What Should You Automate With AI, and What Should Stay Human?

I think AI should handle a lot more of the computer work around the relationship… research, prep, organization, drafting, reviewing documents, the repetitive follow-up stuff. All of that. But the relationship itself? Conversations, trust, judgment, negotiation, empathy, showing up like a real person online and in person… I still want people doing that. I don’t think the goal is automating as much of your business as you can. The goal is automating the right things so you have more room for the work that still needs a person.

I train real estate agents, title professionals, and loan officers on AI, and I’ve delivered more than 500 live sessions up and down the East Coast. Almost every room eventually gets to some version of the same question. We know AI can do more and more of this stuff. The better question is whether we should let it.

Table of contents

• Why are we automating the wrong things?
• What does automating the wrong thing look like?
• What is AI genuinely good at?
• What should stay human?
• Where is the line? Inbound versus outbound
• What does this look like in a real transaction?
• Isn’t the point of AI to automate as much as possible?
• What should you automate first?
• FAQ

Why are we automating the wrong things?

Because the tools made it easy to automate the visible stuff first. The posts. The DMs. The comments. The outreach. Some of the most personal communication we have… and consumers are noticing.

We’re in a trust recession right now. A year ago, 82 percent of consumers said AI search was more helpful than traditional search. That number is down to 54 percent (Fractl and Search Engine Land, Q2 2026). Half of Gen Z have unfollowed, muted, or blocked a brand or creator because the content felt like AI slop (Sprout Social Q1 2026 Pulse Survey). And when people know a brand’s content was written by AI, 32 percent trust that brand less. Only 15 percent trust it more (Meltwater and YouGov, April 2026).

I call it the authenticity deficit. We’re all scrolling past the post we can tell AI wrote. We get the sense somebody’s just reading an AI script on their videos and we scroll right past it. We’re killing our own credibility thinking AI is making us more efficient, and the actual authentic moments to connect get missed.

What does automating the wrong thing look like?

Two things that happened to me as a consumer, not as a trainer.

A while back I reached out about a piece of land for sale. I was thinking about building a house at the time, so I requested a call with the listing agent. When he finally called me back, the whole thing was brief and impersonal. The land’s already under contract, anything else I can help you with, okay, have a good day. He had a guy on the phone who was interested in buying land and building a house… and he asked me nothing. Didn’t explore a thing. Then less than two minutes after we hung up, I get a text. Hey, this is so-and-so’s assistant, we’d love to add you to our distribution list for any of your upcoming needs. So technically the follow-up system worked perfectly… except the guy had just had a real buyer on the phone and never bothered to ask me a question.

Second one. I moved to a new area in Virginia, and a top agent I’d met before messaged me on social media. Welcome to the area, would be great to connect at some point. And it felt genuine. I said yeah man, let’s grab coffee, let me know when you’re around. Within about three minutes I get a response back. That’s awesome! Are you looking to buy or sell… It was his AI agent. And that one automated reply killed everything the first message had built, because I realized I wasn’t having the conversation I thought I was having.

And it’s not just those two moments. I train rooms of agents constantly, and plenty of them add me to their mailing lists afterward. The emails are tone deaf. If you’re looking to buy or sell, we’re here. Nothing personal in them, nothing that shows a human remembers who I am. Teams across the industry are automating proactive sales outreach, and to me that’s the last place AI should be running by itself.

What is AI genuinely good at?

I’ve used this analogy in trainings a lot lately. AI is the crew behind the scenes. Let it help with the prep, the research, the organization, the production… all the stuff that makes the show work. But you still have to show up for the show.

So what does the crew handle? In practice:

• Research, and digging through large amounts of information
• Organizing files, notes, and data
• Prepping you before meetings
• Reviewing documents and flagging what a person should look at
• Drafting, so you’re editing instead of staring at a blank page
• Summarizing long threads, transcripts, and reports
• Repetitive administrative computer work
• Turning conversations into follow-up tasks so nothing falls through the cracks

Look at that list for a second. All of it happens around the relationship, before or after the human moment, on a computer. None of it puts AI in front of your client.

What should stay human?

The relationship itself. Conversations, building trust, judgment calls, negotiation, empathy, reading a room. And showing up as yourself, which counts online just as much as it does in person.

That online part trips people up, so let me tell you where I draw the line. You may draw yours somewhere different. I write my own DMs. I reply to my own comments. When it’s somebody’s birthday, I’d rather send a quick voice memo in my own voice than let a tool post something canned on their wall. And I show up on video as myself. There are AI clone tools now that will generate a video version of you from a script, and I don’t want them anywhere near my outreach or my content. 88 percent of people say AI-generated video tools have eroded their trust in the news they see on social media (Sprout Social Q1 2026 Pulse Survey), and that skepticism does not stay neatly inside the news category. People want to experience the real you before they work with you, and I think we’re making a mistake when we hand that part over.

