At a recent RSU by PriceLabs webinar on June 24, 2026, Boostly founder Mark Simpson laid out how a short-term rental property manager with no coding background can build their own AI agents. He mapped it as a five-level climb, from basic prompting up to a team of AI agents for property managers that handle tasks on their own.

The climb matters because the gatekeeper is gone. Building your own AI used to need a developer; now you can do it in plain English, with no coding background. The cost did not vanish. It moved from a salary to usage: the tokens and connector fees that add up as agents run. What changed is who can start, so the edge now goes to the operators who learn fastest and keep those costs in check.

Most Property Managers Are Still on the First Rung

Simpson calls it the “Jack-In Framework,” a nod to The Matrix. It has five levels. At the entry level sits what he calls ping pong: you ask a question, you get an answer, and you do all the steering. A poll at the session showed most attendees are on the first rungs: 63% use a tool like ChatGPT for writing and basic tasks, 24% run custom workflows, and only 7% use agents that act on their own.

The framework is Simpson’s, and it is a useful map. One caution before you climb: higher is not automatically better for your business. Pick the rung that solves a real problem, then move up one.

Also Read: AI in Vacation Rentals: Build Your Own AI or Wait for Your PMS? 

Level 1: Train the AI on You So It Stops Sounding Generic

Most people take whatever generic answer the AI gives. Level 1 fixes that.

• Custom instructions. Tell the AI once who you are, your rules, and how you write, so its answers fit your business.
• Custom GPTs or projects. Build a version trained on one job. Feed it your best emails and it becomes your copywriter.
• A free trick: tell it to rate its own answer from one to 10 and ask you a couple of questions before it replies, so it stops guessing.

For a rental manager, that means a GPT trained on your house manuals and past messages that drafts guest replies in your voice, or one fed your best listings that writes a fresh listing description and house rules in minutes.

This is the highest-return, lowest-risk level, and the one most operators skip on their way to chasing agents.

Level 2: Turn One Document Into Slides, Audio, or Video

This level uses NotebookLM, a free tool from Google. Drop in one source, like a call recording, and it produces a slide deck, an audio summary, or a short video.

A real use: take the transcript of a call with a homeowner you want to sign, and turn it into a clean recap to send them. It makes a small operator look organised. Other uses: turn your operations manual into a short audio guide for onboarding a cleaner or virtual assistant, or turn a monthly performance export into a tidy slide deck for an owner.

Level 3: Build Simple Guidebooks and Sites From a Prompt, Then Hit a Wall

Here you use AI builders such as Lovable or Replit to make things from a written request: digital guidebooks, one-page sites, simple bots. Some operators built niche listing pages for individual 2026 FIFA World Cup host cities to catch fans searching for a market. Closer to home, that could be a pre-arrival upsell page offering early check-in and mid-stay cleans, or an FAQ bot for your direct-booking site.

The wall is real. Simpson compared these tools to a beautiful car with no engine. They are great for small, throwaway jobs and cannot run your booking website or the channel manager that keeps your calendars in sync across the booking sites. Do not mistake a slick demo for something you can put into production.

Level 4: Hand Routine Tasks to an AI That Works on Its Own

Now the AI starts doing work for you, using connectors (also called MCPs) that let it log into your other apps with no code.

Simpson described an operator whose agent, nicknamed “Big Lou,” handled a late-night broken-plug report end to end: it read the guest’s message, contacted workmen, booked the repair, and coordinated times, all while the operator slept. Other jobs managers hand to a Chief of Staff: a Monday email that pulls occupancy and revenue from all your tools into one report, an agent that flags new reviews needing a reply, or one that chases a cleaner who has not confirmed a turnover.

The limit: an agent at this level can only use connections that already exist. If your tool has not built a connector, the AI cannot reach it. It is a walled garden. Purpose-built software does versions of this out of the box, the kind of AI tools that automate guest messaging and maintenance RSU has covered before.

Level 5: Build a Whole Team of Agents Yourself

This is the deep end, and it removes the Level 4 ceiling. You run your own setup and connect anything you want.

Most people start with OpenClaw, a free, open-source tool from developer Peter Steinberger (hired by OpenAI in early 2026). It is a blank canvas, so nothing is set up for you.

• One lead agent runs the others. Simpson built a single “manager” agent and let it create smaller agents under it, each with one job.
• You do not need to code. Whenever the setup asked for a technical step, he screenshotted it, pasted it into the AI, and asked what to do next.
• Give each agent one narrow job. A clear task like “find old leads” works. A vague brief like “grow my business” does not.
• Never give an agent money access. No bank logins, no card numbers, no crypto. He named his lead agent Icarus, after the figure who flew too close to the sun, as a reminder to keep it in check.

Managers run these for owner prospecting, review responses, and turnover coordination, with one agent owning each job. Be honest with yourself about this level. It breaks, it costs money, and most managers get real value long before they reach it. Treat Level 5 as optional, not the goal.

In the Demo: Agents That Earned Their Keep

The payoff of the session was watching agents do real work. Simpson shared a set of 21 plug-and-play agents built for short-term rental operators that other managers can copy. His sales agents found a lead his company had failed to close two years earlier, researched that operator’s market, sent a tailored message, and booked a call that converted.

By his account, that setup brought in more than $60,000 over 90 days against roughly $5,000 in cost. Those are his own numbers for his own business, and he was candid about the broken builds and wasted hours along the way. Read it for the pattern, not the figure: one agent, pointed at one repeatable task, can run on its own.

Start With One Task You Shouldn’t Be Doing

You do not need a plan. You need one task to hand off.

1. For seven days, write down every digital task you do, on the hour.
2. Score each one. A 1 is critical and only you can do it. A 2 is important but you could delegate it. A 3 is something you should not be doing at all.
3. Your first agent takes over a 3.

Two rules before you start: never give an AI your passwords, bank details, or private guest data, and always review what it produces before you send, publish, or automate it. Do not compare your setup to anyone else’s.

Here is the part that compounds. The first rung is the hardest, because you are learning the tool as much as the job. Each level after that comes faster. Climb one rung this summer and you gain more than saved hours. You gain the skill of climbing, which pays off every time a new tool lands.