Sometimes, when I sit down to write one of these articles, I’m have a sense it’s not going to be a particularly ‘sexy’ topic. I really hope that if you make it to the end, the description of why data matters as part one will make sense.

My co-founder in Greenhouse OS, Alex had this as one of his first conversations with me in 2020.

He tried to explain to me how my current CRM database wasn’t organised well and that there was a much better way to do it.

In the traditional CRM, every seller who contacts us starts life as an applicant with a house to sell, books a valuation then hopefully works their way all the way to be a completion. There the journey ends.

When the current buyer of that property becomes a seller, we start the process again, ending up with many different versions of a property.

Alex made the observation that there are single properties and every property that exists in our patch should exist as a stand alone record in the CRM. So in the Preston Baker CRM we have about 1 million properties. The every house on every street exists in the database – even if we’ve never interacted with it.

We then take all of that old data, where the seller identified themselves as being the owner of that property and attach those people to those properties. We identify that person as the owner of that house.

This is done in real time, inside the CRM so as new valuations are booked, we add more Property Owners. We then check land registry to see when each of these properties sells so we then ‘break’ that relationship as we know that person no longer owns that property.

Today, I know about 73,000 current owners of the properties in my patch. This is live and updated daily.

What have property owners got to do with AI?

The next step is to create triggers or events that would be a reason for a call or a message. There are two big areas:

1. On Market Triggers

These are triggers for competitor stock, where you know the property owner and an event has occurred. That event could be a reduction, fall through or a length of time on the market.

2. Buyer Triggers

These would be things that a buyer has done within your database, such as book a viewing, make an offer etc but critically where they also have a house to sell locally.

How do you convert these into valuations?

This system was built to allow really good, easy propsecting. Opportunities, sorted and filtered in one place, ready to have someone call. But the dates on these reports suggest a problem.

Some of these opportunities sit there for DAYS (maybe even weeks) without anyone calling them.

Team members get busy, and predictably proactive calling takes a back seat. The truth, is that I have nearly 8000 of these triggers that have a call due.

This week we are relaunching our AI negotiator whose job it is to prospect to these opportunities

Below is a text message conversation from our V1 AI. It was good, but it couldn’t do three things that “Ethan” our next generation AI can do.

1. Ethan can now fully book a valuation, creating a time in the diary
2. Ethan can clean the database, removing owners who don’t own the property
3. Ethan can agree a follow up timescale and he can then forward book a task for him to pick back up the conversation

In many ways it’s point 3 that I’m most excited about. What this means is that as we add all of the triggers within 12 months Ethan will be in contact with all 73,000 property owners. Depending on the levels of engagement we get, it will lead to a completely autonomous prospecting system. I have been doing this long enough to know that it will not be as straight forward as that.

The brain has been created, so the medium through which it’s delivered is relatively easy to change. Early evidence points towards messaging being MORE effective than phone calls, so our priority is on building out Whatsapp, SMS and RCS (a more powerful version of text messaging).

Alex has set up the dashboards and we are turning him on next week, so watch this space for updates on how he does.

Once Ethan is working, we will be adding other ‘AI’s into the system.

I would love feedback on which AI’s you think we should be building first

Odi – the viewing feedback AI?
Wren – the sales progression AI?
Mya – the property management AI?
Sienna – the accounts AI?
Talia – the applicant registration AI?
Koba – the tennancy progression AI?

Or any others you can think of! The key is that these AI must be able to actually execute tasks within the CRM, does that mean they have to built by us? In the short term, yes but we want our clients to be able to build their own.