Durable by design: why we invested in Orkes, the agentic workflow orchestration platform enterprises depend on
Enterprises run on workflows. Loan originations, payment processing, patient onboarding, order fulfillment are all examples of core business functions that need to execute correctly every time. Many workflows span multiple steps and systems, some running millions of times a day, others for days on end. They stitch together APIs, databases, and third-party services with little room for failure.
The infrastructure holding these workflows together is more fragile than most organizations want to admit. When something breaks, the cost shows up as delays, manual intervention, and real revenue loss. As systems get more distributed and workflows more dynamic, orchestration is becoming critical infrastructure.
Orkes is building the infrastructure that holds it all together. We were impressed by Orkes’ vision when we first met their CEO Jeu George last March. A year later, AVP is proud to have led Orkes’s Series B.
From open source to enterprise
Nearly a decade ago, developers at Netflix built Conductor to manage the growing complexity of their microservices architecture. They open sourced it soon after and adoption followed organically. Today, Tesla, Oracle, JPMorgan, American Express, and GE Healthcare are among the 3,000+ global enterprises running mission-critical systems on top of Conductor. The original creators of Conductor, Jeu George, Viren Baraiya, and Dilip Lukose, saw an opportunity to take that foundation and make it enterprise-ready. Orkes was founded in 2021 to fulfill that vision with a fully managed, scalable platform built for production from day one.
Most orchestration tools embed workflow logic directly in the codebase. That makes workflows more difficult to see, adjust, and debug. Orkes has taken a different approach, decoupling workflow logic from the code, pulling it into a separate orchestration layer. Developers still build the task workers, but the workflow itself – how systems interact, in what order, and under what conditions – lives outside the code. Teams can trace each step in real time and pinpoint exactly where something failed. Orkes also ships the enterprise features customers require, including access controls, SSO, secrets management, audit logging, and flexible deployment.
Agents make the problem impossible to ignore
Flexible deployment is what makes agents powerful but it also creates an operational challenge . At production scale, agents need infrastructure that handles long-running stateful execution, keeps moving when components fail, and still tells you what happened. Orkes is built for that. Its durable execution model keeps agents on track even when individual components don’t behave, holding state across retries, delays, and partial failures so work picks up where it left off instead of restarting from scratch.
The practical expression of that model is Agentspan, Orkes’ open-source agent runtime. When a developer kicks off an agent, Agentspan compiles the definition into a durable workflow on the server, so the process can die without the agent dying with it. Every running agent carries an execution ID, meaning status checks, approvals, and mid-flight interventions can come from any machine without losing the thread. Teams currently using a variety of agent frameworks don’t need to rewrite anything, they just hand the agent to Agentspan’s runtime and get crash
recovery and full execution history on top. That same runtime governs everything else the platform touches: internal APIs become agent-accessible tools through the MCP gateway, fixed steps and LLM-driven decisions run inside the same workflow, and human reviews slot in as native tasks that resume cleanly once input comes back.
What now?
AVP led Orkes’ Series B with participation from Prosperity7 and continued support from existing investors including Battery Ventures, Nexus Ventures, and Vertex Ventures. The company plans to use the capital to continue investing in its platform for agentic workflows, expand its engineering team, scale go-to-market, and grow its international presence.
The Conductor open source community, maintained by Orkes and in production at thousands of organizations worldwide, is one of the company’s most durable advantages. Enterprises running Conductor have already standardized on the model Orkes is now bringing to market in a fully managed, enterprise-ready form.
As orchestration becomes a critical part of the infrastructure modern distributed systems depend on, the companies that get it right will have a foundation they can build on with confidence. We believe Orkes is building that platform. We’re excited to partner with Jeu, Viren, Dilip, and the entire Orkes team.
