Product Manager, Agentic Products

The Opportunity:

Most companies are adding AI to the software their teams already use. We're building it the other way around.

At Jetson, AI agents already do real operations work. They check rebate paperwork, plan the next steps on a project and draft replies to customers. People review that work, approve it and correct it, and every correction makes the agents better.

That changes the product question. It's no longer "what screen does this team need?" It's "which work should an agent do, which should a person do, and how do they hand it off?"

You'll own that answer for the people doing the work. Your customers are Jetson's own operations teams: rebates, project coordination, system design, customer success, billing, technical service and field operations. Each of them is moving onto one shared platform where people and agents work the same queue. You'll decide what each team sees, what the agents propose, what people approve, and how we know it's working.

The goal is aggressive: take a sale process that typically takes the industry 30+ days and make it three days from paying a deposit to installation. Every workflow you ship gets a home off fossil fuels sooner.

The best outcome is work nobody at Jetson has to do: an agent does it right, or the step disappears.

What You Will Do:

• Bring operations teams onto the platform, one at a time. Sit with a team, map how their work really runs, decide what goes in their first release, and get it adopted. Rebates and project coordination come first.
• Design how people and agents share the work. For each type of task, decide whether an agent suggests it, a person approves it, or it runs on its own. Define what "done" means and what proof shows it.
• Graduate work to autonomy with evidence. Every new agent action starts as a suggestion. When people accept it nearly every time, it moves to one-click approval, then to fully automatic. You set the bar and track whether each action has earned the next step.
• Turn corrections into data. When someone rejects or edits an agent's work, that's a labelled example. Build the eval sets that catch a regression before it ships.
• Read the trace. When an agent gets something wrong, open the trace and find out why: bad data, a missing rule, the wrong tool or a weak prompt. Fix the cause, not the symptom.

What You Bring:

• Experience: 5+ years as a product manager, shipping software people use every day. You care as much about how it feels to use as whether it works.
• You've built with AI, not just used it. You've shipped something with a model inside it, even a side project, and you can tell us where it broke and what you did about it. You're comfortable in Claude Code or Cursor, SQL and Git.
• You think in evals. "It works" isn't a claim until there's a set of real examples it passes. You'd rather show a precision number than a demo.
• You know why it failed. You can tell a model limitation from bad data, a missing rule or a broken workflow.
• You know when not to use an agent. A rule, a script or a better form is often cheaper and more reliable. You pick the simplest thing that works.
• You go to the work. You'd rather sit next to a project coordinator for a morning than read a summary of what they do.
• You write clearly. Your specs read like instructions an agent could follow, because sometimes they are.
• You can say no. With 40 requests in the backlog, you can name the three that matter this week and explain why.

How We Build:

• PMs ship code. Our product team builds with Claude Code and other AI tools every day: prototypes, internal tools, data pulls and fixes. Getting your hands dirty is the job, not a bonus.
• You fix what you find. Spot a broken report or a clunky step? Build the fix yourself, then bring in engineering when it needs to scale.
• We're always on the latest. The newest tools, models and ways of working, and we switch the moment something better ships. You'll have all of it, plus real data and engineers in the room when a prototype needs to become production software.

How We Hire:

Our process includes a 90-minute working session. You'll get a real (anonymized) operations workflow, the AI tools you'd use on the job, and one of us to work with. We want to see how you work, not how you interview.

Job Type: Full-time, Permanent

Pay: $125,000 to $158,000 per annum

Benefits:

• Dental Care, Vision Care & Life Insurance premiums fully paid by Jetson
• Extended Health Care
• Disability Insurance
• Education Support
• Parental Leave Top Up
• Equity

Work Location: In person - Kitchener, ON