Engineers write code faster with AI-native development. That's accurate and widely observed. The bigger leverage comes from pairing it with a reusable integration domain specific language (DSL) — so the speed compounds across every integration Abnormal ships, not just the next one.
Instead of treating every new SaaS integration as custom work, the team defines endpoints, auth, pagination, and ingesters as YAML specs over a generic DSL engine, and carries the same pattern through config mapping, infrastructure, and datapacks. AI makes the first draft cheap. The DSL makes the first draft reusable.
AI Across the Workflow
Abnormal's integrations team doesn't use AI only for coding. The team uses it across the workflow: shaping 1-pagers, accelerating TDDs and test plans, generating and validating first-pass implementations. For DSL work specifically, AI helps produce the first draft of the code, the tests, and the iteration loop around both, while the DSL ensures the result is reusable rather than trapped inside a one-off integration.
Why It Compounds
In a few months, that approach covered more than 85 integration requests spanning Office 365, Google Workspace, Okta, Workday, ServiceNow, Salesforce, OpenAI, Claude, Copilot, and more. New platforms now ship at roughly 5-6x the prior rate. With the same team, Abnormal supports 8 or more product areas without treating each new integration as a one-off investment.
Each new integration builds on the last. That's what compounding at machine speed actually looks like.
See the latest from Abnormal's product and engineering teams.
