8:47 AM - Coffee and the Slack backlog
I'm on my second cup of coffee, laptop open in my home office, and the Slack notification badge is already showing twelve unread messages. Tuesday morning. I'm two months into running Pencil, and this is the routine now: check Slack, check the admin dashboard, check the inbox. In that order.
The Slack channel shows the usual: three notifications that new users signed up yesterday evening, one message from Marcus at Apex Consulting asking how to customize the template library, and one that makes my stomach drop. "Churn alert: Morgan Realty Group account marked inactive."
They were one of my first five customers. Three seats, paid ninety-nine dollars per seat per month. Two hundred and ninety-seven dollars gone. I read the message twice, then make a note to circle back to them later. Sometimes people pause, sometimes they cancel. I need to know which.
I open Gmail next. The inbox filter I built last month is catching most of the noise, but there's an email from Carol Reyes at Reyes Family Practice. She bought a single seat three weeks ago. The subject line is "Thank you, this saves me hours." People don't usually send those emails. I open it.
"I just used Pencil to generate a proposal for a new contract review service. It took me twelve minutes instead of the two hours I usually block. I actually sent it the same day. The prospect signed back in forty-eight hours. I wouldn't have bothered without your tool. I'm telling my network."
I sit with that for a minute. This is why I bought this thing. Not the dollar amount, though I'm watching those too. This.
10:15 AM - The dashboard and the flag
I open the Pencil admin panel. The dashboard is running clean. Two thousand one hundred and forty users across all tiers. Week-to-date revenue is showing five thousand eight hundred and sixteen dollars. The MRR chart is trending up. I know from last month that trending up is the only thing that matters at this stage.
But there's a red flag in the queue. An agent-generated proposal for a customer named Daniel Kim is sitting in the "Review" state, and someone on his team has flagged it for "outdated pricing reference." I click through.
The proposal is for a consulting retainer, ninety thousand dollars annual contract. The agent generated it fine. Scope, timeline, deliverables all make sense. But the assistant flagged a line that says "based on 2024 pricing benchmarks" when it should reference 2025 data. It's a real error. The AI isn't hallucinating; it's just using outdated training data that I haven't updated in the system.
This is the work. This is me. I can't let this go out. I edit the line, add a note saying "Updated to reflect current 2025 market rates," and approve it. I open a separate ticket in Linear, my issue tracker, labeled "Update agent training data: pricing benchmarks Q2 2025." That's a bug to fix before it breaks another proposal.
Then I send Daniel Kim a direct email: "Hi Daniel, we caught an outdated reference in your proposal before it shipped. Corrected version is now ready. This is one reason human review still matters, even with AI helping." It's honest. I'm not hiding the error. I'm showing that we caught it.
12:47 PM - Lunch and the metrics moment
I step away for lunch. I heat up leftovers and sit at my kitchen table with my phone, running the numbers in my head. I've got four months of operating history now. Month one, sixty-two users signed up. Month two hit one hundred and eight. The monthly churn is running at four percent. Some of that is small users who try it and bail. Some of it is bigger accounts like Morgan Realty Group that I need to understand.
The three-seat teams I'm targeting are the sweet spot. They sign up, they get immediate value, they stick around. Solo practitioners churn faster. I'm noticing that pattern.
I check Stripe. Recurly is handling the subscriptions, but I check the actual Stripe dashboard to see the cash. Fifteen thousand and twenty-three dollars in recurring revenue this month so far, with six days left. That's tracking ahead of last month. Eighty thousand ARR was the number I heard before buying. I'm on pace to hit that.
I also see a failed payment from Clarity Design. I flag that in Slack for follow-up, but I suspect they just changed their card number. These things happen. I'll send them an email in an hour.
2:15 PM - The escalation and the manual work
Back at my desk, there's an email I didn't see this morning because it came in at noon. It's from Marcus Yates at Yates Advertising. He's upset. His team generated three proposals using Pencil, and the customer asked him why all three followed the same structure and format. The customer said it felt impersonal, templated.
He's right. That's the issue. The proposals are generated, not written. The thing I'm selling is speed. But if the customer perceives it as lack of personalization, it doesn't matter that it's fast.
I read his full email twice. Then I call him. Actual phone call. No email can fix this.
Marcus picks up on the third ring. I tell him I'm the founder and I want to understand what happened. He walks me through it. He says the proposals did the job. The customer signed. But it felt close. He says if the AI could inject more of his voice into the output, or if there was an easier way to customize the boilerplate, it would fix it.
I take notes. I tell him I'm going to build a feature that lets his team import brand voice samples so the proposals feel more like his firm. He's quiet. Then he says, "You're listening. That's unusual." I tell him I'm small enough right now that I can actually care about this. He says he'll stick around and see what I build.
I'm down fifty thousand dollars in annual value if he churns. I just spent twenty minutes on the phone and committed to a feature I now have to build or hire built. But Marcus Yates isn't gone. And next month when I ship that voice customization feature, he's going to look like my best advocate to his network.
4:35 PM - The pipeline and the pattern
I open the dashboard again. Today's signups are showing three new users. Two are from LinkedIn outreach. People who found Pencil through the content I posted about proposal templates. One is a referral from someone I don't recognize, which means Carol Reyes' network is already moving.
The week-to-date pipeline is showing forty-two prospects in some stage of trial. I track these in a simple Airtable base that I built myself, one column for each stage of the funnel. Of those forty-two, twelve have already generated at least one proposal. Those twelve convert at seventy-three percent. The ones who don't generate a proposal in the first week almost never convert.
So the game is clear: get people to use it in the first seventy-two hours. That's the whole thing.
I update the dashboard view I share with my team. There's no team yet. It's just me. But I'm building the habits for when there is one. I write a note to myself: "Test new onboarding flow emphasizing first proposal generation by day three."
5:55 PM - The wrap
I open Slack one more time. The new signup notifications are coming in steadily, which never gets old. I check the Linear backlog. Daniel Kim's pricing benchmark ticket is tagged. Marcus Yates' voice customization feature is sketched as a concept card. The bug where templates sometimes drop images on certain accounts is marked in-progress. That's a priority.
I send a quick email to Clarity Design about the failed payment, keep it friendly and helpful. I leave the office files open, the dashboards still running, because tomorrow morning I'll do this again. Check Slack. Check the dashboard. Answer emails. Fix what broke. Listen to customers.
This isn't the fantasy of a fully automated revenue machine. This is real work. I'm here because the AI handles the repetitive part. The first draft of a proposal. And I handle the part that actually requires judgment. Knowing when to approve something. Knowing when to ask a customer why they left. Knowing when to build something because I heard the need in someone's voice.
I close the laptop at six-oh-two. Week-to-date revenue is five thousand eight hundred and sixteen dollars. Sixty-two days until I need to show month-three numbers. I learned something from Carol Reyes and Marcus Yates today. I learned something from Morgan Realty Group going silent. Tomorrow I'll learn something else.
That's the job. That's running Pencil.