Poolside sells embedded AI engineering
Poolside
Poolside is trying to sell a managed outcome, not a model endpoint. The important shift is that revenue can come from embedding engineers, wiring the system into repos and CI/CD, tuning agents around a customer’s codebase, and then staying involved to keep reliability and ROI high. That makes the contract look closer to a high end technical services engagement with software attached, than a standard per seat coding assistant deal.
-
The workflow is unusually hands on. Poolside says forward deployed research engineers assess use cases, embed with product and platform teams, integrate into data sources and pipelines, ship agents and developer tools, then operate them with observability and adoption programs. That is classic implementation revenue plus recurring optimization revenue.
-
This model fits the buyer. Poolside targets large enterprises, including regulated environments, where code cannot leave the customer boundary and deployments run on prem or in a private VPC. In those accounts, the hard part is usually not buying seats, it is getting security approval, connecting internal systems, and proving measurable gains in production.
-
Comparable labs show the spectrum. Cohere makes most revenue from private deployments with multi year enterprise contracts, while Mistral layers co development and professional services on top of platform subscriptions. Poolside is pushing further in that direction by staffing deployments directly around coding workflows, which can support much larger contract values than standalone IDE tools like Windsurf or Cursor.
If this works, AI coding vendors will split into two lanes. One lane sells cheap, self serve coding copilots. The other sells an AI engineering team inside the enterprise boundary. Poolside is building toward the second lane, where deeper integration, longer deployments, and ongoing model operations can turn each customer into a much larger and stickier account.