Consumer AI is attracting attention, but its business model still leans heavily on subscriptions and usage charges. Andreessen Horowitz partner Olivia Moore sees a chance to broaden how these products make money—and to build AI tools for parts of daily life that remain largely untouched.
Consumer AI needs more ways to earn revenue
Moore’s report on the top 100 consumer AI apps found that ChatGPT remains far ahead of other players, while Suno and ElevenLabs have shown staying power. The list also reveals gaps: several familiar consumer categories have no entrants among its top 100.
That matters because current AI revenue is concentrated in subscriptions and token usage, which Moore says tend to be more common on the enterprise and prosumer side. She agrees that OpenAI has moved toward enterprise, but describes that as an expansion rather than a retreat from consumer products.
The question is whether consumer services can make money in ways that do not depend only on people paying out of pocket. Moore is interested in a model where users can access a product for free with ads, then choose to subscribe to remove them. That could make a service available to people who would rather not start with a paid plan.
Lower-cost models could fit more everyday uses
Running AI services costs more at the margin than running established internet services such as Facebook or Google Search. Moore says costs are improving, with cheaper models and open-source options increasingly in the mix.
Not every task needs the most capable model. Some consumer uses may work well with lower-cost systems, especially when the model itself is only one component of a product. Moore points to founders building on open-source models and says the change is beginning to show.
Today, however, many of the users generating revenue rely on coding and technical automation. Those tasks may require frontier intelligence, which makes them a different economic case from consumer apps where AI works behind the scenes. As more companies build for broader consumer needs, lower-cost models could become more common.
Many “consumer” apps serve prosumers first
AI has also blurred the boundary between consumer and enterprise products. Moore cites companies such as Gamma, ElevenLabs, and Cursor, which began with consumers and became majority-enterprise businesses within 18 months.
She argues that much of what gets called consumer AI is better understood as prosumer AI: products bought by individuals but used for work. Her examples fall into three groups:
- Product-building tools, including Lovable, Replit, and Fal.
- Product-marketing apps, such as AI ad generators Higgsfield and HeyGen.
- Work-management services, including Manus, Fireflies AI, and Granola.
These products may be purchased by individuals at first, but their work-oriented uses set them apart from older consumer internet services. That distinction helps explain why revenue can appear strong while products still serve a relatively specialized audience.
Everyday categories remain open territory
Moore’s report identifies social apps, dating apps, marketplaces, retail, travel, finance, and health as categories without entrants in its top 100 list. Their absence is striking because they represent areas of ordinary consumer activity, rather than work tasks or technical automation.
Filling those gaps could give AI a broader consumer footing. It would also test whether companies can build useful products for people who are not buying tools for work, and whether those products can support revenue models beyond subscriptions and token charges.
Moore expects more activity in these areas over the next six months. Her view is that the category is still at an early stage: the current leaders and revenue patterns may reflect the first wave of AI products, rather than the full range of services consumers will use.