a16zThe State of Consumer Tech in the Age of AI
CHAPTERS
- 0:00 – 0:29
Why consumer AI breaks out unpredictably (and why this cycle looks different)
The panel sets the stage: consumer tech is hard to predict, and AI is creating new kinds of breakout products that don’t neatly resemble the last era’s social apps. They frame today as early innings where product outcomes can emerge “out of nowhere,” and where execution speed matters unusually much.
- •Consumer is inherently unpredictable; breakthrough products often appear suddenly
- •AI is in an early era where iteration speed can be a key advantage
- •AI’s “peak value” may ultimately be enabling better human connection
- •The conversation will compare past consumer cycles (social/mobile) to AI’s trajectory
- 0:29 – 3:18
From Facebook-to-TikTok breakouts to the AI wave: what counts as a consumer winner now?
Erik contrasts the prior 15–20 years of iconic consumer breakouts with the feeling that new paradigms slowed down. Justine and Bryan argue that ChatGPT is a major consumer breakout, and that image/video/audio AI tools are also winning—just with different dynamics than classic social networks.
- •ChatGPT as the standout consumer outcome of the last few years
- •Rise of modality-specific AI tools (image, video, audio) like Midjourney/ElevenLabs
- •AI innovation was initially research-driven, often lacking strong consumer product layers
- •A new opportunity emerges as models mature and become accessible via APIs/open source
- 3:18 – 5:55
Defensibility and monetization: why AI consumer can charge $200/month (and still grow)
The group debates whether today’s AI leaders are durable or merely early “MySpace-era” winners. They highlight a major shift: consumer AI products can monetize immediately at high price points, and revenue retention can outpace user retention through upgrades, credits, and overages.
- •AI consumer monetization differs sharply from prior $50/year subscription norms
- •High-end SKUs ($200–$250/month) change the need for classic network-effect moats
- •Different models are “pointy” and used for different tasks, supporting pricing power
- •Revenue retention vs. unique user retention diverges due to upgrades/usage-based spend
- 5:55 – 8:00
Consumer spending shifts: software eats entertainment, creativity, and relationships
They argue consumer discretionary spend is increasingly “subsumed by software,” with AI doing real work (research, content creation, personalized media). AI products can justify high prices by saving time and enabling new forms of entertainment and expression.
- •AI tools replace time-intensive work (e.g., automated deep research)
- •Generative video/creative tools feel like a ‘magical mystery box’ for consumers
- •Entertainment and creative expression become software-mediated and paid subscriptions
- •Prediction: future consumer budget categories trend toward “food, rent, software”
- 8:00 – 12:25
AI-native social networks: why ‘AI Instagram feeds’ feel wrong and what might replace them
Bryan and Justine explore why a truly AI-native social graph hasn’t emerged yet. They critique current attempts as skeuomorphic (copying old feeds with bots), noting social needs emotional stakes and authenticity—hard to maintain when content is infinitely generated.
- •Classic social evolved via status updates → photos → short-form video; AI needs a new leap
- •ChatGPT can know users deeply; sharing that ‘essence’ could create new connection modes
- •Most AI-driven social behavior still happens on legacy platforms (Facebook/Reddit/Reels)
- •AI-generated perfection lowers emotional stakes, undermining social dynamics
- 12:25 – 12:59
Synthetic profiles and ‘AI LinkedIn’: when your profile contains what you know
They speculate that future social platforms may shift from static credentials to interactive knowledge representations. Instead of profiles pointing to what you know, people could interact with a synthetic version of you to access your ideas and expertise directly.
- •AI-native LinkedIn concept: profiles that embed knowledge rather than link to it
- •Possibility of talking to a ‘synthetic you’ to extract wisdom or context
- •Reframes social interaction as conversational access rather than content feeds
- •Hints at new identity, reputation, and discovery mechanisms
- 12:59 – 17:23
Enterprise adoption via consumer virality: the ElevenLabs playbook
The panel describes a pattern where consumer buzz becomes enterprise lead generation—sometimes before mainstream consumer adoption. ElevenLabs is used as the key example: early meme/creator usage translated into large enterprise contracts as companies raced to adopt AI.
