Agentic Operations Architect

Gusto, Inc.•Las Vegas, NV

3d•$168,420 - $247,369•Hybrid

About The Position

As an Agentic Operations Architect, you will own the reusable capabilities that make agentic service workflows possible, the patterns, primitives, standards, and tooling that other builders across the org use to automate discrete customer problems. You are not the owner of any single workflow. You are the reason a workflow that took six weeks takes two, and the reason the tenth one is safer than the first. You will prototype workflows yourself, but as a means to an end: to prove a pattern is sound, to find out what builders are missing, and to earn the right to set a standard. The mission of our Service Transformation (CXT) team is to quickly evolve and scale how Gusto delivers outstanding service experiences and leverages AI to do so. Our team designs the systems, governance frameworks, and feedback loops that allow Gusto's service model to improve continuously, operate reliably, and earn the trust of the customers and employees who depend on it.

Requirements

8+ years in operations, product, or AI roles — with at least 2 years designing and deploying AI-assisted or agentic workflows in production
Hands-on experience with agentic tooling, orchestration frameworks, or MCP-style integrations, including multi-agent coordination — conceptual familiarity is not sufficient at this level
Strong fluency with LLM systems: prompt design, evaluation, and how models actually behave at scale
Experience designing human-in-the-loop systems with clear escalation logic, override controls, and compliance-aware checkpoints
Ability to define and diagnose workflow metrics — solve and assist rate, containment, escalation quality — and to tell a real regression from noise
Proven track record influencing Product, Engineering, Legal, Security, and Compliance in writing, in design reviews, and in high-stakes forums
Familiarity with Responsible AI practices, governance, and risk management in production
Prior experience in a function being actively redefined — you have built the plane while flying it before

Nice To Haves

Experience with RAG architectures and retrieval quality optimization
Ability to assess whether underlying data structures, quality, and governance can support reliable agentic performance
Experience in SaaS, fintech, or regulated environments where compliance constraints are real
Exposure to AI governance or safety-oriented review processes

Responsibilities

Design the reusable patterns for agentic service workflows, orchestration, HITL protocols, override and escalation logic, quality safeguards, so builders assemble from proven components
Set the standards builders work to, concrete enough to apply without you in the room
Prototype workflows selectively: to validate a pattern, expose a gap, or unblock a hard problem, then extract the reusable part and hand it off
Pull learnings from every build back into the shared patterns, so each one is faster and safer than the last
Influence Product on what to build next: turn what's slowing builders down into specific, prioritized asks backed by evidence
Prototype around gaps the platform doesn't cover yet, proving a capability is worth building before USP invests
Be honest in both directions about the distance between what builders need and what exists
Rank automation opportunities by volume, risk, and customer impact, mining SOPs, operational metrics, and interaction data, to determine what gets built and where builders go
Work with Data to define how workflow performance gets measured, so every builder measures the same way
Identify where there are processes which exist which should not be automated
Bring what's at stake for customers into Legal, Security, Compliance, and platform reviews, so requirements land in the patterns instead of getting retrofitted per workflow
Represent how the portfolio is performing to senior stakeholders, briefly, and in a way that drives action