Tom, Dick, and Harry - our AI helpers

Tom, Dick, and Harry - our AI helpers

Rental Property Investor · Member since 2025 · 85 posts · 35 votes

After months of building, testing, and refining AI workflows, we finally developed a model that promises to work! We launched three agents, each focused on its own aspect of STR management, and a team member who loves musicals called them Tom, Dick, and Harry (lol).

In a nutshell:

Tom — Guest Communications & CRM.

Tom operates a dedicated guest‑communication line; his job is to automatically send welcome messages immediately after booking, payment reminders, check‑in instructions, mid‑stay follow‑ups, and checkout guidance. Tom also handles inbound guest messages. If he knows the answer, he responds instantly. If not, he escalates to a human. He is particularly useful in registering and prioritizing maintenance requests and complaints.

Dick — Back‑Office Operations & Analytics.

He is essentially Copilot (comes with the MS Office suite) with a set of unique structured prompts. He scans Outlook, extracting what's important from a sea of spam. Integrated with Excel, Dick tracks financials, generates performance reports, and runs revenue and expense analysis. He also supports creative tasks like social posts, image generation, and website updates.

Harry — Calendar Coordination & Dynamic Pricing

Harry manages calendars, availability, and dynamic pricing. Over time, our goal is to train Harry to orchestrate Tom and Dick — to create a fully automated STR management ecosystem.

As of now, we are excited about watching these guys start working. I'd love to hear from other STR operators who are embracing AI — what models or workflows have proven effective for you?

3Reply
707 views

Most Popular Reply

Ryan MoyerBusiness Member
Property Manager · Orlando Kissimmee Davenport Salt Lake City, Park City · Member since 2019 · 991 posts · 1k+ votes
1w

Which one of them wrote this post?

Next level of training for whichever one it was is to stop using the "--" that all AI agents overuse when writing content.

Cosmic Vacations4.9174 Reviews
See this reply in the discussion

18 Replies

Jump to latestLatest
  • Ryan MoyerBusiness Member
    Property Manager · Orlando Kissimmee Davenport Salt Lake City, Park City · Member since 2019 · 991 posts · 1k+ votes
    1w

    Which one of them wrote this post?

    Next level of training for whichever one it was is to stop using the "--" that all AI agents overuse when writing content.

    Cosmic Vacations4.9174 Reviews
    • Rental Property Investor · Member since 2025 · 85 posts · 35 votes
      1w

      Dick the Copilot, obviously. No secret, we do use AI to polish content, but we never simply cut and paste. Good note about using the "--" , will keep it in mind.

  • Technology · Member since 2026 · 74 posts · 27 votes
    1w

    Interesting way to split the operation, Dan. The part I’d be most curious about is the handoff between Tom, Dick and Harry. Individual agents can work well, but things get more interesting when a guest issue handled by Tom automatically changes something in operations, financial tracking or availability without someone manually passing the context along. That’s usually where I’ve found the biggest value in workflow systems.

    • Rental Property Investor · Member since 2025 · 85 posts · 35 votes
      1w

      The handoff will definitely be the next step. Although a technical solution is available, we do hesitate to give these guys full power over a workflow for now. We will come to that when we are confident they can each handle their assigned areas.

    • Technology · Member since 2026 · 74 posts · 27 votes
      1w

      That makes sense.

      I’d probably keep the agents scoped too rather than give them unrestricted access.

      The interesting piece to me is having a controlled orchestration layer decide what context can move between them, what actions are allowed, and when a human needs to approve something.

      That way Tom, Dick and Harry can stay in their lanes while still working as one system.

      That’s actually the kind of architecture I enjoy building.

  • John UnderwoodPro Member
    Investor · Greer, SC · Member since 2014 · 13k+ posts · 17k+ votes
    1w

    very cool!

  • Andrew SteffensBusiness Member
    Tampa, FL · Member since 2022 · 3k+ posts · 3k+ votes
    1w

    Probably a loaded question but how did you achieve this? We have been using AI tools that come with a lot of our existing software but have not been able to make our own like this.

