About LawnStarter LawnStarter is the nation's leading on-demand marketplace for lawn care and outdoor services, with over $100M in annual bookings. We're expanding beyond lawn care to become the one-stop shop for all home services — operating across three brands (LawnStarter, Lawn Love, Home Gnome) on a single shared platform. About Service Delivery at LawnStarter We make a promise on both sides of the marketplace: customers get a job done well, and Pros get paid fairly for doing it. Service delivery is everything that happens after a customer books — keeping that promise through completion, support, and the moments when things don't go to plan. This is the hard part of running a marketplace: we broker work we can't directly see, between two parties whose interests sometimes collide. Get it right and people stay for years. Get it wrong and you lose customers, Pros, or both. Trust is the product — and increasingly, the systems that protect it are AI-powered. Requirements The Role This is a broad Senior PM role on the quality, trust, and communication side of service delivery — setting the right expectations on both sides, steering Pros to deliver great work, resolving conflicting interests fairly, and making our AI-powered support genuinely good. You'll work on live, high-scale systems with a mandate to make them better. Service delivery is a big area with more than one PM in it. You'll work alongside them; your center of gravity is the trust between both sides of the marketplace. What makes this role different: It's a real area of impact, not a single feature. You'll shape strategy, policy, and the systems behind a whole slice of the post-booking experience — not optimize one screen. It's inherently two-sided. You work for customers and Pros at once, and you're often the one arbitrating between them. Pleasing one side at the other's expense is failure. AI is a core tool, not a side project. Support and outreach already run on AI. You'll push how far that goes — and build the evals that prove it's good before we trust it with more. Problems to Solve Getting expectations right before the work ever starts. Most service failures aren't bad work — they're mismatched expectations. The grass grew a tier past what was booked; "deep clean" meant something different to each side; the yard didn't match the photos. You'll make sure what's promised to the Pro matches what the customer actually expects, that those expectations are the right ones for the property and service, and that Pros are steered toward delivering great work. Every mismatch you prevent up front is a conflict you never have to resolve later. Arbitrating genuinely competing interests. A customer wants a date the Pro can't commit to. The grass is long enough to need a different price tier than booked. The job wasn't what the listing photos suggested. These aren't bad actors — both sides are right from where they sit, and the marketplace has to make a call. You'll build the policies and the (increasingly AI-assisted) decision systems that resolve these fairly and consistently at scale — and make the judgment call yourself when there's no clean answer. Making AI trustworthy enough to do more — which means evals. Customer-side AI already handles ~34% of support at roughly a penny per message, and Pro-side support is still to be built. Expanding either is gated on one hard, ongoing, technical thing: can we prove the AI resolved a case rather than just closed it? You'll own the eval systems — golden datasets, automated judges, regression detection, human-review sampling — that define what "good" means for an open-ended conversation, gate every change, and catch quality drift before a customer feels it. This is the deep, unglamorous work that makes a slick demo safe to scale. Messaging that connects both sides — and quietly protects the marketplace. The inbox is how 500K+ customers and 20K+ Pros coordinate across all three brands — and the record we lean on when something goes sideways. It also has to moderate : catching when a relationship is drifting off-platform (disintermediation) or a conversation is heating into conflict. You'll keep communication easy for the legitimate 99% while spotting the patterns that quietly cost us customers, Pros, and revenue. What Success Looks Like (Year 1) Expectations match on both sides : Pros know exactly what each job requires, customers get what they expected, and fewer jobs go sideways from misalignment in the first place. Competing interests resolve faster and more fairly : A clearer resolution model with measurable consistency — and more of these conflicts prevented up front by better expectation-setting. AI is provably good, and does more : An eval suite gates AI changes so no quality regression ships unseen — and on that foundation, higher resolution on customer support plus a first version of Pro-side AI support live. Messaging connects and protects : Measurable improvement in messaging reliability and engagement, plus working moderation that catches off-platform leakage and conflict early. Who You Are AI-native. You use AI daily in your own work, and you have real intuition for how to measure whether an AI experience is actually good — not just whether it shipped. You've built or owned evals, or you're hungry to, because you know that's what separates a trustworthy agent from a demo. This is unlikely to be a good fit if you treat LLM quality as a vibe check or as engineering's problem to figure out. Comfortable making two-sided calls. You can hold both the customer's and the Pro's interest in your head at once, and you're willing to make the call when they conflict — clearly, and with a rationale you'd defend to either side. This is unlikely to be a good fit if you're a people-pleaser who can't say no, or if you instinctively optimize for one side and forget the other exists. You design for the right outcome up front. You're not satisfied grading work after the fact — you'd rather get the expe