Marketers dump what they no longer want in AI SaaS companies

Investors have been pouring billions into AI companies over the past few years, as the technology continues to take off in the Valley and thus the world. But not all AI companies are capturing the attention of investors.

Indeed, while it seems that every company these days is reinventing itself to have “AI” in its name, some startup ideas are no longer favorable to investors. TechCrunch spoke with VCs to learn what investors are looking for in AI startups as a service.

Popular SaaS groups that are investing now include AI-based architecture startups, data-driven SaaS solutions, systems (that help users complete tasks), and platforms focused on core business needs, according to Aaron Holiday, managing partner at 645 Ventures.

But he also offered a list of companies that are considered boring to investors today: Startups making smaller parts, horizontal hardware, lightweight inventory management, and surface analytics — basically, anything an AI assistant can do.

Abdul Abdirahman, an investor at the F Prime company, added that any kind of software “without any kind of data” is no longer popular, and Igor Ryabenky, the founder and director of AltaIR Capital, went into depth on this point. He said investors aren’t interested in anything, really, that doesn’t have a lot of depth.

“If your differentiation is in the UI (user interface) and the system, it’s not enough,” he said. “The barrier to entry has come down, making canal construction more difficult.”

New companies entering the market now need to create “real ownership of the process and fully understand the problem from day one,” he said. “Large codebases are no longer an option.” The most important thing is speed, focus, and the ability to change quickly. Prices should also be flexible: fixed models for each seat are difficult to protect, while used containers make sense in this environment. ”

Jake Saper, general partner at Emergence Capital, also had thoughts on ownership. For him, the difference between Cursor and Claude Code is “the canary in the coal mine.”

“One has a way of working, the other just does the work,” Saper said. “Developers are becoming more and more a decision maker rather than a strategy.”

He said anything related to “forced service” – meaning trying to get as many customers as possible to use the drug continuously – could be found to be at risk if agents take over.

“Pre-Claude, getting people to work within your program was a powerful strategy, but if agents are doing the work, who cares about how people work?” He told TechCrunch.

He also thinks integration is slowing down, especially since Anthropic’s model context protocol (MCP) makes it easier to connect AI models to external data and systems. This means that one does not need to download multiple plugins or create their own plugins; they can only use MCP.

“Being a connector was like a tunnel,” Saper said. “Soon, it will be useful.”

Also, what is no longer new includes “workflow systems and management tools that make the coordination of human tasks unnecessary if, in the long run, the assistants just work,” said Abdirahman, citing examples, especially SaaS companies whose stocks are at the bottom where new AI startups start with better and more efficient technology.

Ryabenky said SaaS companies that are struggling to scale right now are the ones that can easily be emulated, he said.

“Generic design tools, project management software, basic CRM clones, and thin AI wrappers built on top of existing APIs fall into this category,” he said. “If these products often lack deep integration, proprietary data, or aggregated knowledge, powerful AI teams can quickly rebuild them. This is what makes investors wary.”

Beyond that, what remains attractive about SaaS is depth and sophistication, with tools integrated into complex workflows. He said companies should focus on integrating AI more deeply into their products and changing their products to reflect that, Ryabenky continued.

“Investors are also investing in businesses that have services, data, and technology areas,” Ryabenky said. “And far from things that can be copied without much effort.”