8:42 AM - Inbox triage
I'm at my desk with coffee, opening my laptop. The HotelAI admin dashboard is already open from yesterday. I scroll to the top.
4 new signups overnight. Revenue shows $782 so far this month. Week-to-date pipeline: 8 active demo conversations, 3 in "decision" stage, 2 in "negotiation" over pricing.
My Slack channel - configured to receive product alerts - has three unread messages. One is a customer flagging an issue. One is a trial-expiration notice. One is a billing edge case I didn't write.
First, email. Gmail inbox shows 12 new messages since last night. Filtering to customers only gives 6.
One is from Carol Reyes at the Reyes Family Bed & Breakfast in Santa Cruz. She's been on the product for three weeks. Her subject line: "Guest messages getting repetitive?"
I open it. She's right. The agent is using the same closing line in guest responses: "We're so glad you chose to stay with us." Four out of five recent messages have it. I see why that would feel off.
I check her configuration in the admin UI. It's a template setting I can fix in two minutes. I reply to Carol: "Saw this. I'm checking your settings right now and will have a fix ready today. Thanks for catching it. The agent learns from feedback, and you're helping us improve for everyone."
That's the owner work. I'm not hiding behind "the system does this." I'm here.
Second email is from Thomas Kwan at the Westside Inn. He signed up four days ago. Subject: "When do you charge us?"
Thomas is asking about billing. He's on a 14-day free trial ending Friday. The email is polite but cautious. He wants to know if he'll be automatically charged, what the process looks like, and if he can upgrade to annual.
I open Stripe. He's attached a card but hasn't been charged yet, which is correct. I reply with the specifics: we charge monthly on the 15th, he can modify his plan before Friday, and we offer 15% off annual plans. I include a Calendly link to a quick pricing call.
That's another piece people don't see. A SaaS engine running on automation still needs a human to confirm details, reassure the customer, and make them feel like they're talking to a real person.
10:15 AM - A flagged conflict
I toggle back to the admin UI. Under "Agent Outputs Pending Review," one item is flagged: "Needs Human Decision."
It's a review response drafted by the agent. Context: Rachel Kim at the Portland Modern Hotel received a 3-star review on Google. The guest complained about noise from a neighboring room. The agent drafted:
"We're sorry you experienced noise. We've since installed soundproofing in all guest rooms and take guest comfort very seriously."
The flag is correct. We haven't installed soundproofing. The agent hallucinated a solution. If Rachel sends this, the next guest checks in expecting something that doesn't exist.
I click "Revise" and rewrite it:
"We're sorry you experienced noise during your stay. Guest comfort is a priority for us, and we take feedback like this seriously. We're exploring solutions to improve the soundproofing in our guest rooms and would love to speak with you directly about your experience if you'd like to discuss it further."
It's honest. It doesn't over-promise. It invites dialogue.
I approve the revision. The system notifies Rachel that the review response is ready to publish. The agent doesn't miss reviews, and it gets the tone right 85% of the time. That 15% where it hallucinates or misrepresents. That's why I'm here.
11:50 AM - A new booking
Slack pings me. New demo booked. I tap the notification, and it takes me to our Calendly integration.
Prospect: Gabriela Santos, owner, Santos Suites, Las Vegas.
Thursday, 2 PM. I skim her intake form. Small property, 24 rooms, she's answering guest messages manually in the mornings before housekeeping duties start. Textbook HotelAI customer.
The system has already sent her a confirmation with demo details. I don't need to do anything except show up Thursday.
But I make a note. This week's pipeline is solid. 8 demos, 3 of them with decision makers who've stated budgets. If even 2 close, we're looking at $250-300 in new MRR this week.
That's the work that matters. The actual pipeline conversion. That's my real job.
12:30 PM - Lunch and the metrics check
I step away for lunch but my laptop stays open. While I eat, I pull up our tracking sheet - a Google Sheet with month-to-date metrics.
- Total customers: 47
- New customers this month: 9
- Churn this month: 1
- Current MRR: $6,840
- Projected MRR by month-end (if 2 demos close): $7,340
The churn sticks with me. Frank Morrison at the Cornerstone Inn canceled yesterday. His feedback: "The agent responses don't match our voice. It's too formal for a quirky, indie hotel."
I replied to Frank this morning: "I get it. This kind of personalization takes configuration. I'd like to help you dial it in before you go. Can we schedule 15 minutes?"
He hasn't replied. Churn happens. But it reminds me that the product is only as good as the fit.
2:08 PM - Customer escalation
Back at my desk. My phone buzzes. A customer call coming in through Slack.
It's Patricia Hoffman at the Beacon House Hotel in Boston. She's been a customer for 8 weeks. Friendly woman, quick to adopt the product, good revenue at $160 per month.
Her issue: "The housekeeping coordinator sent a message to a guest at 11 PM last night asking about their laundry request. The guest was upset. Can the system be set to quiet hours?"
The agent doesn't have a built-in quiet-hours feature yet. I take the call and walk Patricia through a workaround. She can set her team's manual notifications to follow a schedule. I'll flag this as a feature request in our Linear board.
I pull up Linear while on the call. I add a new ticket: "Feature request: quiet hours for housekeeping notifications (9 PM to 7 AM default, customer-configurable)." I tag it P2 and link Patricia's name.
She feels heard. I'm not making promises I can't keep, but I'm showing her that her feedback goes into the roadmap in real time.
4:30 PM - Pipeline review
I block 45 minutes to review this week's demos and pull together an update for myself.
Of the 8 demos this week:
- 3 are hot (detailed pricing questions, onboarded their teams)
- 2 are warm (interested, asking about competitors)
- 2 are cool (seemed interested in the call, haven't replied to follow-up)
- 1 is dead (booking was a mistake, they run a managed property)
For the 3 hot ones, I draft follow-ups. Amber Costa at Costa Family Lodging in San Diego asked specifically about integration with her existing staff scheduling tool. I checked our integrations and we don't have it, but we have an API. I reply:
"Great question about that software. We don't have a native integration yet, but our API is solid and several customers have built lightweight webhooks to sync guest requests with their scheduling system. I can connect you with a reference if you'd like to see how it works in practice."
This is the honest demo-to-close work. Not overpromising. Showing a path forward. Making the prospect feel like they're solving a real problem, not buying a feature checklist.
6:15 PM - Wrap
I close the demo notes and check the metrics one more time before I shut down.
Today:
- 4 new signups
- $782 month-to-date revenue (tracking toward $7,500 by month-end)
- 1 customer issue fixed (Carol's template)
- 1 agent hallucination caught (Rachel's review)
- 1 escalation handled (Patricia's quiet hours request)
- 3 hot demo leads for this week
It's not glamorous. I'm not sitting back while a robot does the work. I'm reading emails, reviewing agent outputs, catching mistakes, pushing the pipeline forward. Some days it feels like babysitting. Some days it feels like running a small business, which is what this actually is.
I close the laptop at 6:18 PM.
The thing I'd change: more time to personalize each customer's onboarding instead of jumping from firefight to firefight.
The thing that worked: showing customers I'm real, accessible, and responsive. That's worth more than perfect automation.