LLMs Threaten Niche Productivity Apps
Mike Knoop, co-founder of Zapier, on Zapier's LLM-powered future
LLMs compress a whole category of narrow workflow software into a feature inside a larger assistant. If an app mainly takes plain language, calls Gmail or another API, and returns a faster way to sort or draft messages, then Zapier, ChatGPT, or the inbox itself can absorb that job. The surviving products are the ones that either own the main work surface, or provide deep infrastructure that many assistants rely on.
-
Zapier was already exposing 5,000 plus apps through a natural language actions layer by 2023, turning requests like find my latest email into a direct Gmail action. That makes thin wrappers around integrations easy to replicate from inside a general purpose assistant.
-
The real split is interface owner versus capability provider. Mike Knoop describes developers getting pushed either toward owning the front end where users spend time, or toward being the back end service that powers many other products. Commodity apps stuck in the middle get aggregated away.
-
Email triage can still be a business, but only when it expands beyond a single shortcut. Fyxer grew to an estimated $9M ARR in May 2025 by pairing inbox triage with drafting, scheduling, meeting context, and team deployment, while also spreading across Gmail, Outlook, calendar, and notes to avoid getting replaced by one inbox feature.
This pushes productivity software toward two endpoints. One endpoint is bigger assistants and work hubs that bundle many simple tasks into one surface. The other is specialized systems with richer context, stronger data access, and repeat workflows that are hard for a generic assistant to copy. The middle layer of single use wrappers keeps shrinking.