The Opening in AI Software
June 29, 2006
The AI market has a strange shape right now. Usage is everywhere, but dependency is still thin. Edison Research and SSRS reported that more than half of Americans were using AI tools weekly in early 2026, while Bank of America Institute found that only about 3 percent of households were actually paying for AI services. That gap is not a small detail. It is the market telling us that people are interested in AI, but most have not yet found a product that becomes necessary.
That should bother anyone watching this market closely, because the technology itself is obviously powerful. People are not rejecting AI. They are using it, testing it, sharing it, asking it questions, and folding it into pieces of their day. What they are not doing, at scale, is treating it like infrastructure. They are not yet saying, “I need this every month the way I need my phone, internet, banking, payments, calendar, or cloud storage.”
The reason is that most AI products still feel like places you visit rather than systems you depend on. You open a window, ask a question, get an answer, and leave. The interaction may be useful, but it usually does not carry forward. It does not become the connective layer across the person’s day. It does not remember enough, connect enough, or reduce enough of the burden of managing a modern digital life.
That is where the consumer opportunity sits. The individual market is not a side market or a toy market. It may be the cleanest place to prove what AI software is supposed to become, because the buyer is closest to the pain and the buying decision is simple. There is no procurement cycle, no internal committee, no implementation office, and no long change-management exercise. There is just a person deciding whether the product makes life easier enough to pay for it.
That market is large enough to matter. In the United States, roughly 236 million adults own smartphones. At $100 per month, a foundation-level AI service for that population represents about $283 billion in annual platform potential at full penetration before any marketplace or domain-specific services are added. The exact penetration rate can be argued, but the size of the opportunity cannot be dismissed. If AI becomes a base layer of daily life, the consumer market is not an afterthought. It is the first proving ground.
The mistake would be treating this as another app category. The better comparison is infrastructure one layer above the internet connection. The internet connection lets devices communicate. The next layer has to help meaning move across the person’s digital life. That is what today’s tools do not do. They answer questions, generate text, summarize information, and automate pieces of work, but they do not yet hold the person’s context in a way that compounds over time.
Small business is the natural next step because it is the same problem with real money attached to it. Credence Research estimates the AI market for small and medium businesses at $194.6 billion in 2024, growing to $567 billion by 2032. The owner is already managing work across disconnected tools, and there is no department to absorb the mess. If today’s mostly shallow AI use is already producing meaningful savings, that does not mean the problem is solved. It means the larger opportunity is still sitting underneath the way the business actually runs.
That is why the path from consumer to small business can move quickly. Once a foundation proves that it can organize the individual’s context, the same foundation can support work, money, customers, commitments, projects, and knowledge with a different set of services attached to it.
Enterprise comes later because the structure is harder. Large companies have the same underlying problem at a much greater scale. Their data is spread across departments, permissions, platforms, teams, and workflows, and they are now trying to make AI operate across all of it. Cloudera and Harvard Business Review Analytic Services reported that only 7 percent of enterprises say their data is completely ready for AI.
Enterprise is already showing the same pattern, just at a larger scale. The data is spread across too many systems, the definitions do not line up, and every new AI layer has to spend part of its effort reconstructing what the business should already know. That is why the response has become so expensive. Enterprises are not only paying for intelligence. They are paying to assemble context over and over again on top of data structures that were built for SaaS and not AI.
Enterprise AI keeps getting stuck because pilots are controlled and businesses are not. Once the system moves into daily operations, it has to deal with current data, permissions, definitions, accountability, and trust at the same time. When that context is not already part of the foundation, the company pays to rebuild it every time AI is asked to do real work. Workato named the pattern directly: AI is often bolted onto existing workflows instead of built into how the business operates.
This is the opening in AI software. The money has poured into models, chips, agents, and enterprise experiments, but the product that makes AI necessary across daily life and work is still largely missing. The market is using AI, small businesses are adopting it, and enterprises are trying to govern it. What they are all reaching for is the same thing, even if they describe it differently.
They are reaching for a foundation that makes AI useful beyond the moment.
