Consumer-facing AI products are attracting renewed interest with the unexpected success of Meta’s personal assistant Muse and mascot Jolly, along with the rise of OpenAI’s newly released Dots and Instinct. Instinct, which handles tasks such as booking travel, making restaurant reservations, and canceling subscriptions, has reached a $10 billion valuation. These developments support the view that agentic AI has become reliable enough for everyday tasks and offers users real value.
However, while consumers’ willingness to pay for AI remains limited, it is unclear whether better models create a more profitable consumer business. As a result, the industry is turning toward the Anthropic model, which focuses on enterprise contracts and different verticals. Although Muse and Instinct appear to fall outside this trend, the fundamental economic problems of consumer AI remain.
According to the State of Markets report, based on Andreessen Horowitz’s PNC research, 2.2% of consumers were paying for AI services as of May, with average monthly spending of $31. Although adoption and spending are increasing, the growth has been linear. Even the major performance leap from GPT-5.2 to Astra was not clearly reflected in the data. Bank of America found in March that approximately 3% of U.S. consumers were paying for AI, representing a 40% increase from the previous year. Menlo’s September survey found that one-quarter of adults used AI every day, and that half of these users paid for it.
The main problem is operating costs rather than revenue. AI is expensive compared with previous technologies such as social networks or cloud computing; even hundreds of millions of paying users do not guarantee reaching the break-even point.
OpenAI has adapted to these conditions by turning toward the enterprise market; its enterprise bookings have reportedly doubled since July. Dots was also introduced with use cases aimed at software engineers and agency workers. Meta has more revenue options for Muse thanks to its personalized advertising infrastructure and is also exploring the enterprise market. Instinct plans to take a share of purchases made through its agent and may avoid the cost of training a frontier model. Nevertheless, there are limits to growth in consumer AI without enterprise revenue.
Why it matters
This picture shows that increased usage in consumer AI does not, by itself, translate into a sustainable business model. While users’ budgets for services remain limited, having agentic systems perform tasks exposes companies to high operating costs. As a result, product success is increasingly measured not only by model performance but also by non-subscription revenue channels and opportunities for enterprise use. OpenAI’s focus on enterprise customers, Meta’s advertising infrastructure, and Instinct’s plan to take a share of transactions reveal different responses to the same economic dilemma. The open question is whether these channels can generate enough revenue to cover the costs of consumer products and how much the paying user base will expand.