    • Technology · Member since 2026 · 74 posts · 27 votes
      1w

      Andrew,

      that’s actually the interesting part. We built the agents around specific workflows rather than trying to make one AI do everything.

      The bigger challenge is giving the AI access to the right business data, tools and actions while keeping the important decisions controlled.

      Happy to share how I’d approach it for a vacation-rental operation.

    • Andrew SteffensBusiness Member
      Tampa, FL · Member since 2022 · 3k+ posts · 3k+ votes
      1w

      That would be great - I have no idea how to build AI agents, just use tools like ChatGPT and other AI tools in existing softwares

    • Rental Property Investor · Member since 2025 · 85 posts · 35 votes
      1w

      Andrew - we can connect, I’ll show you some details

  • Rental Property Investor · Member since 2025 · 85 posts · 35 votes
    1w

    CRM and guest communication (Tom) is built on Botpress and Twilio; back office and Analytics (Dick) use the Copilot that comes with Office, and we just spent effort building the prompts. Calendar coordinator and Dynamic Pricing (Harry) - required the most technical expertise; we have an IT resource on the team and used Upwork to hire an expert to build code with Claude.

  • Fayetteville Arkansas · Member since 2026 · 38 posts · 12 votes
    1w

    We have Archie for running our construction project.

  • Member since 2020 · 9 posts · 2 votes
    1w

    Amazing! Giving power back to the individuals is what I love about AI. What problems did you run into (if any) as you added the AI agents? Any hallucinations/mistakes?

    • Rental Property Investor · Member since 2025 · 85 posts · 35 votes
      1w

      A lot of trying and errors at first. Tom, the CRM was easiest as we trained him on standard communications, and instructed to handoff to a human if encounter a request he cannot handle. I anticipate, however, more work once we assign him more complicated tasks. Dick, the Analyst , most of the initial work was to construct adequate prompts, but once done they can be reused and no more work is needed. Harry, the Calendar and Pricing is our main focus as of now. Still work in progress and will take time to ensure he does not double book (right now we verify everything manually) and prices optimally.

  • Englewood, NJ · Member since 2018 · 461 posts · 82 votes
    1w

    Dan, interesting setup with Tom, Dick, and Harry. I'm not in STR but I do use a similar approach for tax deed auctions in Florida. I wrote Python scripts that pull property data from county websites, analyze things like tax balances and assessed values, and score which properties look distressed enough to bid on. It's not as organized as giving each agent a name, but the principle is the same: break a big workflow into smaller pieces that each do one thing well.

    The handoff problem Paul mentioned is something I run into too. My scripts can flag a property as a good candidate, but I still have to manually verify the title status and check if there are any liens before I actually bid at auction. The automation can get you 80% there, but that last 20% of judgment still needs a human.

    Curious how Harry handles the dynamic pricing. Are you feeding him comp data from AirDNA or similar, or is it more based on your own historical booking data?

    • Rental Property Investor · Member since 2025 · 85 posts · 35 votes
      1w

      For dynamic pricing, AirDNA is not reliable; I got a better solution. Scrubbed comps actual booking data from AirBNB and VRBO (two years worth of data) and applied a forecasting algorithm.

  • Real Estate Consultant · Melbourne, FL · Member since 2019 · 205 posts · 113 votes
    3d

    How does Tom get somebody's attention when he hands something off? A guest locked outside needs more than a message sitting in an inbox. Curious whether you've got a phone alert or someone actively watching it.

    • Rental Property Investor · Member since 2025 · 85 posts · 35 votes
      3d

      We made it simple, actually. Once a guest tells the bot, "I want to talk to a human" the conversation is forwarded to a property manager on duty as a text alert. In the next phase, we plan to elaborate more, i.e categorize inquiries into those needing an immediate response and those that can be queued and answered in the order it were received and during regular work hours.

Join the conversationCreate a free account to reply, vote on answers and follow this thread.