Builder Product Manager I

ECI Software Solutions is hiring a Builder Product Manager I focused on its Lumber & Building Materials products. The role translates ERP business logic and customer needs into AI-agent missions, prototypes, evaluation criteria, and priorities for modernizing legacy products.

Responsibilities:

• Work with the Industry PM and the Value Stream PM to guide you in overall strategy goals and key metrics for the applicable value stream are achieved
• Work with UX Research to define what "good" looks like when the interface is an AI agent, not a button
• You write the behavioral specs (tone, escalation logic, failure modes) for generative UI experiences talking to customers to validate what you have built
• You co-own the eval framework with the QA Builder, writing the "ground truth" definitions that judge agent output
• Use AI to run your own user research, synthesize support tickets, usage telemetry, and interview transcripts, along with UX research and information from Signal to help prioritize insight
• Replace traditional PRDs with structured "Mission Specifications", agent goals, tool constraints, success evals, and failure conditions that give Builders everything they need to ship without a kickoff meeting
• Interview subject matter experts and translate their knowledge into system prompts and agentic constraints that prevent hallucinations in workflows
• Co-build the eval datasets and rubrics that the QA Builder uses to verify agent output against real ERP business logic; not vibes
• Turn a "what if we automated X" into a working proof-of-concept prompt chain fast enough to validate before the squad commits to a build cycle
• Use AI-assisted analysis of usage data, support volume, and codebase complexity to continuously prioritize which legacy modules the squad attacks next based on overall strategy objectives

Requirements:

• The PM will need to be curious about the workflows end to end
• You use LLMs to synthesize user research, generate acceptance criteria, and stress-test your own assumptions before a single line of code is written
• You can specify and pressure-test the behavior of autonomous agents
• You understand what LangGraph state machines do, and you can write the "Mission Brief" that constrains one correctly
• You aren't afraid of 20-year-old ERP workflows
• You understand that "the PM said it was fine" isn't a guardrail
• You define measurable acceptance criteria, quantitative evals, and the ground-truth definitions that let engineers prove correctness
• You can take a 50-page accounting regulation or a decade of tribal knowledge from a SME and with the help of the Value Stream PM, turn it into structured system prompts, agentic constraints, and eval rubrics the squad can build against
• You don't just define features; you define how correctness is measured
• You have a physical reaction to 40-page PRDs nobody reads
• If a requirement can't be expressed as a testable agent instruction, you question whether it's real
• You don't think in "features"; you think in "capabilities."
• You have already used an LLM to spec, critique, or prioritize your own work
• You find ERP business logic genuinely interesting
• You understand the basics of how customer workflows work and leverage the Value Stream PM or other experts to drive why it matters and what breaks when it doesn't
• You know the difference between a compelling AI demo and an agent that's safe to run on a workflow