Here’s the split I teach, on one page:

Work	Automate with AI?	Why
Market research and prep before a meeting	Yes	Computer work around the relationship
Organizing files, notes, and transaction details	Yes	Tedious on a computer, exactly what AI is for
Drafting documents, posts, and email starting points	Yes, with your edit	You edit instead of starting from blank
Summarizing calls and building follow-up task lists	Yes	Nothing falls through the cracks
Answering inbound status questions on an active file	Carefully, yes	Fast accurate answers are the service
Proactive outreach to prospects and your sphere	No	This is the relationship itself
DMs, comments, and social conversations	No	People can tell, and it costs trust
Negotiation, advice, and judgment calls	No	This is what they hired a person for
Showing up on video	No clones	People need to see the real you

Where is the line? Inbound versus outbound

There’s a massive difference between inbound and outbound, and that distinction settles most of the gray areas for me.

Inbound… a client on an active transaction asks what time closing is, what documents they need, where things stand. I think there’s a real place for AI handling that, as long as it’s accurate and honest about what it is. In that moment the client wants a fast, correct answer more than they want a personal touch, and it frees your team up.

Outbound is different. That’s you initiating contact with a prospect, a past client, your sphere… trying to build or maintain a relationship. That’s where automation starts feeling weird to me, because the whole value of that message was that another human chose to reach out.

The rule I’d hang on the wall: if automating something gives you more time to be human somewhere else, it’s probably worth exploring. If automating it removes the human part your client values, slow down.

What does this look like in a real transaction?

Take one client meeting, whatever seat you’re in.

Before the meeting, let AI do the digging. The property, the market, the loan scenario, the title file… have it prep you with a summary and the questions likely to come up. Whether you’re an agent walking into a listing appointment, a closer getting ready for a settlement, or a loan officer ahead of an application call, that prep is the same kind of computer work, and you’ll walk in more prepared than you’ve ever been.

Then you get into the meeting, and this is where I want you present. Listen. Ask the next question. Read the room. Don’t spend the whole conversation thinking about what the system is going to do afterward.

When it’s over, AI earns its keep again. Your notes or your recording turn into organized follow-ups, a drafted recap, a task list so nothing gets dropped. The follow-up itself still needs your judgment and your voice though. AI can draft it… you decide what it says, and it should sound like you before it goes out.

Isn't the point of AI to automate as much as possible?

No, and I’d push back on that whole framing. I don’t think the goal is to automate as much as we possibly can. I think the goal is to get some of this computer work off our plate so we have more time for the stuff that still needs a person.

I tell rooms all the time… if AI gives you three hours back this week and you spend those three hours scrolling your phone, AI didn’t accomplish anything for you. If I give you three hours back, what are you doing with the three hours? That’s where the ROI is. More conversations, more appointments, more face time, more of the work only a person can do. The tools don’t get you better results by themselves. What you do with the time they free up does.

What should you automate first?

Try this. Look back at your last week and find the computer work that annoyed you. The status you chased three times, the document you reformatted again, the twenty minutes you spent digging through email for details you knew were in there somewhere, the same explanation you typed out for the fifth time. Pick one of those and figure out how AI or a better process takes it off your plate.

Then find the other thing in your week. The hard conversation that went well, the negotiation, the in-person moment you wish you had more time for. That’s what the recovered time is for. Start with one of each, and you can be running this by Friday with the AI you already have.

FAQ

Should I let AI write my social media posts? I’d let it draft them, not write them. Feed it your own words… transcripts, stories, the way you talk… and let it build a starting point, then edit until it sounds like you. Posts that are straight AI-written are exactly what people are learning to scroll past, and about a third of consumers say they trust a brand less once they know the content came from AI.

Is it okay to use an AI assistant to respond to leads? For inbound questions on active business, handled honestly, I think so. For outbound relationship building, I wouldn’t. The moment somebody realizes the personal message that reached out to them was automated, you’ve lost more trust than the automation saved you in time.

What about AI answering my phones or chat on my website? Same test. A visitor asking your office hours or a client asking for a status update is inbound service, and AI can handle that fine as long as it’s accurate and not pretending to be you. Once it moves into advice, negotiation, or trying to win new business, that belongs to a person.

Are AI video clones ever acceptable? For marketing and relationship content, I wouldn’t use one. Consumers already report that AI video has lowered their trust in what they see online, and your face is the last thing you want them second-guessing. Record the real you, imperfections included. In my experience it performs better anyway.

How do I know if I’ve automated too much? Ask yourself what your client would think if they saw exactly how that message or touchpoint got produced. If the answer makes you uncomfortable, you’ve probably crossed the line somewhere.

Doesn’t refusing to automate outreach put me behind competitors who do? I’d argue the opposite. As more outreach gets automated, the genuinely human message stands out more, not less. Everybody’s inbox is filling up with automated touches, so yours being real is the advantage.

Where does the time AI saves really go? Wherever you decide to put it, and that’s kind of the whole point. Decide up front that the hours go to conversations, appointments, and showing up. If you don’t decide, they’ll disappear into the scroll, and you won’t be able to point at anything AI did for you.