Staff Product Manager, Agentic Quoting Tools

Xometry (NASDAQ: XMTR) powers the industries of today and tomorrow by connecting the people with big ideas to the manufacturers who can bring them to life. Xometry’s digital marketplace gives manufacturers the critical resources they need to grow their business while also making it easy for buyers at Fortune 1000 companies to tap into global manufacturing capacity.

We are hiring a Staff Product Manager to own that experience end to end. You sit between Sales and the quoting teams and you serve the customers who matter most to Xometry. Your job is to rethink how a request becomes a priced quote: every channel it can arrive through, every hand-off between a person and a system, and every step that does not need a human at all. You will design the agents that read the request, build the quote, and propose the price, the Model Context Protocol (MCP) integrations that let those agents work across our quoting systems and Enterprise Resource Planning (ERP) platform, and the staff tools where people review what the agents did and decide what they should not. The measure is simple: how fast, how accurately, and with how little effort our best customers get a quote they can act on.

This role sits in the Pricing and Manufacturability (P&M) group alongside the product managers for cost prediction, pricing, and Design for Manufacturability (DFM), and reports to the head of P&M product.

You will be given a metric and a problem space, not a fully scoped backlog. Comfort defining your own roadmap from ambiguous signals is a requirement, not a bonus.

This role requires a hybrid schedule (3 days a week) at our Waltham, MA or North Bethesda, MD office location.

How You'll Contribute

Own the request-to-quote experience for our most important customers. Map every path a quote request takes today, from the moment a customer emails, texts, uploads, or sends an agent, through Sales, quoting, pricing review, and back to the customer. Know the behind-the-scenes steps better than anyone outside the teams that run them. Remove the steps that do not need to exist. Your scorecard: quote turnaround time, cost per quote, quote accuracy, and quote-to-order conversion on the human-assisted path.

Redesign the funnel, not just the tools. Treat each intake channel as a pathway with its own drop-offs and its own fix. Decide where a request should be read by an agent, where it should be routed to a person, and where the customer should be able to self-serve without either. Build the pathways so a request arrives structured, complete, and already half-priced by the time a human sees it.

Make agents do the work people should not have to. Own the roadmap for agentic quoting: agents that read a request and its files and build the quote, assistants that propose margin and lead time, and automation that promotes human-priced parts into the instant-pricing engine over time. Decide what the agent does, what the human confirms, and what stays human. Define "good" and "failure" for every AI feature before engineering starts, and hold that bar through launch and after.

Connect the agents to the business through MCP and the ERP. Own the integration layer that lets agents and people exchange work across Salesforce, our quoting systems, file storage, and the ERP: Model Context Protocol (MCP) servers, tool definitions, permissions, and the audit trail that shows who (or what) changed a quote and why. The exchange between humans and agents is a product surface; design it like one.

Build the staff tools the quoting and sales teams actually use. Own the internal quote review, pricing, and quote summary tools. Treat adoption as the product metric: a tool the team routes around has failed, regardless of what shipped.

Instrument everything. Every human pricing decision (override, margin change, lead-time adjustment) becomes structured data that improves the next prediction and the next agent run. Write your own queries, build your own dashboards, and put the events you need into the requirements document, not into an analyst request.

Run discovery continuously, not on a research calendar. Sit with quoters and account managers. Read the inbound email and text queues. Watch session recordings. Synthesize the signal with AI tools and validate it against live customer and staff conversations before acting on it. A working prototype, tested with the people who will use it, is your default artifact for a build decision.

Set product strategy and secure buy-in. Define a 6 to 12 month roadmap for agentic quoting, sequence it by customer impact and speed, and align Engineering, Data Science, Sales, Quoting Operations, and Finance behind it. Present progress against named outcomes to technical and non-technical audiences.

Build and ship with AI agents. Operate spec-first. Read the relevant systems with agentic tools, ask sharper feasibility questions as a result, and write specs that an engineer can hand to a coding agent. Prototype in Cursor, Claude, and Figma Make before you write requirements.

Raise the bar for peers. Share the AI workflows that worked and the ones that failed. Your patterns should show up in other product managers' artifacts.

What You'll Bring to Xometry

You would be a good fit with:

8+ years of product management experience, including direct ownership of a workflow or operations product that sits between a sales or service organization and the customers it serves, in a B2B, marketplace, or enterprise context.

Evidence of shipping AI or automation into an expert human workflow with guardrails, and of knowing when not to automate.

Experience integrating products across systems of record (CRM, ERP, file storage, messaging), and a working understanding of how agents connect to those systems through tool protocols such as MCP.

Multi-channel thinking: you have redesigned how requests enter a business (email, text, forms, APIs, agents) and can show the drop-off you removed.

AI-native practice, not just AI tool use: you can show where AI changed a decision your team made, not only where it saved you time. Hands-on fluency with Cursor, Claude, or equivalent agentic tools, and a point of view on evaluating large language model (LLM) features before and after launch.

Strong analytical fluency: you write your own SQL, you have built your own dashboard, and you have caught a broken metric before it misled a decision.

Operational curiosity: you learn how a process actually runs by sitting with the people who run it, and you earn their trust before you change their tools.

Exceptional written communication: your documents let an operations lead, an engineer, and an executive act on the same page.

Comfort making decisions under ambiguity, translating a loosely defined problem into a testable hypothesis and a plan.

Bachelor's degree in Business, Computer Science, Engineering, or a related field.

Signals we look for

You can walk through a specific request-to-fulfillment funnel you redesigned across a sales team, a service team, and their customers: the channels involved, the steps you removed, the before and after numbers, and how adoption went.

You have built or owned an integration between an AI agent and a system of record, and you can explain what the agent was allowed to do, what it had to ask a human for, and how you knew when it was wrong.

You have shipped an AI or machine learning (ML) feature into a human-in-the-loop process and can describe its evaluation criteria, its failure modes, and what you monitored after launch.

You can point to a case where your team decided differently because of an AI-driven workflow you built (continuous customer signal, prototype-first decisions, evaluation criteria written before build).

You have worked in manufacturing, supply chain, pricing, or quoting, or you can show how you learn an unfamiliar operational domain fast.

Pace

This is a startup pace inside a public company. You will