As both Drew and now Nathan have written, there is a lot of nascent interest in bots for real estate. Most technologists tend to focus on the consumer-facing use cases: helping buyers figure out which homes to look at, answering basic questions about listings, etc.
At Virgent, we took a little bit of a different tack. Since we exclusively focus on home sellers, we handle a lot of showing requests from buyers’ agents – in some cases it can be dozens of requests a day.
It only took us two weeks after launch to determine that coordinating showings via phone and text message was not super scalable. We knew we needed an automated system to handle the throughput, but we also knew we had to be very careful in developing it to ensure that buyers’ agents would use it instead of making a phone call or sending an email and that it was dead simple for our clients to use as well.
In order to do this well, it meant really understanding our user, who in this case was the buyer’s agent. This person is extremely busy, and almost always scheduling showings while on the go. They don’t have the time or interest in downloading a new app, but they do want the latest information on the subject listing instantly (“Does it have any offers?”)
We landed on the following guiding principles (consultants call these the Business Requirements):
- The bot must accept natural language
- The bot must be entirely mobile, without requiring agents or our home sellers to download anything new
- The bot must be quick to use – no more than 3 “screens”
- The bot must be flexible enough to handle different occupancy scenarios (e.g. Owner Occupied, Tenant, or Vacant) and adjust the showing confirmation process accordingly.
With that in mind, we built a fully automated, text-messaging based bot that now handles about 80% of all the showing requests we receive. Here’s how it works:
How Our Bot Works
Properties we list are auto-assigned a text-message capable phone number which is noted in the private remarks of the listing. Buyer’s agents can text any sort of showing request to the number. For example:
“I’d like to show this home tomorrow at 3pm”
“Saturday, May 16 @ 12:30pm”
“Is the home available for showing at Friday at 2?”
Our servers parse the agent’s request, turn it into a standardized time format, and send back the current offer status of the house (e.g. the house does or doesn’t have active offers on it).
In the case of an owner- or tenant-occupied property, our bot immediately sends a text message to the resident asking them if they can be out of the house at the requested showing time. A “yes” response triggers confirmation texts and emails to both parties, as well as other information the buyer’s agent needs to know, such as alarm codes.
A “no” is passed back to the agent for rescheduling along with better times for the resident.
If the property is vacant, the showing request is automatically approved and the homeowner is notified via email that a showing has been scheduled.
About 15 minutes after the showing ends, our system automatically texts the showing agent and asks for feedback on the property. Any response is instantly shared with our home sellers and our team via email.
The results? Our average showing is confirmed about 5 minutes after it’s requested, entirely via text message, without any intervention by our team.
We were actually surprised by the reaction we got once we released the system. This bot is one of the most noted features in our reviews on Zillow:
They have an amazing feature that will text you when an agent wants to show your home and you can respond with a time that works best.
The text-to-show feature was excellent because it allowed us to share showing information about our 3 dogs and alarm codes codes directly with selling agents
It’s always great to get that feedback, but personally I was even happier that our bot passed what I’ll call “the poor man’s Turing Test.”
For those of you who aren’t familiar, the Turing Test is a test of a computer’s ability to simulate a human’s behavior so the user can’t tell if they’re talking to a human or a computer.
After launching the bot, we had a number of agents contact us afterwards following up on their discussion with the bot, which they had assumed with a member of our team on the other end.
More importantly, our counterparts on the buying side have told us the bot saves them a ton of time in scheduling showings, gives them the info they need instantly, and makes the showing process much less hectic.
The Technical Nitty Gritty
For those of you who are curious, we built the system using Twilio, which is a service that can programatically handle incoming and outgoing text messages and phone calls. The platform is fully integrated in our proprietary software, built on Ruby on Rails, so all showing data is stored in our systems and available to our clients online 24/7.
Shoot me a note if you ever want to talk about Twilio or RoR.