia Moore: AI tools justify $200 monthly fees by replacing manual work
“Now with products like Deep Research, for example, that could replace 10 hours of generating a market report by yourself, and so that kind of thing is easily worth, I think for many people, 200 dollars a month, even on like one or two generations.”
Justine Moore: Users pay $250 monthly for AI video generation tools
“I mean, I think things too, like VO three, like, people are paying 250 a month, and I'm happy to pay that because it's like you have this suddenly, like, it feels like a magical mystery box that you can, like, open it and get whatever video you want, only for …”
Anish Acharya: AI models will intermediate entertainment, creativity, and relationships
“So I think all the entertainment is being subsumed by it. A lot of the sort of creative expression work that you would do outside of software is now being subsumed by it. A lot of the sort of relationship intermediation, which might have been a place for dispo…”
Bryan Kim: ChatGPT knows more personal context about him than Google
“I pour my heart and soul into ChatGPT. It knows more about me than probably Google, potentially. Which is an insane thing to say. Like Google, I've been using, using Google for a decade plus, and ChatGPT may know more about me than Google, because I type more,…”
Justine Moore: AI social sharing still occurs on legacy platforms
“So far, the social behavior that has come from, like, the AI creative tools, largely, but also things like ChatGPT, is still happening on the existing social platforms, and not in the new AI platforms.”
Olivia Moore: AI social networks fail because they lack real emotional stakes
“To work, a social network has to have, like, real emotional stakes. And if you can generate the content in a way that you like it, and you always look amazing, and you always look happy, and you're always in a cool background, like, it doesn't have the same se…”
Anish Acharya: LinkedIn points to user knowledge rather than containing it
“LinkedIn is a pointer to what you know, instead of actually containing what you know.”
Anish Acharya: Future social platforms will feature knowledge-rich AI avatars
“And with this tech, we can create a profile that actually contains what you know, so I can talk to you synthetic, you know, ET and get all of your wisdom. Perhaps that's what future social looks like as well.”
Justine Moore: Consumer AI virality directly drives enterprise lead generation
“There's an initial consumer virality moment, and then that actually leads to lead generation in, Enterprise sales in a way that we did not see with the last generation of products.”
Justine Moore: Image generation market will support multiple persistent winners
“So, like, in image, for example, there's not just one best image model. There's, like, Best image for designers. There's best image for photographers. There's best image for people who can only pay 10 dollars a month versus the people who can pay 50 or a hundr…”
Olivia Moore: Enterprises are adopting AI voice to replace offshore call centers
“Real enterprises have picked up voice so quickly to replace human beings on the phone or to augment what human beings are doing on the phone. Even in really kind of sensitive and critical categories like financial services, because previously they were using o…”
Anish Acharya: Voice is the primary AI insertion point for the enterprise
“It feels like voice is the AI insertion point for the enterprise period.”
Anish Acharya: AI will handle high-stakes business negotiations and sales pitches
“The most important conversation that happens in a business in a given day, week, year is going to be intermediated by AI because AI will just do a better job with the negotiation or the sales pitch or the persuasion or the friendship.”
Justine Moore: Consumer AI clones will expand to everyday real people
“But I think we'll start filling in everything in the middle that's not just like a character, a fictional character, not just like a human thought leader, but like all of the real people in between.”
Olivia Moore: MasterClass launched RAG voice agents trained on instructor content
“Masterclass launched kind of an interesting beta where they take people who have already recorded courses on the platform and turn them into voice agents, where then you can ask questions that are really specific to you, and from my understanding, it basically…”
Justine Moore: Digital creators will split into human stars and AI personas
“I, my take is probably there will be fragmentation into, like, two types of creators or celebrities. One type is like a Taylor Swift type where, like, the human experience of it, I think, matters in some ways. Like, a lot of people not only love her song, but,…”
Olivia Moore: Making high-quality AI movies takes as long as traditional filming
“Which is like, yes, anyone can generate art now easier than ever before, but it's still, Takes an enormous amount of time to make great AI art. Like, we hosted an event with a bunch of AI artists last summer, and many of these people, when they walked you thro…”
Olivia Moore: Top-tier AI creator talent will have very low conversion rates
“So I think we're gonna see something similar happen where we're gonna have pools of AI talent and pools of human talent, and the very best of each is gonna rise to the top, and it's gonna be a really low conversion rate on both, which is probably how it should…”
Anish Acharya: AI music feels mediocre because models average historical data
“I think a lot of music, the problem is that the music that the AI generates is it just feels very mid. You know, and definitionally, these things are averaging machines, and culture is supposed to be at the edge.”
Anish Acharya: AI models cannot invent new musical genres like hip-hop
“If you trained, you know, a model with all the music, you know, up until, but just prior to hip hop, would it, like, infer hip hop? I don't think so, because music is the intersection of past music and culture is critical to it. So you sort of need something t…”
Justine Moore: Expect more specialized vertical AI companion apps
“We're gonna see even more vertical companions moving forward.”
Bryan Kim: Gen Z averages just over one close friend
“There is a very clear trend of average Number of friends that you can talk to over time going down. I think the youngest generation is something above one.”
Olivia Moore: Studies show Replika AI app reduces depression and anxiety
“Actual studies were showing depression and anxiety and kind of suicidal ideation were going down in users.”
Anish Acharya: Overly agreeable AI companions degrade real-world social skills
“Highly agreeable AI does not set you up well for that. So I think there's sort of a fine balance between being just agreeable enough to help you, like, engage and get better at this, Versus being so agreeable that you're actually worse at this.”
Olivia Moore: Tech-focused youth are increasingly adopting wearable AI pins
“Now when I go to tech parties, like, a lot of the under-twenties are wearing pins, That record what they're saying and doing, and they find, like, real value from them.”
Anish Acharya: AirPods are the most widely adopted post-smartphone device
“I mean, the device that has been most widely adopted post phone is the AirPods.”
Olivia Moore: Cultural norms will emerge around wearable AI recording devices
“But I think that's why there'll be, like, a new set of cultural norms. Like, when the cell phone was introduced, like, there's places where it's rude to take a loud call. Like, the same set of things will emerge around these recording devices.”