In many ways, I am at my happiest when I’ve got a problem or question in front of me, that I can try to unpick by using data.

That’s one of the reasons I love working with Salesforce and AI together. You can ask any question you want of the data and then make informed choices about how to improve it.

Last week I had the opportunity to do a deep dive into Ethan’s performance to work out what tweaks we can make to improve effectiveness.

So what have we learned in the 6 weeks since Ethan was first deployed.

AI Booked vs AI Assisted

Part of the challenge is not making Ethan too pushy. That’s basically the default setting for an AI tasked with closing for a valuation.

The consequence is that Ethan backs off (currently) before I would. When someone tells him that they want to give it a couple of weeks, that’s exactly what he does.

This means that customers who have expressed their dissatisfaction with their agent and want to give it more time, our business development team pick up this warm lead, call it and close it.

Our experience is that we need to make about 40 dials to book one valuation from our normal dataset with that team.

These opportunities we are calling AI assisted are converting at 1 in 2 and we are calling these AI assisted.

Timing Really Matters

We’ve taken a blanket approach to the testing mornings / afternoons and evenings and testing 7 days per week.

The conclusions are that we get 79% of responses outside of office hours.

The best times for sending a message that gets a response are below:

You can see that weekday lunchtime and weekend mornings are clearly the best times. We are now going to be weighting our messages so that a higher percentage are sent during these peak times.

The Potential is Huge

The triggers below are the triggers we have based Ethan’s early testing on. You can see that it’s a limited number. They are high value triggers, but low in volume. Excluding new listing triggers then convert at around 1 in 20 conversations.

As we adopt new triggers our best estimate is that if the ‘high quality’ triggers continue to perform at the same rate then we can move then percentage of valuations to around 26% of all valuations booked at a cost per valuation of £2.40.

When we start to capture the lower grade triggers, and assuming that they are 5x less likely to book a valuation we believe we can move the numbers to 44%.

There is clearly a lot of assumption and extrapolation built into that number, but every step we take will move us closer to that number.

1. My assumption assumes we make the AI no more effective than it is today
2. All of these valuations are generated from the agents’ own database – there is no external marketing
3. The customer continues to respond in the same way as the technology spreads across the industry.

Final Thought

My final thought on this is that there is going to be a huge first mover advantage as businesses on legacy CRM’s scramble to catch up. How long the advantage will last is anyone’s guess only that change is the one thing we can count on.

If you want to look at how AI prospecting would work with your business then you can book a demo here.