8:42 AM - Inbox triage
I open Slack and scroll through the overnight pings. Three alerts in the Spa AI operations channel. One's a green flag - a med spa in Portland converted their trial to annual, first customer to lock in yearly. Another is yellow: Carol Reyes at Reyes Family Spa never logged in yesterday. The third is a surge notification: fifteen new trial signups hit between 2 and 4 AM, which means yesterday's cold email batch landed cleanly.
I pull up the admin dashboard. MTD we're at $2,985 in new MRR. Current active trials: 23. Pipeline in sales conversations: 8. It's only Tuesday and we're already up 34 percent from last week's daily average. I make a mental note to double-check the email that drove the surge - something worked.
Gmail is next. The AI system prepped four retention outreach drafts overnight for customers who've gone quiet. One is flagged for my review: a client at Luxe Skin in Denver who churned six months ago, now showing new booking activity in their integrated calendar. The email subject line reads "We noticed you're booking again." I read through it. It's not pushy, but it's not quite right either. The tone's too formal. I edit it down, make it warmer, change "We noticed you're booking again" to "Hey Marina, I saw your therapist Elena has a waitlist again. What changed?" and approve it to send.
The other three emails look solid. Approve, approve, approve.
10:15 AM - A flagged conflict
Slack again. Alert from the conflict-detection layer: James Park at Park Wellness in Austin has two retention sequences scheduled for the same prospect on the same day. The AI was trying to thread two separate behavioral signals - a client who hadn't booked in 45 days and another who hit a price-sensitive threshold - and queued both campaigns.
This is exactly the kind of thing I need to catch. I open the admin panel and look at the full picture. James's med spa has fifty-four active customers. The duplicate sequence would've hit Margaret Chen, one of his high-value clients, with two separate "come back" emails three hours apart. She probably would've noticed.
I disable one sequence - the 45-day-silent one. The price threshold signal is stronger. I note it in Linear, mark it as a product bug that needs the batching logic tweaked, and send James a heads-up in a quick email: "Caught a double-sequence issue on your account this morning. You're good now, but I'm flagging it to our product team so this can't happen again." Small touch, but it buys trust.
11:50 AM - The metrics ritual
Every morning, I check Stripe. Today's revenue cleared: $1,247 from recurring subscriptions, $89 from a new monthly customer who signed up yesterday. The math still feels strange. Two months in and I'm watching money land in my account in real time. Some days it feels like I'm checking a lottery ticket.
I open the metrics tab in the admin dashboard. Average response time on our escalation queue is 4.2 hours. That's creeping up. Tomorrow I'll need to add another escalation slot or prioritize differently. New trial activation rate this week is 62 percent, which is solid. Churn is holding at 8 percent, which is where I want it for now.
I pull the conversion report. Out of the 23 active trials, 8 are in real sales conversations. Two of them - Erica Chen at Chen Aesthetics and a new prospect, Diana Lopez, running a boutique laser clinic in Sacramento - both asked detailed questions about integrations yesterday. These are hot. I make sure their white-glove onboarding sessions are confirmed for Thursday.
12:30 PM - Lunch and the escalation
I'm halfway through a sandwich when a Slack notification lands. Carol Reyes responded. She didn't log in yesterday because she was debating whether to keep the subscription. Revenue is down. She's worried about whether the AI can actually move the needle for a smaller clinic.
This one stings. Carol's been with me since week two. She's sent me a voice message, which means she's serious.
I call her back. Fifteen minutes later, I understand the issue. Her team didn't implement the recommendations the system generated two weeks ago. The AI flagged four customers for targeted outreach, and Carol's staff just... didn't do it. The system works, but it only amplifies what you actually execute on.
I walk her through it. I show her the three clients who re-booked after the first round of outreach worked. I offer to help her team set up Slack reminders so the recommendations actually land in their workflow. We reset the trial timeline. She'll stay for another month, commit to actually running the outreach. It's not a done deal, but it's not a loss.
I make a note: "Product roadmap - auto-assignment to staff calendars" and send it to Linear.
2:08 PM - The bug
An email lands from James Park again. One of his customers got sent a retention email for a treatment they actually just booked. The AI pulled historical booking data, but it didn't check the most recent entry correctly. James's tone is professional, but I can feel the frustration.
I dive into logs. The issue is in the booking-sync layer - a timezone parsing error that's causing a four-hour drift on some integrations. It's a small thing, but it breaks trust. I file an urgent ticket, assign it to myself, and spend thirty minutes writing a fix. Test it against James's data. Confirm it works. Deploy at 2:47 PM.
I email James back: "Found it. Fixed it. Deploy went live fifteen minutes ago. Should be clean going forward. Thanks for flagging it - this was a real issue and I appreciate you being direct about it."
His response comes back in four minutes: "Appreciate the fast turnaround. Let's talk about the annual plan next month."
That one lands different.
4:30 PM - Pipeline deep dive
Back to the dashboard. I pull the sales pipeline report and open the detail view on all eight hot conversations. Cross-reference with Gmail to see where emails are stuck. Three of them need follow-ups. Two are waiting on me to send custom ROI projections. One - Diana Lopez in Sacramento - is asking for a demo of the churn-prediction model.
I record a seven-minute Loom walkthrough of exactly what Diana asked about, send it with a personal note, and CC our head of onboarding so she's ready to jump in if Diana moves to a trial.
The pipeline feels real now. In month one, I was hunting to fill it. Now I'm triaging it. Different problem, better problem.
5:15 PM - The thank you
A customer from month one - Priya Sharma at Priya's wellness clinic in Portland - sends an email with the subject "You weren't joking." She's got numbers now. In six weeks, the system flagged twenty-one customers she was about to lose, her team executed the outreach, and she kept eighteen of them. At her average service value, that's about $2,400 in prevented churn. The email is short and honest: "This thing actually works. Not hype. Genuinely changed how I think about my business."
I read it three times. I write her back: "Priya, this is exactly why I built this. Thank you for putting in the work on your end - the system is only as good as what you actually do with it. Your story is going in the next case study."
She replies: "Happy to. Hope it helps you sell this."
5:50 PM - Close out
I scan through the Slack channels one more time. No new red flags. The Portland annual customer is already activated and uploaded their booking data. Carol Reyes's trial is re-engaged. James Park's bug is live. The pipeline has three hot leads and two more warming up.
I close the laptop. It's 6:04 PM.
This day felt different from last Tuesday. Less hustle, more operating. I spent time on three customer problems - not all of them easy - and fixed one in code. The AI system worked hard overnight, but every output that mattered crossed my desk first. I approved, edited, rejected, or escalated. That's the job. That's what running this engine actually is: reviewing what the system surfaces, making judgment calls, sending one personal email that shouldn't be automated, and catching the conflicts before they hurt someone's business.
The revenue is good. The retention is holding. But the real test is whether I can stay on top of this as we scale. Right now, it's manageable. Eight hot leads is manageable. Eight hundred won't be. That's a problem for next month.
For now, it works. The work is real, and it's worth doing.