The bar has moved
Investors have poured money into AI, but not every AI startup gets attention. In a March 2026 TechCrunch piece, reporter Dominic-Madori Davis asked venture investors what they are no longer looking for in AI software as a service companies. Their answers draw a clear line between products that are easy to copy and products with real depth.
This post summarizes what they said and what it means for anyone building or choosing AI tools.
What is falling out of favor
Aaron Holiday of 645 Ventures listed categories investors now find uninteresting: thin workflow layers, generic horizontal tools, light product management and surface level analytics. In short, anything an AI agent can now do.
Abdul Abdirahman of F-Prime added generic vertical software without proprietary data. Igor Ryabenkiy of AltaIR Capital went further. "If your differentiation lives mostly in UI and automation, that's no longer enough," he said. "The barrier to entry has dropped, which makes building a real moat much harder."
Ryabenkiy named the categories struggling to raise: generic productivity tools, project management software, basic CRM clones and thin AI wrappers built on existing APIs. If a product is mostly an interface layer without deep integration, proprietary data or embedded process knowledge, he said, strong AI native teams can rebuild it quickly.
What investors still want
Holiday listed the categories that remain popular: AI native infrastructure, vertical software with proprietary data, systems of action that help users complete tasks, and platforms deeply embedded in critical workflows.
Ryabenkiy said new companies need "real workflow ownership and a clear understanding of the problem from day one." He added that speed, focus and adaptability now matter more than large codebases, and that rigid per seat pricing will be harder to defend than pricing based on usage.
Execution over process
Jake Saper of Emergence Capital offered a sharp framing. He compared two coding tools and said one owns the developer's workflow while the other simply executes the task. "Developers are increasingly choosing the execution over process," he said.
His broader point is that getting people to spend their day inside your software used to be a strong advantage. "If agents are doing the work, who cares about human workflow?" he asked. Abdirahman made a similar point about tools that coordinate human work, which may matter less if agents simply complete the tasks.
Integrations as a utility
Saper also argued that integrations are losing value as a moat. With standards like the Model Context Protocol making it easier to connect AI models to external systems, he said, "Being the connector used to be a moat. Soon, it'll be a utility."
What this means for builders
Taken together, the investors are describing a shift from software that helps people do work to software that does work. A few practical lessons follow.
Own an outcome, not a screen. Products that complete tasks hold up better than products that only organize them.
Go deep on a problem. Understand the domain well enough to encode real process knowledge.
Build on data and context. Products that learn from the specific data of the people using them are harder to copy.
Price for value delivered. Usage or outcome based pricing fits a world where agents, not seats, do the work.
A note on context
These are the views of a handful of investors, shared in one article. They reflect where venture money is flowing, which is not the same as what every customer needs. Plenty of teams still rely on project tools and simple automation every day. Still, the direction is worth taking seriously. When the people funding new software say that execution is beating coordination, it signals where the next generation of products will compete. Builders and buyers alike should expect tools to be judged more on the work they finish than on the time people spend inside them.
Questions to ask about your own product
If you build software, a few questions follow from the article. Could a capable team rebuild your product quickly? Does it rely on data or process knowledge that others lack? Does it finish work, or does it only help people organize work? Honest answers point to where to invest next.
What this means for teams buying tools
The same lens helps buyers. When you evaluate AI software, ask whether it actually completes work or just adds another interface. Ask whether it fits your real processes and can use your own context. And ask whether it lets you combine tools rather than adding another island.
Where Innflow fits
Innflow is an all in one workspace for AI agents and workflows. The shift these investors describe, from coordinating tasks to executing them, is the idea Innflow is built around. Instead of tracking work for people to do, teams use Innflow to set up agents and workflows that do the work across their connected tools, with people approving the steps that need judgment.
The investors' critique is also a useful check for any product, including ours. Value comes from agents that complete real tasks inside a team's own processes and context, not from another layer of screens. That is the standard worth holding every tool to.
Focus on what gets done
The message from investors is clear. Interfaces and simple automation are no longer enough. What lasts is depth, data and the ability to get real work done. If you want to see what it looks like when agents and workflows handle work across your tools, try Innflow at innflow.ai.
Source: Dominic-Madori Davis, TechCrunch, "Investors spill what they aren't looking for anymore in AI SaaS companies". Read the original.












