Day 1 operating ManufactureAI · Wishdeal Factory

8:42 AM - Dashboard and morning triage

I open the ManufactureAI dashboard before I've finished my coffee. This is still the part that feels good. Week-to-date: 23 new signups, four converted to paid in the last 14 days. Today's carryover is sitting in the queue: 47 quoting requests waiting for agent review. I've learned that 47 is manageable. Seven months ago, that number would've crushed me.

The Slack notification from my integration pinged at 6:15 AM. It reads: "Flagged: 2 quotes require manual review before sending." I click through to the dashboard. One's a false alarm - the system caught a material substitution that was actually fine. The other is Marcus Chen at Chen Precision Metal Works requesting a custom lead time that breaks our standard template. I'll circle back to Marcus in my review queue.

Inbox: 12 new emails since I left yesterday at 6. Three are customer responses to quotes their shop didn't request through the system - they came through email directly. That's the gap I haven't fully closed. The AI only processes requests that land in the web form. These outliers are still on me. I save those threads to a folder labeled "Manual Follow-up."

One email is from Patricia Hoffman at Hoffman Manufacturing, a shop that signed up three weeks ago. Her subject line is "Thank you. This is working." I open it first. Patricia says our system quoted a complex aluminum bracket assembly in 22 minutes that used to take her 90 minutes to price manually. She lost that job to a competitor last quarter because her quote came in two days after the customer needed it. She's already recovered the software cost on one new contract. The email is short and direct. I read it twice. This is why I'm doing this.

The fourth email is flagged: "Billing issue." I forward it to my spreadsheet where I'm tracking customer issues that don't fit neatly into the product. This is Rebecca Liu at Liu's Tool & Die. She's three months in, paid her first invoice through Stripe cleanly, but the second month she says our invoice landed in her spam folder and she doesn't want to miss a payment. She's asking if I can send it to her partner instead. No automation for this yet. I make a note to reply before lunch.

10:15 AM - A flagged conflict

The Marcus Chen quote is more complicated than I thought. His shop does aerospace-spec work. The system quoted him on a standard aluminum alloy and a 14-day lead time. Marcus needs a titanium variant with 21-day turnaround for a customer's certification. Our pricing template doesn't have titanium in it yet. This is the kind of thing the system should learn to flag, but it doesn't quite yet.

I open Google Sheets where I'm tracking edge cases. This is my informal bug list. Titanium pricing gap is already there with three other instances. I pull Marcus's quote, adjust the material cost by hand - I know his shop's markup from his intake form - and rewrite the lead time section with a note explaining the certification timeline. I add my name to the quote and send it to him directly through Gmail, not through the system.

"Marcus, attached is your revised quote. Given the titanium spec and your certification needs, we've factored in a 21-day window. If you need faster, we can discuss it. Let me know if this tracks with what you're scoping."

This is the work that doesn't scale. But it's also the work that kept him from churning. I open Linear to create a ticket: "Titanium pricing model - add to material database." I tag it as medium priority. There are three other material gaps below it. One day I'll hire someone to own the material library. Today, it's me.

12:32 PM - Lunch and the metrics check

I step out for a sandwich and eat it at my desk because Tuesday afternoon is when Stripe sends the daily revenue summary and I want to see it come in. Today's revenue from new billing is $597. Three customers who converted last week are on their first billed month. One of those is Patricia from this morning. Her small shop is on our $199/month tier. The other two are on $299. The system is working.

Week-to-date pipeline: $8,400 in potential monthly recurring revenue from leads in active conversations. That's 14 shops currently in some stage of evaluation. Five of them went through free trial and are in close calls. The largest is Carol Reyes at Reyes Fabrication. Carol's been in the trial for six days. She's run the system on five of her recent RFQs and is still deciding. Her potential ARR is $2,400 if she converts.

I pull up my follow-up schedule in Calendly. I have a 30-minute call with Carol at 3:15 PM to walk through her trial experience. I should prep for that. But first, the email to Rebecca Liu.

I open Gmail and draft the response: "Rebecca, I got your note about the invoice going to spam. I'm going to add your partner to the billing contact list in your account. From next month, you'll both receive the invoice. Can you send me their email address so I can add them properly?"

It's 47 words. It could be a Zapier automation. I could template it. Instead I'm writing it fresh because Rebecca's anxiety about this matters more than my efficiency. She's a three-month customer. She's keeping her renewal on track. I send it.

3:15 PM - Carol and the close

Carol's on the call at exactly 3:15. She's direct. She's tried the system on five jobs. Four of them matched her internal estimates exactly. One was 8% lower, which made her nervous. She asked her team to verify it and they confirmed it was accurate - she'd been overpricing that job class.

"The question I have," Carol says, "is whether I can trust this when I've been doing it my way for eight years."

This is the objection I've heard four times now. It's not really about the price. It's about control and confidence. I tell her what I've learned: the system is trained on thousands of actual job completions from shops like hers. It misses outliers. It can't read customer desperation or market timing. But on spec and material, it's more consistent than human estimation. I ask her: in those five jobs, if four were accurate and one made her smarter, what was the cost of her old process - in time or lost bids.

She doesn't answer directly. She asks if she can talk to one of our other customers. I know Patricia Hoffman would take a call. I offer to introduce them. Carol agrees to a one-week trial extension while she talks to Patricia and makes a decision.

I follow up the call with an email to Patricia asking if she'd be willing to speak with Carol. Patricia responds in 11 minutes: "Absolutely. Tell her to call me." I forward Patricia's number to Carol and cc Patricia. That's the closing pattern I've learned works: peer validation beats sales pitch every time.

4:45 PM - The bug and the day's final push

A customer, David Torres at Torres Manufacturing, reports that his quote history isn't showing three jobs from last week. I log into his account and see the issue immediately. The filter is set to "This Month" by default, and our system date settings rolled over last night - he's looking at May when those jobs were assigned in late April. It's a UI bug, not a data loss.

I check my Linear backlog. I can see this is the second report of this exact issue. I open the dashboard code locally and find the date filter logic. The bug is in line 487 of the quote-list component. The default filter should be "Last 30 Days," not calendar month. It's a one-line fix. I change it, test it locally, and push it to staging. The fix deploys in six minutes.

I send David a quick email: "Hey David, we found a bug in the date filter. It's fixed now. You should see all your quotes when you log back in. Sorry for the confusion." David's the type who doesn't respond to service fixes - they're just baseline. But he won't churn because of it, which is the point.

6:15 PM - The close

I close the laptop at 6:42. The dashboard shows 49 new signups today and $3,200 in new ARR committed for next month from conversions that closed. One customer - Elena Rodriguez at Rodriguez Sheet Metal - churned yesterday. She said the system couldn't handle her custom edge-case quoting rules. I marked it down. That's data I need.

The week-to-date pipeline is now $9,100. Carol's decision comes in two weeks. Marcus's updated quote is sitting with him. Rebecca's billing reconciliation is handled. Patricia's word-of-mouth is working. The material library still has gaps.