8:42 AM - Inbox triage
I open the laptop to three Slack notifications from the Church AI alerting system. First one: Redeemer Church in Portland had an error generating their Sunday visitor follow-up email yesterday. The AI template injection went wrong - it grabbed last week's attendance count instead of this week's. The second notification is neutral, just confirming our scheduled daily digest generated 47 bulletins overnight for churches on the standard plan. The third one makes me smile. Carol Reyes at Redeemer sent a note: "Our admin said this saved her three hours Tuesday. She can't stop talking about it."
I flip to Gmail. The inbox shows 23 new messages since yesterday evening. Twelve are auto-forwarded demo requests from the Facebook group outreach campaign. One is an invoice query from First Reformed in Ohio asking why they were charged twice last month - I can see in our records they had a failed payment, retried it, and the retry went through, so we actually owe them a credit. One message is a cancellation notice from Mount Carmel Assembly. The subject line just says "Not using it." That stings. They were on the plan for six weeks. I add it to my mental list to follow up with a cancellation survey.
I open the Church AI operator dashboard. Today's numbers: two new signups came in overnight from the Facebook outreach - churches in Arkansas and Wisconsin. This week's running total is four new trials. The pipeline shows seven demos booked for the next ten days. Yesterday's revenue was $594, which means our current customer base is running at roughly that monthly recurring rate. Six active customers, paying seventy-five to one hundred twenty-five dollars depending on congregation size. It's not nothing. It's real money from real people who are using the tool.
10:15 AM - A bulletin under review
I click into the pending-review queue. There are eight bulletins waiting for final approval before they go out to customer churches. I start with the first one: Grace Community Church in Charleston wanted a four-paragraph Sunday bulletin with a specific tone - warm, welcoming, not corporate. The AI drafted it and I need to read it before it ships.
The draft looks clean. Paragraph one announces the Sunday service and invites first-time visitors. Paragraph two highlights a grief support group starting next month. Paragraph three has a giving reminder, which the AI knows to keep subtle because we've flagged overly aggressive giving language as a quality issue in the past. Paragraph four closes with the week's event schedule.
I approve it and it queues for delivery.
The next one from First Assembly in Dallas has a problem. The AI generated a visitor follow-up email but it references "your donation on Sunday" when the church's giving data shows the recipient didn't actually give. I flag it for manual review and send a note to Dallas: "Your follow-up email needs a quick edit before it goes out. I'm sending you the draft separately so you can approve the fix." This is the work nobody tells you about. The AI does the heavy lifting, but I'm the quality gate.
I open Linear, our bug tracker, and log this as an issue: "AI giving-reference false positive when parsing donation data from Google Sheets integration." It's a known gap - we're pulling giving records from a sheet that doesn't always sync cleanly with the attendance data. I tag it as low-priority because it only affects maybe five percent of our customer base and we already have a workaround, but it's something I want to fix this month.
12:30 PM - Lunch and the metrics check
I grab a sandwich and sit back at the laptop with a cup of coffee. Time to check the real metrics. I open our Stripe dashboard. This week's billing shows:
- 6 active subscriptions
- $394 MRR from existing customers
- 4 new trial signups (not yet paid)
- 2 failed payment attempts (one from a church whose card expired, one from a church that looks like they've gone quiet)
I send a quick email to the church with the expired card: "Hi there. Your payment method declined this month. It usually means the card's outdated. Reply with an updated card and we'll charge you right away, or let me know if you want to pause the subscription for a month."
In the sales Slack channel, I check what's happened with last week's demos. Three companies said they needed to think about it. One church actually decided not to move forward - budget freeze at the end of the fiscal year. Two are actively in the evaluation phase and asked for custom integrations with their existing database system, which is above what we can do in month two. I make a note: "Custom integrations are a common request. This might be a product expansion in quarter three."
One demo from four days ago converted. Parkside Community Church signed up yesterday and their first trial bulletin generated overnight. They seem engaged.
2:08 PM - Customer escalation
A phone call comes in through the booking link. It's Patricia Morse from Calvary Baptist. She's spent the last hour trying to set up the AI to generate visitor follow-ups that match her church's existing email templates. The system isn't pulling her template variables correctly and she's frustrated.
I take the call. She walks me through the setup. I can hear the frustration, but also that she really wants this to work. "I don't want to manually write these emails anymore," she says. "That's why I bought this."
I walk her through our Google Docs integration - she needs to name her template file a specific way so the AI can parse it. It's a documentation problem on our end. I email her the exact file naming convention and schedule a fifteen-minute call with her for Thursday to verify it's working. Before I hang up, I say, "I'm going to add a note in the setup wizard to make this clearer for the next person. Thanks for catching it."
After the call, I open our admin UI and make a quick edit to the onboarding checklist so the next customer gets better instructions. Small fix. Takes five minutes. Saves someone else the frustration Patricia just had.
3:45 PM - Following up on churn
I draft an email to Mount Carmel Assembly, the church that cancelled. I don't use a template. I write it personally.
"Hi, I saw that you cancelled your subscription yesterday. I'd love to understand what didn't work for you. Was it the interface? The output quality? Budget? Or something else? I'm still working through bugs and feature requests, and your feedback would help me build something that actually solves the problem. If there's anything I can fix to bring you back on, I'm here for it."
I send it. I don't expect a response, but sometimes people reply and sometimes that reply tells me something important about what I'm missing.
4:30 PM - Pipeline review and the demo prep
I check my calendar. There are three demos scheduled for tomorrow and Thursday. One is a large church with 450 members - the biggest prospect we've had. They want to see the bulletin generation and the visitor follow-up workflow in action. I spend twenty minutes building a demo narrative in my head, then I open the actual product interface and walk through it like a customer would, taking notes on what's clear and what's confusing.
The interface is clunky in one place - the settings for congregation size aren't obvious and a new user might miss them. I log it. Then I write a one-page email outlining the demo flow for the big church, with specific examples and a timeline. I send it to myself as a reminder to review it tomorrow morning before the call.
5:45 PM - Wrapping
I close the pending-review queue. All eight bulletins have been approved or flagged. I check the Slack alerts one more time - no new errors. I look at the dashboard again. Today's numbers: two new signups, $594 in recurring revenue today, one customer call handled, one churn followed up on, one bug logged, one setup issue fixed.
By month two, I thought this would feel more automatic. It doesn't. The AI does the drafting work, the template generation, the heavy lifting. I do the decisions. I approve, I fix, I read customer emails and respond to them personally, I chase the churn, I track the pipeline. This is not passive income. This is a business.
The part that worked today: Carol's thank-you note. Knowing that an admin at a church actually got three hours back in her week. That's why this exists.
The part that was hard: Mount Carmel leaving, and knowing I don't fully understand why. Patricia's call - a smart person frustrated because my documentation wasn't clear enough. That's my fault.
What I'd change: I need better template documentation before customers even start. I need to figure out why some churches churn after six weeks. I need to get the giving-data sync actually working correctly so I'm not catching false positives in manual review.
I close the laptop at 6:12 PM. Tomorrow there are three demos. By Friday, I want to have fixed the template naming issue in the wizard and sent out an updated help doc. Small moves. Steady progress. This is the work.