In early 2025, MultiCare Health System in Tacoma, Washington pulled the plug on fourteen Moxi robots it had deployed across its hospitals. Nurses complained they got in the way. Administrators said the economics didn’t stack up. The program was shuttered.
There is a quiet erosion happening inside organisations that have embraced artificial intelligence most enthusiastically. It is not visible in
Real estate AI splits clearly into two distinct markets that barely overlap agent & brokerage tools (lead gen, CRM, staging, listing content) and Investment & commercial tools (data, underwriting, portfolio management, lease abstraction).
AI is commanding capital and headlines, yet employers from carmakers to hospitals report they cannot find enough people to perform
Most AI systems are extraordinarily adept at finding patterns. What they cannot do, by mathematical necessity, is tell you what causes what. Multiple causal structures can produce identical data. More compute does not resolve that. More data does not either. This restriction is not a flaw in the technology. It is a fundamental limit on what passive observation can tell you.
The Creative Writing niche has a clear hierarchy. At the top are prose specialists like Sudowrite for generating quality fiction text. Below that are structural tools like Novelcrafter and World Anvil for organizing complex story worlds. Then screenwriting tools with their own formatting conventions. And finally editing tools that refine what’s been written.
E-commerce AI is one of the most saturated and fast-moving niches: almost every workflow from first click to post-purchase has 5–10 competing tools. The most important distinction is whether a tool is built for e-commerce specifically (like Gorgias or Prisync) versus a general AI tool applied to e-commerce (like Jasper or ChatGPT). The former almost always wins on depth; the latter wins on flexibility.
Metrics are often “low-resolution photographs of a high-resolution reality,” and when a measure becomes a target, it ceases to be a good measure. The danger is that the public will lose the judgment needed to evaluate whether the system is serving them because the “cognitive infrastructure” of struggle and confusion is being optimized away by technocratic skill.
They’re B2B/prosumer tools targeting students, researchers, and academics: a smaller, quieter market compared to the consumer AI boom. They don’t have the marketing budgets or hype cycles of OpenAI/Google/Anthropic, and they rarely make splashy product launches.
