Internal Automation Vulnerable to AI
Zachary Kirby, co-founder of Vessel, on building the Vercel for integrations
AI is most likely to squeeze the low stakes workflow layer before it displaces the reliability layer. Internal automation tools win by letting a team describe a task like sync leads, enrich a record, or route a ticket, then fixing the occasional mistake in house. Customer facing integrations are different, because the product vendor owns the failure when data is wrong, auth breaks, or a sync lags, which keeps demand high for code level platforms built around repeatable APIs, ETL, webhooks, and tight control.
-
The practical split is who absorbs errors. In an internal workflow, an ops team can catch a bad field mapping and rerun the job. In a customer facing integration, the end customer sees the broken Salesforce or HubSpot sync immediately, so the integration itself becomes part of product quality.
-
That is why older leaders like Zapier and Workato started in internal automation, with drag and drop workflows for non technical users. Vessel is positioning around the opposite need, developer owned integrations embedded in product code, where teams need fine grained control over auth, rate limits, scopes, refreshes, and edge cases.
-
AI still expands the market. OpenAI function calling makes it easier for models to invoke tools and external systems, which raises the value of clean public APIs. At the same time, Zapier and Workato are already packaging AI agents on top of their automation layers, showing where internal orchestration is heading fastest.
The next phase is a stack split. Internal automation platforms will become more conversational and agent driven, while customer facing integration platforms become deeper infrastructure, with better schemas, observability, sync guarantees, and API abstractions. As AI increases the number of software actions companies want to automate, both categories grow, but the reliability premium on external integrations grows fastest.