Why Building Consumer Companies Has Been So Difficult

The primary challenge for consumer companies over the past decade has been the closing of viable distribution channels. In the early 2010s, emerging platforms like Facebook and Instagram offered organic growth opportunities with reasonable customer acquisition costs. I remember when I first started as a PM at Blue Apron in early 2014, Facebook was just emerging as a wide-open paid marketing channel with manageable CACs.

There were brief moments of opportunity – the podcast boom, for instance – but these windows have largely closed. Before that, mobile as an emerging platform helped propel a generation of consumer companies. Since the mid-2010s, however, building de novo consumer businesses has become exceedingly difficult due to saturated distribution channels and skyrocketing acquisition costs.

What AI Brings to the Consumer Table

Despite these challenges, the current generation of AI presents several intriguing opportunities in the consumer space:

1. Multimodal Interaction Experiences

Today’s AI isn’t limited to language models – it encompasses audio and image generation as well. This multimodality opens doors to entirely new interaction paradigms that go beyond the mere utility of content creation.

We’re starting to see examples of this with apps like Character.AI, where users develop ongoing relationships with AI personas, and Midjourney, which has created a community around collaborative image generation. These hint at the possibilities, but I believe we’re just scratching the surface of truly novel interaction models.

I envision game-like, novel forms of digital interaction where AI models serve as conduits for engaging one-on-one or group experiences. While admittedly abstract, I’m thinking about fundamentally new user experience paradigms or content units that could catalyze fresh consumer behaviors.

2. Distribution Through “Wow Factor”

Historical consumer breakouts have relied on insanely viral behavior – products so compelling that users can’t wait to share them. ChatGPT exemplifies this in the AI space as OpenAI became an “accidental application layer company” with 300M monthly uniques (granted, this is a mix of consumer and professional users). Its success stemmed from positively shocking users with capabilities that wildly exceeded expectations about how human-like a machine could sound.

We’ve seen this pattern with other early consumer AI applications like Lensa AI’s Magic Avatars, which generated a massive wave of social sharing when users discovered they could create artistic renditions of themselves. Similarly, AI voice cloning tools like ElevenLabs captured attention by demonstrating uncanny capabilities that users eagerly shared.

This suggests a path forward: consumer AI applications that create genuine surprise and delight might overcome distribution challenges through sheer virality. However, as we’ve seen with apps like Clubhouse and BeReal, initial virality isn’t enough, and retention is harder than ever. After novelty attracts users, the product must prove sustainably useful or entertaining to retain them.

3. The Next Frontier of Personalization

We’ve been discussing personalization in commerce for two decades – from Amazon’s early recommendation models to countless startups promising to match users with perfect products. Yet progress has been surprisingly limited.

Current attempts like Stitch Fix use a combination of human stylists and algorithms, but even they haven’t fully cracked the personalization code. Pinterest has leveraged visual AI to improve discovery, but the gap between understanding user preferences and consistently delivering value remains substantial.

The challenge isn’t just algorithmic – it’s about machines truly understanding users while balancing privacy concerns and minimizing the friction of explicit user inputs. Today’s AI applications show promise as general-purpose, chief-of-staff-like utilities (e.g. our portfolio company M1) that could, as a side effect, develop deep understandings of users and their life contexts.

This suggests two promising avenues for consumer AI:

Two Primary Opportunity Spaces

Opportunity 1: New Forms of Consumer Entertainment

The first opportunity revolves around broadly defined new types of consumer entertainment. These would compete for attention minutes currently dominated by social media giants like Meta and TikTok. AI could enable entirely new entertainment formats that feel more personal, adaptive, and engaging than today’s social feeds.

We’re seeing early versions of this with apps like Replika, which offers AI companions, and AI-powered storytelling apps like NovelAI. The recent phenomenon of AI “parasocial relationships” – where users develop emotional connections with AI characters – suggests untapped potential for deeply personalized entertainment experiences.

Opportunity 2: Transaction Matching

The second opportunity space lies in transaction matching – not just better personalization, but essentially next-generation marketplaces. Instead of aggregating supply and demand in the traditional marketplace model, AI could act on behalf of the demand side, deeply understanding what each consumer needs based on personal context, then proactively finding the appropriate supply.

Some early examples include Perplexity, which simplifies information discovery by directly answering questions rather than requiring users to sift through search results. In e-commerce, companies like The Yes (acquired by Pinterest) have attempted to build shopping experiences that learn user preferences over time to present only relevant items. 

Today this might be reactive (a consumer asks for X, and the agent finds the best X), but tomorrow it could be proactive, anticipating needs before they’re explicitly expressed.

The feasibility of this approach depends on whether supply attributes are locked up or if there are opportunities to use generative workflows to bypass traditional supply aggregators. Take Airbnb as an example – finding accommodations still requires hours of searching because marketplace matching remains imperfect. Could AI radically reduce this friction?

Conclusion

I believe we need more experimentation combining different modalities and interaction methods. The very fact that I’m creating this post using a voice note rather than typing demonstrates how new input methods can make tasks not only significantly easier, but also more engaging and fun.

There’s vast territory to explore in how we interact with AI models in household and consumer contexts. I’m particularly excited to see what founders might create when they think beyond the enterprise and focus on bringing AI’s capabilities into our daily lives in ways that truly enhance, surprise, and delight.

I’m particularly bullish on AI’s potential to break through the distribution deadlock that has stymied consumer innovation. The combination of novel experiences, viral delight, and unprecedented personalization might finally give consumer technology the tools it needs to thrive again. If you’re a builder tinkering in this arena, I’d love to meet and jam on ideas.