8:42 AM - The First Alert
I crack my laptop open with coffee still steaming. Slack pinged at 8:37, which means at least one of the overnight inquiry batches triggered a flag. The Rental AI dashboard loads - I've memorized this sequence by now. Top right: today's stats still blank. Three new signups since midnight. One of them is Marcus Chen at Chen Properties, a twelve-unit portfolio in Arizona I'd been chasing on LinkedIn for six weeks. That's twelve more tenant inquiries I won't have to answer at 11 PM on a Tuesday.
The Slack notification was worse. Sarah Kim at Northgate Apartments flagged a conflict at 2:47 AM her time. One of my agent responses drafted to a tenant asking about move-in costs. The AI suggested a monthly payment plan for the deposit. Sarah's lease doesn't allow deposit splitting. That's a liability risk. I open the draft in my admin dashboard and read it. The agent nailed the tone - friendly, thorough, empathetic - but missed a critical detail Sarah had entered during setup: "deposit must be paid in full, non-negotiable."
I write Sarah an apology email and a corrected draft. It takes eight minutes because I want to show her I actually read her setup documentation, not just rubber-stamped her system. I don't know if she's thinking about churning yet, but the care matters. She responds within ninety seconds. "Thank you for jumping on this. This is exactly why I hired you guys." I screenshot it. Good moment.
10:15 AM - The Review Batch
Overnight, the agent ran 426 inquiries across my seventy-three active customers. Of those, 68 got flagged for human review - gaps in the customer's setup, tone decisions that needed judgment, edge cases. I won't review all 68 today. But I pull the top fifteen by priority: three customers near churn, four new customers in their first week, eight miscellaneous issues.
First in the pile: Todd Wilson at Wilson Rentals. A tenant asked whether a service dog was covered under his pet policy. The agent drafted a response that explained the difference between a service animal (not a pet, covered under ADA) and an emotional support animal (classified as a pet). Smart distinction. But the tone came off slightly condescending - "As you may know, the ADA defines" - and Todd's voice is gruff and direct. I rewrite the middle of the response to match his register. "Here's the thing: service animals aren't pets under the law, so your policy covers them. If they claim emotional support status, that's different and we should talk about it."
I approve it and the agent pushes the rewritten version to Todd's outbox queue. Six minutes of work. The agent saves Todd from writing an hour-long email that gets the law right but offends a tenant. That's the core of the job.
Second item: Alicia Santos at Santos Property Group submitted a template update yesterday. She wants all her responses to mention a lease amendment for guests staying longer than fourteen days. The agent didn't catch this as a standing instruction, so three replies went out without it. I flag this as a bug in my Linear workspace - "Guest-policy standing instructions not being pulled into overnight batch." I assign it to myself. That's code work, not operator work. I'll fix it before Friday. For now, I manually send Alicia corrected versions and ask her to override the originals in her email thread. She'll have to do light cleanup. Not ideal. She's a $249-per-month customer on an annual plan, and I'm asking her to fix my system's mistake.
12:47 PM - The Metrics Check
I step away from the review queue to look at the week. My Gmail is synced with Stripe for payment notifications, so I glance at my inbox: seven payment receipts this week. That's consistent with my baseline. My Stripe dashboard shows $1,043 in MRR arrivals for today alone, which is high. I pull the weekly summary. Three annual plan conversions this week. That's $897 of that $1,043. Two of them are new customers - the Marcus Chen signup this morning, plus a company called Meadows Management that converted from a free trial yesterday.
Year-to-date, I'm at $34,200 in revenue. I had a goal of $40,000 by the end of month two. I'm nine days out. The pipeline shows seventeen active trials. If I close half, I hit $40k by Friday. That's not a margin call. It's a modest miss by a few days, and the trend is still up. I finish my lukewarm coffee and go back to the review queue.
2:13 PM - The Churn Conversation
Email comes in from Jennifer Ng at Ng & Sons Property Management. The subject line is "Thinking about pausing." I read this three times. She's been a customer for six weeks. $179 per month. She says the agent gave her tenant some advice about local rent-control ordinances that didn't match her state law, and now she's worried about liability. I check the agent's draft. It's technically correct, but it's opinionated about a grey area, and the agent should have flagged it for Jennifer to decide.
This is not a system failure. This is a judgment call the agent got wrong. I write Jennifer back and offer three options: I can dial back the agent's specificity on legal questions and have it stick to her policy documents only. Or I can refund her last two months and call it even. Or we jump on a call and I show her how to tighten the boundaries on the system.
She picks the call. We talk for twenty-three minutes. I show her the setup form and explain where the agent is pulling from. I explain what instructions she can tighten. She asks three hard questions about edge cases. I answer two from experience and admit I don't know the third - it's a state-specific landlord thing - and I promise to research it and send her a followup email by tomorrow morning. She agrees to keep the subscription.
I do the research. Thirty minutes on the state bar association website. I send her the three-option framework she needs. She replies "You sold me on this product as a tool, not a robot. I appreciate you being honest about the limits." That's the narrative I need to stay in. This costs me an hour and saves me $179 monthly. The math is real.
4:35 PM - Pipeline and Personal Emails
I open the admin dashboard to pull the week's pipeline view. I track this manually in a Google Sheet alongside our Stripe data, cross-referenced with the emails I'm sending. Five of the seventeen trials are property management companies. One is a landlord co-op in Colorado with twelve units. The odds are good. I email three of them with a "how's it going" message. Light touch. No aggressive close. Two come back within the hour with questions. One asks if I can integrate with their Rent Manager software. I don't have that integration. I tell him the truth: "Not yet, but it's on the roadmap. The API work is maybe six weeks out, and I've got two other customers asking for it." He appreciates the honesty more than a fake promise. I put a sticky note on my desk to check that integration by June. It's a simple REST call. Doable.
5:58 PM - Closing Out
I review my to-do list. Bug on standing instructions: flagged in Linear for Friday. Email follow-ups: six sent, three responses waiting. Customer success calls: one unscheduled, one pending (Jennifer Ng's state law question). Customer escalation emails: all resolved. The backlog of overnight agent reviews: still at forty-two remaining, I'll touch them tomorrow morning.
I close my Slack. The message threads go quiet. I shut down my Stripe dashboard and Gmail. I close the admin panel. The laptop screen goes dark.
Seven weeks in, and I've sold Rental AI to seventy-three customers. None of them believe a robot is running their business. They believe I'm using a tool to move faster and catch things they wouldn't. That's the honest thing. The robot does the volume. I do the judgment. Some days feel like glue. Some days feel like leverage. Today was both. That's the work.