iBooked Online

G’day Accommodation Professionals,

We’re back for a brand-new season of The Accommodation Show! Welcome to Season 24.

Over the break, one topic dominated almost every conversation I had with property managers: where is this AI wave actually going, and how do we use it in our day-to-day operations without breaking our business?

Right now, most operators are stuck in one of two camps. Either you’re using ChatGPT as a polite copywriter to reword guest messages, or you’re falling down rabbit holes at 2 AM trying to “vibe code” custom tools you don’t fully understand, only to spend days fixing broken scripts to save two minutes of daily work.

The gap between basic chatbots and real, functioning automation is wide. But deploying autonomous workflows doesn’t require a computer science degree—it requires knowing which tasks to hand over, which vanity metrics to ignore, and where your business data actually lives.

To kick off Season 24, I sat down with Jan Sahagun, co-founder of Trellis. Jan recently came through Y Combinator in San Francisco, where his team has been building autonomous AI agents specifically designed to handle short-term rental operations.

Jan’s first filter for automation is what he calls the 20-minute rule. If a task only takes a team member two minutes to do, leave it alone. Operators waste days building and troubleshooting automations for minor tasks, creating maintenance headaches that burn far more time than they save. Real efficiency comes from targeting the repetitive, friction-heavy workflows that quietly eat 20 to 30 minutes every single day, and using tools to compress that work down to 30 seconds.

He is equally blunt about vanity metrics. Bragging about a 100% automated guest messaging rate sounds impressive on paper, but it is often an operational trap. If an automated response fails to solve the guest’s immediate problem or dings your review score, that automation worked against you. The metrics that actually matter are response and resolution times, guest sentiment, and protecting your review average—not whether a human never touched the keyboard.

Underpinning all of this is where your data actually lives. An AI tool connected to Slack or your inbox without direct access to your PMS is running blind. The PMS is your source of truth—holding live calendar availability, guest contact details, and reservation histories. By using open standards like Anthropic’s Model Context Protocol (MCP), you can bridge that gap, allowing models like Claude to safely read data from your PMS, dynamic pricing software, and internal spreadsheets before taking action.

What you’ll learn in this episode:

• 🤖 The difference between a simple chatbot and an autonomous AI agent

• ⏱️ The 20-minute filter for choosing what to automate (and what to leave manual)

• 🔗 How to connect Claude directly to your PMS data with MCP

• 📊 Why chasing a 100% messaging automation rate hurts your reviews

• 🛠️ Building internal dashboards (like cleaner scorecards) without writing code

• 🧠 Outcome-based prompting advice direct from Y Combinator

•  

Jan Sahagun is the co-founder of Trellis (trellispm.com), an AI platform built to run autonomous operational workflows for property managers. A recent graduate of Y Combinator, Jan works with accommodation operators worldwide to connect AI tools cleanly into everyday management stacks.

It is great to be back with you for Season 24. We have an incredible lineup of operators, founders, and practical operational breakdowns planned for the weeks ahead.

📺Watch the episode here