Workato published a fireside session with DMSi’s chief product officer, and the useful thing about it is the admission at the center. “I don’t think anybody has really figured out the pricing,” said Jason Niemi, who says he has tried to redo DMSi’s pricing about ten times. Each attempt runs into a customer base that spans five-user lumber yards to a company with 7,000 users: “This customer is going to go up by 300x. And then our biggest customer went down by half.” So DMSi, a 350-person ERP vendor with roughly 650 customers, is not starting with a price sheet. It is sending everything, API call or agent, through a single access point, then adding monthly usage allowances, then working out the price.
This is a customer story on the vendor’s own site, ending in a call to explore Workato’s product, and the figures carry that weight. The “10x cost reduction” from moving to open models is the claim of Preston Badeer, DMSi’s AI partner, and the “roughly 87% cheaper” is Badeer’s own estimate of Badeer’s own company’s offering. Neither is a benchmark. The sequence survives without them: control the plumbing, drive down the cost, then monetize. So does the operational advice. Start on a frontier model, work down through smaller ones with evals, treat tool definitions as prompt engineering, and catch doom loops, where “the agent gets stuck and can’t proceed,” with a hard cap on retries.
The catalog shows what that sequence looks like as a surface. Workato has 11 API pages, and the ones this story leans on are easy to name: the Embedded Partner API for a vendor putting Workato “in every user’s hands, whether they know it or not,” the Recipes API for the integrations DMSi is rebuilding, and the MCP Servers API, Tools API and Policies API for the agent-facing side. The catalog description agrees with the post’s positioning, recording Enterprise MCP, Genies and Workato Embedded layered on the iPaaS foundation. The order-entry agent that drafts an order and holds it for review has a parallel in the record as well: the agentic access profile marks 5 of 60 operations human-in-the-loop, with 40 acting.
Workato’s Kin Score is 65.7, strong, with developer ergonomics at 76.8 and contract quality at 70.6, and contract governance trailing at 45.5. Agent Readiness is 49.5, and it lights dimensions most of the market leaves dark: MCP server, delegated identity, protected resource metadata, dynamic client registration. For a platform selling a controlled access point, that is the right set to have lit. The unlit list is the one to read beside the doom loop advice: idempotency, reversibility and dry-run mode. A retry cap limits how many times an agent repeats a call. Only idempotency decides what each repeat does.