Out Nurture's Story

📖 5 minute read

Out Nurture started as a conversational AI agent that texts leads and hands them off to humans once they're "ready." It featured intelligent scheduling and follow-ups so that leads don't "fall through the cracks."

Since then, we've learned a few things. One is that this works great. Leads are addressed according to their inquiry, rather than obviously automated drip campaigns, which frustrate people. Another thing is that there are many companies and CRMs trying to do this exact thing. The only one which I can say is exactly what we were doing is Apten. Knowing that there was a company dedicated to outbound conversational AI with the same level of quality as us was important. It let us focus on what we're really trying to do, not just the bottleneck.

Let's talk more about what I mean by "bottleneck." Without automating conversations, any grand marketing efforts would be limited by the number of human employees! Sure, you could force leads into traditional scalable processes like VAs, endless forms, or call centers, but this is expensive and a bad experience for the lead.
Even worse is that people have two bad traits when it comes to following up:

  1. Don't handle rejection/no response well (patience)
  2. Don't research the lead in detail to personalize each message (diligence)

We experienced this with our real estate clients as a marketing agency. Once we helped them get their online advertising working, they thought money would just...appear. There was this expectation that getting leads meant getting customers because this had been their experience to date. However, online leads are colder and less certain about what they want and how they want it. Many real estate agents got frustrated and even burnt out from "following up" on these leads because they were "bad leads."

It gets worse: I even heard from a real estate Google Ads veteran that

The average time from "click to close" was over 700 days.

That means real estate agents would need to follow up for, on average, 2 years to get a return on ad spend. Imagine us as the marketers trying to sell Google Ads services with numbers like that. Of course, this isn't the case with all real estate ads across every market and offer, but the pain is real: Who is going to follow up for years? What happens if they decide last minute they're going to go with their cousin who's a real estate agent?

The point here isn't to dissuade from running ads, shame "lazy" realtors, or just complain. It's to observe the bottleneck in this process isn't paying for leads. It's paying for buyers. Isn't that the point of marketing at the end of the day? Everyone wants a reliable way to get business. We pay marketers to bring customers to business, or to bring the business to the customer.

So our solution? It was to create an AI agent that has conversations with leads with the goal of handing off to a human real estate agent. Maybe the lead opted in from an ad or they signed up at an open house; we would use that context and whatever notes about them we could get to send them personalized texts. This way, realtors could focus on lead generation and on closing.

This worked well. We had dozens of real people get connected with realtors and lenders because of our product. It was great to see people get helped and not spammed. Some people opted out explicitly; others, we had simply nothing to offer. Even still, real deals were coming from this. We thought to ourselves, "what do we really have?"

A fully automated system to qualify and follow up with leads over long periods of time was in our hands. But as with any new solution, new problems arose:

  • What if our clients have slightly different campaigns? Not everything is an open house.
  • How can we ensure the leads are opted in and a fake number wasn't used?
  • What if a lead doesn't need to buy a house, but instead needs to do a refinance? Do we say "Sorry, can't help."?
  • Should we only support real estate use cases? Can we let our users make prompts?
  • What if the lead doesn't want to talk to anyone but our assistant?

These are individually tractable problems, and we addressed each of them as they arose. But our initial design wasn't scaling as we fixed these problems, so we began redesigning. Everything.

Were we going to just make conversational AI but for every industry? How would it look? Do we expect our users to be writing prompts? How can we scale our business?

Around this time, we started seeing the inevitable: CRMs and companies began making the same solutions. Some were good and some were terrible. CRMs are especially poised to succeed here since they're already focused on managing customer context. And for what it's worth, building this isn't really that hard. You can get to 80% of a conversational sales agent with this setup:

  • prompt: goals + lead context
  • tools:
    • schedule_followup
    • handoff_to_human_agent
    • set_do_not_contact

The rest of the 20% of implementation is bookkeeping (e.g., handling message delivery status) and prompting.

So where do we fit into this? What's the grand vision? To be honest, having a conversational outbound agent was something that I couldn't believe didn't exist. I started down that route knowing it'd be well-traveled soon.

Grand Vision

I won't say much for now but this chart is the pretty much the whole shebang. You problably can see what I see 🙃