Chatbot Copywriting Jobs (NOW HIRING)

Job Description: About LawnStarter LawnStarter is the largest on-demand marketplace for lawn care and outdoor services in the country, booking more than $150M a year. We run three brands — LawnStarter, Lawn Love, and Home Gnome — on a single shared platform, and we're expanding beyond lawn care into the rest of home services. About Lifecycle Marketing at LawnStarter Our customers come to us with a problem they want solved now: the lawn needs mowing. Many of them start a signup and then stop, and what we say to them in the minutes, days, and weeks that follow is lifecycle marketing's job. At our volume, a tenth of a percentage point is real money. The program spans three brands and roughly 1,750 unique message templates going out to customers today. It was built to convert, and it does. We've already put AI to work inside it — every SMS reply a lead sends us now routes to an AI rather than a person, and that change moved the numbers. That is a starting point rather than a finished program, and improving on it is your job. The Role You'll own the end-to-end customer lifecycle marketing across LawnStarter, Lawn Love, and Home Gnome, driving quantifiable growth through customer acquisition, reactivation, and cross-sell expansion. Ownership encompasses the strategy, roadmap, copy, execution, and measurement of every message we send, along with the health of the transactional messages our customers depend on. AI is central to how you'll do this. You'll decide where and how AI does the segmenting, the copywriting, and the campaign build, so that you produce more messages and return more revenue. The AI Bar One of our core values is AI or Extinction: every task, AI-first, no exceptions and no excuses. We mean that literally, and this is the role where we expect it to show most clearly. How the work actually gets done here. Our messages get built by driving agentic AI against our internal repo. That means writing the skills and prompts that generate email templates, Content Blocks, and the Liquid behind them, pushing them to Braze through its API, running them, checking the output, and fixing it when it's wrong. Campaign and Canvas structure gets assembled in the dashboard, but the content those steps reference is generated, versioned, and updated from the repo. We run Claude Code today, and we'll keep using whatever is best for each part of the work. What you've used elsewhere doesn't matter to us — Codex, Cursor, Claude Code, something newer. What matters is that you've written skills and agents that do real work, rather than only prompting a chat window. You will not be writing application code and you don't need an engineering background. You do need to be comfortable working against a repo and version control, treating an agentic tool as your normal way of getting work done, and able to read and correct the Liquid, HTML, and JSON it produces well enough to catch what the model got wrong before it ships. We're not looking for someone who drafts a little faster with a chatbot. We're looking for someone who changes what lifecycle marketing is capable of doing: Personalization at a level of detail no person can produce by hand - Writing and personalizing for segments too small to manually create, and then testing at that size. Conversations rather than sends - Our inbound SMS for leads already routes to an AI, and you'll take that further into the lifecycle. Tests we wouldn't have thought of - Using models to find the segment, write the hypothesis, and interpret the result. Agents that handle the build - Skills you write that generate templates and Content Blocks, wire the Liquid behind them, check them against our rules, and catch a broken message before a customer ever sees it. Canvas structure still gets assembled in Braze, but everything it references comes from the repo. Monitoring that runs without you - Roughly 400 transactional messages fire 63,000 times a day, and no person can check that by hand. You'll use AI to monitor the send data so a message that arrives late, sends twice, or contradicts another channel gets caught and diagnosed on its own — and where it's safe to do so, corrected before a support ticket tells us about it. Ideas that aren't on this list - We expect you to arrive with approaches we haven't considered, and if the tooling shifts six months from now, we expect to hear about it from you first. Whoever takes this role should end up being the person the rest of us come to when we want to work this way. If you've been the most advanced AI user everywhere you've worked, this is a job where that really counts for something. We'll ask you to show this work, so come prepared to walk us through what you've built and what it did. What makes this role different You do the work yourself - There's no agency and no contractor pool. You write the message, build it in Braze, send it, and read the results. The build is generated, not hand-made - Message content gets into Braze because you wrote the skill that puts it there, not because you hand-built the hundredth variant in the composer. If that sounds like the interesting part of the job rather than something to get through, we should talk. The program already works - You're not starting from nothing, and you're not cleaning up after someone else. You're taking something that already performs and making it considerably better. You’ll implement brand strategy - None of our three brands has a written voice or design standard applied to its messaging, which leaves roughly 1,750 messages waiting to be rewritten and redesigned (AI first). You’ll partner with the brand team to create the look and feel for three distinct brands. What You'll Own How we use AI, and the tooling that does it What gets automated, what gets personalized, what an agent handles, and what stays human. You set that direction and then build it yourself: the skills, the prompts, the Braze API calls, and the checks that run against the output. We've tested the lead drip hard for two years and there are no easy gains left. What remains is personalization and segmentation at a level of detail no person can produce by hand. We know where the opportunity is. Nobody has built the thing that captures it. Strategy and roadmap The full customer journey across all three brands, covering lead conversion, onboarding, retention, and winback. You decide what gets built and in what order. Most of our testing has gone to the start of the funnel, so we have tested retention and winback far less — even though a customer we keep is worth considerably more than a lead we convert. That's the clearest open opportunity in the program. Every message, and the words in it Email, SMS, in-app, and push: who gets it, what