8:42 AM - Inbox triage
I open the Relay dashboard and pull my coffee closer. The metrics widget shows 14 signups overnight - about normal for a Tuesday. Revenue counter reads $3,847 for the week so far. I check Slack and there are three alerts waiting: two customers have approved their handoff packages and are ready for export, and one flagged a conflict that the agent couldn't resolve. That one always gets my attention first.
My email inbox has overnight notifications from Stripe and two customer messages. The Stripe one is routine - a retry on a declined card went through. The customer messages are different: one is a thank-you note from Carol Reyes, who bought Relay last week, and one is someone who bounced off during signup asking if we support Figma plugins yet.
I read Carol's email first. She's a product designer at a small studio in Portland, and she used Relay to wrap up a rebrand project for a local real estate firm. She's saying the asset inventory alone saved her three hours, and the support terms block gave her a template she's now using for all her clients. That's the email I needed this morning.
The bounce email I'll flag in Linear as a feature request. We don't support plugins yet, and it's on the roadmap for Q3, but I want to make sure Rui sees the pattern if we're getting multiple asks. I add a note: "Third plugin request this month. Prioritize when planning Sprint 7."
10:15 AM - A customer in limbo
The flagged conflict is from Maya Torres, a freelancer in Austin who's wrapping a Shopify site redesign for a local retail client. The agent inventoried her GitHub repo fine - found the theme files, the custom CSS, the deployment notes - but when it tried to draft the maintenance playbook, it flagged something it couldn't verify: the original developer left inline comments saying the custom checkout flow "needs a rewrite in six months," but the client has no record of that conversation.
I pull up Maya's project in the app and read the agent's draft note: "Unable to confirm scope of future work. Recommend clarifying with client before finalizing handoff."
This is exactly the kind of escalation the agent should catch. If Maya hands off without clarifying that the checkout code is temporary, the client will be surprised in six months. I fire off an email to Maya: "Hi Maya, the system flagged something in your GitHub comments that might be worth a quick clarification call before you deliver. The custom checkout code has a 'rewrite needed' note in it, but your client contract doesn't mention that timeline. Want to clarify that with them before you hand it off, or mark it as post-launch scope?"
She replies within ten minutes. Apparently she and the client had talked about it informally but never documented it. She wants to add it to the playbook under "Year 1 considerations." I update her draft in the app and mark the conflict resolved. The agent re-processes and confirms the section now looks clean.
11:47 AM - Metrics and the churn notice
Before lunch I pull up the Relay metrics dashboard - a habit now. The week-to-date pipeline shows $28,000 in customers who've connected a project but haven't approved yet. That's solid. Active trials are 34, up from 28 last week. Churn rate is holding at 2.1 percent, which is fine for a product this new.
But then I see the churn email that came in overnight. A customer named Dmitri Volkov in New York bought Relay two weeks ago, ran one project through it, and decided it wasn't for his workflow. His reasoning: he said the agent's asset inventory was too comprehensive for his needs - he prefers handoff packages to be minimal and client-focused, not exhaustive.
I read his cancellation request carefully. He's not angry, just not the right fit. His workflow is literally the opposite of what Relay is built for. I approve the refund in Stripe and send him a personal note thanking him for trying it, wishing him well, and letting him know if he ever needs something more comprehensive, we're here. I also flag his feedback in Linear as validation for a "lightweight mode" some users have requested.
That one stings a little, but it's the cost of niche software. Not everyone will be a fit, and I'd rather know now.
12:47 PM - The manual workaround
A customer named Jeff Chen emails asking if Relay can integrate with his custom CMS where he stores design tokens and component documentation. It's a legitimate question, but it's outside our scope right now. We work with Figma, GitHub, Google Drive, and Notion. His CMS isn't on that list.
Instead of just saying no, I write back: "Hey Jeff, we're not connected to custom CMSes yet, but here's what I'd recommend: export your component docs as a Markdown file and drop them in a GitHub repo folder alongside your Figma file. When you run Relay, the agent can read both and weave them together in the inventory. Does that workflow make sense?"
He replies an hour later saying that would work. Problem solved without building anything new. That's the kind of small thing that builds goodwill and keeps customers thinking of you as someone who actually cares about their use case.
I add this pattern to a quick operational note for myself: "Customers often hit walls that aren't product bugs. Take five minutes to propose a workaround before escalating as a feature request."
2:30 PM - A clean handoff
Two customers have both finished their reviews and approved all four sections of their packages. One is getting exported to PDF, the other to a Notion page. The agent does the assembly in real-time, and I spot-check each one before they leave the system.
The first customer is a visual designer in Vancouver who built a component library for a brand studio. Her inventory is detailed, her usage guide is warm and human-written, her playbook lists quarterly audits and annual design system refreshes with realistic time estimates. I review it, click approve, and it generates. The PDF downloads in three seconds.
The second is a developer in Berlin who wrapped a SaaS dashboard rebuild. His inventory lists all the new API endpoints, his guide is technical, his playbook covers database migrations and deployment procedures. Straightforward. I approve that one too.
Both customers get an automated email from Relay with their package link and a note that I'm available if they hit questions. I also see in my analytics that both are now in the "post-handoff" cohort. I'm curious if they'll come back and use Relay again for their next project.
4:15 PM - The pipeline review
I pull up Linear and review the seven open issues for this week. Four are feature requests from customers - plugin support, Webflow integration, better API documentation. Two are bugs that Rui flagged as low-priority. One is a sprawling roadmap item for "teams and billing tiers" that keeps getting pushed because it's big work.
I update the status on the Figma plugin request to "prioritized for Sprint 7" based on the feedback pattern I saw this morning. I also note that three customers have asked about team plans in the last two weeks, which is the first real signal that demand is there. I draft a Slack message to Rui: "Team billing is moving up the priority list. Three customers asked about it this week, and the plugin request is now top three. Want to scope it together tomorrow morning?"
I also check the weekly cohort report. Of the 14 signups yesterday, seven have already connected a project. That's a 50 percent conversion from signup to first action, which is good. Of last week's 18 signups, 11 converted to a paid trial. That means our organic GTM motion - Twitter posts about freelance workflow, ProductHunt engagement, design community Slack channels - is attracting the right people.
5:50 PM - The one thing I'd change
I close the Relay dashboard and reflect on the day. The work today was exactly what I expected: reviewing agent output, handling the 5 percent of cases that needed a human judgment call, watching metrics, and doing one real piece of work - the manual workaround I built for Jeff.
If I'm honest, the part that felt inefficient was the flagged conflict from Maya. I had to write an email, wait for her response, manually update her draft in the app, and trigger the agent to re-process. That loop took 45 minutes. In another month or two, I want to build a faster resolution flow - maybe a quick-reply interface in the app where customers can answer yes/no questions and the agent just re-runs the flagged section without me needing to step in.
The part that felt great was Carol's thank-you email. And Dmitri's churn - even though it stung - taught me something about who we're not built for, which is just as valuable as knowing who we are.
I have 34 customers in trial right now. If even half convert this month, we'll be at roughly $16,000 MRR. That's the number I'm watching. The infrastructure is stable, the product works, and the operator work - the part I do - is manual but manageable at this scale. At 100 customers, I'll need to hire someone to handle escalations. At 200, we'll need to automate the conflict resolution flow I described.
For now, it's sustainable. I close the laptop at 5:52 PM and feel like I earned the evening.