Lead AI Workflows Product: Strategy, Prototyping & Outcomes

Mural; United States; Remote; USD 150.000 - 210.000; Vollzeit

Zusammenfassung

Mural is seeking a seasoned Product Manager to own AI Workflows from idea to value proof. You will partner with the Engineering Manager and the Agentic Design Director to define strategy and drive end-to-end experiences with engineers, ML engineers, designers, and product managers.

You will prototype, test, and decide what ships based on user data and evals. You will collaborate with growth, finance, legal, and security to align shipping decisions, set up data loops, and measure usefulness,

Qualifikationen

• 7+ years in product management, including shipping at least one AI product from concept to real users.
• Recent hands-on experience building agentic or LLM-based products, including what went wrong and what you changed.
• Technical fluency to design and read evals, and to work as a peer with engineers and ML engineers on context management, latency, reliability, and failure modes.
• Sound judgment on model decisions: comparing models, routing across them, adopting new ones, and deciding when fine-tuning or a custom model is worth it.
• Understanding of AI unit economics, including cost to serve, usage-based metering, and how model choices change margin.
• A record of launching, rebuilding, changing direction, and killing ideas based on evidence
• Comfort making high-stakes decisions with incomplete information, and being candid about the tradeoffs.
• Experience owning outcomes while partnering with expert engineering and design leaders as true counterparts.
• Experience creating structure where no playbook exists.

Aufgaben

• Set strategy for AI Workflows together with your engineering and design counterparts, own the outcomes and the evidence behind what we build, and engage deeply in the end-to-end user experience
• Find the team workflows where AI adds clear value, at the intersection of collaboration, alignment, and decision making.
• Build working prototypes with modern AI tools (Claude, Cursor, Codex, or similar) to test ideas with users before committing further investment.
• Turn each release into a hypothesis with pass/fail criteria, and make the call on whether to iterate, scale, or kill it.
• Define product-quality evals with engineering that measure usefulness, trust, and reliability, and own the call on whether an experience meets the bar to ship.
• Decide with engineering and ML when to use frontier models, when to route across models, when to bring in a new model, and when a model of our own is worth building.
• Understand the cost to serve each workflow, and work with growth and finance on how usage is metered and what the unit of value is.
• Set up the data loops that improve quality, including what we collect, how we label it, and how it feeds evals and model work.
• Design agentic experiences that work in real multiplayer sessions, where people need to understand, steer, and correct what the agent does.
• Talk with users every week and bring what you learn back into the work.
• Work with design, engineering, marketing, sales, legal, and security teams to align on shipping decisions.
• Share what you learn across the product organization so other teams can adopt the practices that work.

Kenntnisse

Product management

AI products

Agentic/LLM-based products

Evals design

Model decisions

AI unit economics

Launch history

High-stakes decisions

Cross-functional partnership

Agent interoperability

MCP interoperability

Collaboration software

Customer-embedded product work

Tools

Eval tools

Jobbeschreibung

Mural is seeking a seasoned Product Manager to own AI Workflows from idea to value proof. You will partner with the Engineering Manager and the Agentic Design Director to define strategy and drive end-to-end experiences with engineers, ML engineers, designers, and product managers.

You will prototype, test, and decide what ships based on user data and evals. You will collaborate with growth, finance, legal, and security to align shipping decisions, set up data loops, and measure usefulness,