8:42 AM - Inbox triage
I hit Command+1 to pull up Slack before I've finished my coffee. Three alerts. The dashboard shows Ashton & Sons Roofing had a job cost come in 18 percent over margin yesterday. The system flagged it automatically. I note it and move on for now - I'll ping their ops manager later. Two new signups came through overnight. One from a HVAC shop in Michigan called Reliable Mechanical, the other from a plumbing outfit out of Phoenix. Both filled out the discovery form at 2 AM and 3:45 AM respectively. Neither has been onboarded yet.
I open my email client and scan the inbox. There's a reply from Carol Reyes at Reyes Family Practice - she's been a customer for about three weeks. Her message is short: "This system just caught something I've been losing money on for months. Thank you." I feel something good in my chest. I take a mental note to ask her for a referral next week. No need to be pushy about it today.
Below that is a message from my payment processor flagged by Stripe. One of my churned customers from last week, Torrence Construction, has a failed renewal. I click into their account. I can see from the logs they only used the system twice before going quiet. The onboarding probably didn't stick. I make a note to audit my onboarding flow later - maybe there's a pattern I'm missing.
9:30 AM - AI draft review and approval
I switch to the Margin Protection dashboard and navigate to the drafts queue. The system has prepared three outreach emails from last night's prospecting runs. Each one is tailored to a different customer archetype: a concrete contractor, a dentist's office, and a small electrical firm. I've learned to trust the framing about 75 percent of the time, but the language sometimes misses. The concrete one is solid - it mentions a specific pain point about material costs outpacing labor margins. I approve it and it queues for sending.
The second email is too aggressive. It's pitching features instead of leading with a problem. I mark it for revision and leave a note: "Focus on the cost bleed, not the features." The system will regenerate it in the next cycle. The third one is fine. I approve it.
By the time I'm done here, it's 9:50 and I've got eight items on my task list for the day.
10:15 AM - A flagged conflict
I open Slack again and there's a message from the system about Ashton & Sons. It's not just a margin miss - the daily digest shows they're trending toward 12 percent below target for the entire week. One job is normal variance. But this pattern suggests something structural. I know the owner, Derek, from my sales call. He's sharp but doesn't have an ops person. That's actually why he bought the system.
I draft a Slack message to Derek: "Hey, I'm seeing a trend with your margin on jobs this week. Not saying anything's wrong, but if you want to hop on a call and talk through it, I'm free this afternoon." I don't want to be alarmist. But I also want him to know I'm watching and that I'm a partner, not just a vendor.
He replies within two minutes: "Yeah, we're having a supplier issue. New lumber pricing. Can we talk at 2?"
11:45 AM - Onboarding follow-up
I open Linear to check my bug queue. There's one item flagged from yesterday: the onboarding workflow is missing a step where customers connect their job costing system. I'll need to log that as a technical task, but for now I decide to manually walk through onboarding with both new signups.
I start with Reliable Mechanical in Michigan. I send them an email with a Loom video I recorded last week walking through the first setup. It's about four minutes. More personal than documentation. I add a line: "Give me a call if anything's unclear. Here's my Calendly." I include a booking link for a 15-minute sync.
12:30 PM - Lunch and the metrics check
I grab a sandwich and open the Stripe dashboard on my phone. This is something I do now without thinking. Today's revenue is 2,140 dollars. That's about right for a Tuesday. Week-to-date I'm at 10,480. The run rate is tracking toward about 110K for the year if this holds, which was my starting target. I've picked up three new customers in the past two weeks, and I haven't lost anyone except Torrence yet.
I flip back to the admin dashboard and check signups for the week. Seventeen so far. My pipeline is stronger than I expected. I've got seven discovery calls scheduled for the next eight days. Three of them look warm - they've already mentioned specific pain points in their intake form.
What's harder to quantify: I have no idea if I'm actually building a good business or just riding luck. Two months in, I'm thinking about things I didn't expect to think about. The churn on Torrence is making me wonder if my ICP is really as tight as I thought. Service businesses from 5 to 50 people sound like one category, but a five-person plumbing outfit operates completely differently from a 30-person HVAC empire.
2:00 PM - The customer call with Derek
Derek picks up at exactly 2 PM. We talk for eighteen minutes. The problem is exactly what he said: a lumber supplier increased prices by 30 percent, and it's flowing straight through his job costs. The margin loss is real, but it's temporary. The system is working - he caught it in real time instead of discovering it in month-end accounting like he used to.
He's thinking about raising his estimates, but that could lose him bids. I don't have a solution for him. But I listen. I ask good questions. I tell him I want to revisit this in a week to see if the trend stabilizes. He seems grateful that someone is watching.
After we hang up, I feel better. This is the kind of call that reminds me why I bought this business. The system isn't doing Derek's job for him, but it's giving him visibility.
3:30 PM - A manual billing edge case
An email comes through from Phoenix Medical Services. They're asking about a refund because they've been charged twice for the same month. I log into Stripe and I can see exactly what happened. They were on a monthly plan. When they upgraded to the annual plan last week, the system didn't cancel the monthly subscription first. So they got two charges. My system should have prevented this, but there's a gap in the logic.
I process a manual refund of 150 dollars and send them an apology email. I also log the issue in Linear as a bug: "Prevent double billing during plan upgrades." It's a small issue, but it matters. Losing that customer over 150 dollars would be stupid.
4:30 PM - Pipeline review and next week's plan
I pull up my pipeline view. Seven warm leads, three lukewarm, fourteen cold outreach runs scheduled to go out tomorrow. The system is generating prospecting lists every evening, and I'm reviewing them in batches now. I used to do this manually, but that was eating three hours a day.
I check the calendar for this week's calls. Tomorrow I've got three discovery calls. Thursday I've got two more. Friday is lighter. Next week is starting to fill up too.
I think about what Derek said on the call and make a mental note: I should start segmenting my outreach by company size and industry. The system is smart enough to do this, but I haven't told it to yet. That's a config change that could save me a lot of follow-up time on mismatched leads.
5:45 PM - Closing up
I send a message to Carol Reyes asking if she knows anyone else who might benefit from the system. I say it plainly: I'm building this business and I'm looking for customers who are going to get real value. If she knows someone, I'd appreciate the intro.
I close Slack, then Gmail, then the dashboard. I look at my notes from the day. Three clean things: the AI-drafted email that worked, Carol's thank-you note, Derek's problem solved on the call. Two things that were harder: Torrence's churn, which I still need to understand better, and the billing edge case, which I've logged but which meant I had to do manual work today.
I think about month one and month two. The business feels more real now. It's not magic. It's me reviewing an AI system's work, talking to customers who have real problems, fixing small bugs, and keeping an eye on numbers that matter.
I lean back in my chair. I'd be lying if I said this was easier than I thought it would be. But it's working. And it's real.