triggers it, when it arrives, how often it repeats, and whether it's right before it sends. You write the copy yourself, and what you write is what customers read. The brand build Roughly 1,750 message templates were built one message and one test at a time, each built to convert, but no voice or design standard was ever applied because none was ever written. You'll partner with the brand team to write those standards, then apply them to every message across all three brands — sequenced, tracked, and shipped as a program. The writing isn't the hard part. The hard part is doing that much work while the program keeps running, and having the discipline to leave the messages that already win alone. Transactional health Roughly 400 messages go out about 63,000 times a day, telling customers their Pro is on the way, the work is finished, their payment failed, or their Pro has changed. They're the messages we send most often and the ones customers trust most, and nineteen of those journeys fire across email, SMS, and push simultaneously. When one arrives late, says the wrong thing, sends twice, or contradicts what another channel said an hour earlier, the customer notices before we do. You'll build the monitoring that catches it and stay with engineering and support until it's actually fixed. Measurement, and what counts as a result The testing roadmap, the analysis behind it, and the call on what becomes the new control. You set the bar for significance and you don't lower it, including when the answer is that your own idea lost. The division of work with CRO They own what happens on the page; you own what happens in the message. When something involves both — a quote abandoned on-site, a price that surprised someone — you'll work it out together. What Success Looks Like (Year 1) AI is doing work we couldn't do before — more than one capability in production, each handling something no person here could have done by hand, and each with a measured result Conversion is up— more leads become customers in the window you own, and the tests demonstrate it Retention has moved— a winback and retention program tested as rigorously as the lead drip is today, with impact you can measure Test velocity is considerably higher — many more tests each quarter than we run now, held to the same standard for what counts as a result Transactional problems get caught, not reported — you learn a message has broken from your own tracking rather than from a support ticket, and there's a record of what broke and what you did about it Every message carries the brand — all three brands rewritten and redesigned, with conversion performance maintained RequirementsWho You Are You build with agentic AI Claude Code, Codex, Cursor, or whatever you've settled on — an agentic tool is already how you get work done, and you've written skills or agents with it that run when you're not sitting there. You know exactly how they fail, because you've fixed them. When a tool can't do what you need, you build what you need rather than waiting for the vendor to ship it. This is unlikely to be a good fit if your AI use amounts to drafting emails and summarizing documents. You start from what the customer is feeling. Someone who abandoned a quote because the price surprised them needs a different message than someone who's been waiting three days for a Pro to show up. You work from what the person is thinking, then decide what the message has to do about it. AI will write you a hundred variants and a test will tell you which one won, but neither one tells you which hundred were worth writing. This is unlikely to be a good fit if you choose the tactic before you understand the motivation, or if you'd rather run tests than form a view about the person first. You write, and you read the numbers. You'll draft the SMS and interpret the significance test on it, because both belong to you and you're genuinely good at both. This is unlikely to be a good fit if you write well but hand off the analysis, or analyze well but hand off the words. You work in the tool. You'll be in Braze rather than reviewing someone else's build, and you'll be writing the skills that put things into Braze. This is unlikely to be a good fit if you've spent your career directing an agency and would find the build work beneath you, or if anything resembling code is something you'd rather hand to someone technical. You change your mind when the data tells you to. You hold strong hypotheses and give them up quickly, and you say so out loud when you were wrong. This is unlikely to be a good fit if you need to be right, or if you import practices from another company without testing whether they apply here. You investigate errors. When a customer reports getting the wrong message, you'll follow it back through the canvas, the event, and the data until you find where it broke, and then stay with engineering until it's actually fixed. This is unlikely to be a good fit if you regard debugging as beneath the strategy work, or if you'd rather build the next campaign than understand why the last one went wrong. This Role Is NOT A place to learn AI on the job - We need someone already working this way who has results to show for it. If you're curious but early, this isn't the role. A management role - No direct reports and no agency to direct. If your next step is leading a team, this isn't it. Brand strategy - You'll apply each brand's voice across every message, but you won't decide what that voice is or write the guidelines, and you'll be judged on conversion and retention rather than on creative. If you want to own the brand, you'll be frustrated here. An execution-only role - Nobody hands you the roadmap. You decide what to test and what to build, then defend it with results Demand generation - Lead volume and lead quality belong to paid acquisition, and you own what happens after the lead arrives. If you want the top of the funnel, that's a different job. Pro lifecycle - This is the customer side; another team handles the roadmap, strategy, and copy for Pros, although you’ll partner to handle the execution. Benefits Base salary: $135k-$165k Equity: Conversion and retention determine a good part of what this company is worth, so you should own a piece of it. Healthcare: Medical, dental, and vision Fully remote (US): Writing well and reading tests carefully both take uninterrupted time, so we trust you to work where you work best. Flexible PTO: We judge results. LawnStarter provides equal employment opportunities (EEO) to all employees and applicants for employment without regard to race, color, religion, sex, national origin, age, disability, or genetics. We comply with applicable state and local laws governing nondiscrimination in employment.