- •Consumer virality can precede mainstream adoption but still drive enterprise demand
- •Enterprises monitor AI Twitter/Reddit/newsletters to source tools and ideas
- •AI mandates inside companies accelerate willingness to buy ‘consumer-looking’ products
- •Tactical growth loop: detect workplace clusters of users (e.g., via payments) and sell in
- 17:23 – 20:42
Moats in early AI: ‘velocity is the moat’ (and what network effects look like so far)
Bryan reframes classic defensibility: in early AI markets, speed of shipping and model/product iteration drives mindshare, usage, and revenue. They discuss how true social network effects are still nascent, but “marketplace-like” data advantages (e.g., voice libraries) are emerging.
- •Traditional moats still matter, but early winners often win by moving faster
- •Velocity drives mindshare → traffic/users → revenue → ability to keep iterating
- •Closed-loop creation→consumption→social network effects aren’t fully formed yet
- •Early compounding advantages appear via assets like voice/character libraries (ElevenLabs)
- 20:42 – 23:05
Voice as the next major interface: from a failed substrate to an AI primitive
Anish explains why voice is newly viable: it’s fundamental to human interaction but historically didn’t work well as a tech platform. With generative models, voice becomes a first-class primitive, opening both consumer experiences and broad enterprise transformation.
- •Voice has always mediated human interaction, but prior voice tech (e.g., VoiceXML) fell short
- •Generative AI makes voice finally usable as a core interaction layer
- •Early consumer promise: always-on coach/therapist/companion in your pocket
- •Surprise: enterprises adopt voice quickly, even in sensitive regulated industries
- 23:05 – 23:48
AI in the most important enterprise conversations: negotiation, persuasion, sales
They push beyond the idea that AI voice is only for low-stakes customer support. Anish argues the highest-leverage business conversations—sales pitches, negotiations, relationship-building—will be AI-intermediated because models can outperform humans in persuasion and consistency.
- •Common misconception: AI voice is only for low-stakes support calls
- •Claim: AI will mediate the most important business conversations
- •AI advantage in negotiation, persuasion, and structured selling
- •Sets up broader implications for roles, workflows, and business outcomes
- 23:48 – 31:51
AI clones for learning and personal development: from Masterclass agents to ‘scaling yourself’
The discussion turns to personalized learning and advice via synthetic versions of real people. Examples include Delphi-style clones and Masterclass voice agents that let users ask targeted questions, making expert knowledge more accessible than long-form content.
- •AI clones could extend beyond celebrities to ‘everyday experts’ and undervalued talent
- •Masterclass experiment: turn recorded courses into interactive voice agents via RAG
- •Short conversational learning may beat long-form attention requirements
- •Open question: prefer a known person’s clone or a fully synthetic ‘perfect match’ mentor?
- 31:51 – 38:47
AI companions go mainstream: vertical companions, emotional support, and real-world outcomes
They argue companionship is a foundational LLM use case, spanning therapy/coaching to NSFW relationships and niche “vertical companions.” The panel emphasizes that many users simply want someone to talk to, and that well-designed companions can improve mental health and even help people build better real-world social skills.
- •Companionship was an early mass-market LLM behavior—even when products weren’t designed for it
- •Vertical companions expand the definition beyond ‘AI girlfriend’ to nutrition, coaching, and more
- •Story: Character.AI user credits AI girlfriend with helping him find a real-life partner
- •Caution: overly agreeable AI may harm users’ real-world relationship skills
- 38:47 – 43:22
New AI platforms, devices, and social norms: always-on assistants, AirPods, and recording culture
In closing, they speculate about hardware and ambient AI: devices that see/hear what you do, act on your behalf, and provide ‘human insight’ benchmarking and recommendations. They also discuss the emerging (and contentious) norm of recording conversations, predicting new etiquette and cultural rules will form around always-on AI.
- •AI may stay phone-centric, but local/on-device models could enable privacy-first experiences
- •Ambient AI: pins, screen-aware agents, and always-on assistants that can take actions (emails, tasks)
- •AirPods as the post-phone device hiding in plain sight—limited by social protocol today
- •Recording norms will evolve; context and culture will determine what becomes acceptable
