Product Marketing Manager

What you'll actually do

• Build the source-of-truth substrate. Audit, organize, and unify what Sift knows about its products, competitors, ICP segments, sales motion, win/loss patterns, and customer language. Decide what lives where, what gets retired, what needs to be created, and what needs to be made machine-readable instead of just human-readable.
• Ship internal agents that move the needle. Identify the GTM workflows where an agent saves the most hours per week or unlocks the most consistency: competitive Q&A for AEs, first-pass launch briefs, draft messaging for new campaigns, RFP support and build them. Measure adoption and quality. Iterate.
• Equip the PMM team to operate at higher leverage. Help evolve the product marketing organization to focus less on hand-crafted deliverables and more on activities with strategic importance: sales enablement, message refinement, and researching the competitive landscape.
• Set evaluation standards. Define how we measure agent quality at Sift: accuracy, voice fidelity, helpfulness, hallucination rate. Make sure what we ship gets better over time, not just bigger.
• Be the team's AI tooling lead. Evaluate, recommend, and integrate new AI tools as the landscape shifts. Build the relationships and the muscle so we're early to what matters and skeptical of what doesn't.

Who we think you are

You're someone for whom the last two years have been a step function in what you can produce. You've built things for fun, for past jobs, on the side. Not "I asked ChatGPT to write a term paper" things. Real things: an agent your old team still uses, a workflow that saved you a significant time, a custom GPT that does one job exceptionally well.

Specifically:

• 3–8 years in product marketing, content strategy, sales enablement, marketing ops, growth, or an adjacent role where you produced GTM artifacts and felt firsthand where the work was repetitive
• You've shipped at least one internal AI assistant, whether Glean, Claude, custom GPT, or similar that other people actually used, and you can speak to the design decisions you made
• Comfortable with prompt engineering, retrieval design, and evaluation frameworks. You know why a system prompt isn't a JD, and why "looks good to me" isn't an eval
• Have evaluated and selected AI tools for a team before, including the unfun parts: vendor procurement, auditing outputs, and getting people to actually use the thing
• Fraud, trust & safety, or risk industry background strongly preferred. Time on the buy side, the sell side, or at a Sift competitor or adjacent player is a real plus
• You write well. You can hold a brand voice. You're not allergic to positioning, messaging, or launches, you just believe the leverage point has shifted
• You're comfortable being early. This role is being invented in real time, and your second month here will not look like your first

Bonus signals (not requirements):

• Light coding fluency: you can read and modify Python or JavaScript, are comfortable with APIs, and have wired things together via MCP, Zapier, n8n, or similar
• Specific experience with Anthropic's Claude: Projects, Skills, Agents, or the API
• You've written or spoken publicly about your AI work

What we're not looking for

• Someone whose primary excitement is owning a product line's positioning and writing its narrative deck.
• Someone who treats AI as a productivity hack instead of a craft. We want people who think about agents the way good PMs think about products.
• Generalist enthusiasm without shipped work. The bar here is "show me what you built," not "tell me what you've read."

How to apply

Send your resume and include a link to or write-up of something you've built. A Claude Project, a Glean assistant, a custom GPT, a screencast of a workflow, a GitHub rep. Give us anything where we can see how you think when you're designing for a user that isn't you. One real thing beats ten polished bullet points.

Applications without a "what I built" artifact will get less attention than those with one.