Principal Product Designer, Agentic Chat

Principal Product Designer, Agentic Chat

You'll lead the design of how people experience agents inside GTM One, ZoomInfo's unified go-to-market platform for sellers, marketers, and revenue teams. In agentic chat, the agent's behavior is the experience, and you own that experience end to end: how it acts, how it asks, what it costs, and how it proves itself before a customer finds out otherwise. This is a Principal seat defined by surface ownership, not feature delivery — you'll set direction for agent experience across every way a customer reaches an agent, and you'll be expected to bring proposals forward rather than wait for a brief.

You will:

• Own the agent experience across universal chat, including how the agent communicates status, asks for input, and reports back on long-running work.
• Design the interactions that let agents create artifacts, apps, and automations built on context, and define how those outputs show up to the customer.
• Build cost estimation and credit-approval moments into chat, so customers understand what an agent action costs before and after it runs.
• Establish evals as a design instrument, using evaluation results to prove an experience is ready before launch and to guide what changes after real users touch it.
• Design the trust mechanics that let customers calibrate confidence in agent output, including how failure, uncertainty, and partial results are shown.
• Prototype in code, open pull requests, and ship working proposals directly into the codebase alongside engineering.
• Partner with product management and engineering on orchestration and tool infrastructure while owning what the customer actually experiences of it.
• Define problems alongside solutions, operating with thin briefs and acting as a product thinker on your surface area.

What You Bring
Must-Have:

• At least 2 years designing AI or agentic products where the model's behavior was the product itself (not AI features layered onto a conventional surface), and you can speak candidly to what went wrong in front of real users.
• Experience running evals: you have built or used an evaluation setup for an agent, and you can explain how you knew it was ready before launch and what changed after real users touched it.
• A track record of shipping in code: you prototype with AI coding tools (not only in Figma), you open pull requests, and you can read a backend contract, even if you are not an engineer by title.
• A working understanding of how LLMs behave in production, including latency, streaming, tool calling, confidence, and failure modes, and you design for those honestly rather than defaulting to the happy path.
• Experience working directly with engineering in their channels and on their timelines, as a core part of how you ship.

Preferred:

• Domain depth designing for sellers, marketers, RevOps, or CS inside a CRM, sales-engagement, data, or revenue-intelligence product.
• Experience with enterprise complexity, data-dense interfaces, bulk operations, or long-running jobs without losing clarity for the end user.
• Experience designing for a product where usage directly costs the customer money.
• Comfort blending UX research with agentic design — using research to predict and evals to confirm.

How We Evaluate
You'll be assessed across four axes, shared with you in advance: context engineering and agentic systems (including eval experience); the blend of UX research and agentic design; aesthetic quality and experience craft; and engineering fundamentals. No one is expected to score highest on all four — we want to know where you'd place yourself so we know how to work together. Interviews include walking through a real agent-surface decision (including an argument you lost), an eval you ran and what it changed, a live conversation about a go-to-market workflow, a working session on an open problem, and a look at your own tooling. No spec work, no unpaid take-home, and no design challenge we'd actually ship.

#LI-Remote
#LI-ST2