AI is everywhere right now. Everyone's talking about how it's changing the world, replacing jobs, writing emails, and running businesses. But let's cut through the hype for a second.
If you're like most loan officers, you've probably wondered: can AI really handle client communication as well as a human? Can it follow up on leads, build trust, and actually get people to book appointments?
I wanted a real answer to that question, not marketing talk or theory. So I ran a live test.
The Test
To keep things fair, I wanted enough data to see real patterns, not random luck. Here's how it went down:
Batch 1: Generic "Out-of-the-Box" AI Platform
I sent emails to 3,049 people using a built-in AI email follow-up platform. The system asked me questions about my services, tried to understand my goals, and then automatically replied to every incoming email.
Results:
Out of 3,049 emails sent, I received 56 replies. When I read through each conversation, 11 replies were completely off-base. The AI misunderstood context, tone, and intent. It even replied to "out of office" autoresponders as if it were speaking to a real person. A few responses were acceptable, and I did get 2 legitimate leads, but overall, it felt more like babysitting a robot than running a campaign.
In short, it performed poorly enough that I would never trust this type of AI to handle real lead engagement again.
Batch 2: Custom-Built AI System (My Own Tech Stack)
Next, I sent another batch to 2,986 contacts using a custom AI system I built myself. This one was trained on my real conversations, tone, and intent. It had clear rules for how to handle replies, objections, and next steps.
To make things consistent, I sent this batch on the same day of the week, within 17 minutes of Batch 1's send time. That eliminated time-based bias completely.
Results:
I received 51 replies, and only 2 responses showed confusion. The rest were accurate, on-topic, and felt surprisingly natural. I ended up with 6 appointments booked, plus several quality conversations still in motion.
This one was good. Not perfect, but good enough to use again with confidence, especially because I know it can improve over time with proper training and feedback.
Batch 3: Human-Only Follow-Up
Finally, I manually replied to every email myself. I sent 3,002 emails, got 53 replies, and booked 8 appointments.
No surprise here: the personal touch still wins. But now I had something valuable: real numbers to compare.
The Results at a Glance
| Batch | Emails Sent | Replies Received | Quality of Replies | Appointments Booked | Grade |
|---|---|---|---|---|---|
| Batch 1: Generic AI | 3,049 | 56 | 11 off-topic, low accuracy | 2 | C |
| Batch 2: Custom AI | 2,986 | 51 | 2 off-topic, mostly accurate | 6 | B+ |
| Batch 3: Human | 3,002 | 53 | All accurate | 8 | A |
What This Test Actually Proved
Here's the takeaway: I'm not anti-AI. Far from it. AI absolutely has a place in modern mortgage marketing and is rapidly becoming a requirement if you wish to compete. But it's how and where you use it that determines whether it helps or hinders your progress. Unfortunately, a lot of what is being taught to loan officers falls in the "hinders" category.
Batch 1 showed exactly what happens when you bolt on a generic AI and hope it "just works." It doesn't. Hope is not a strategy. These types of platforms often react without understanding nuance, tone, or intent. That's not intelligence. That's automation pretending to be smart.
Batch 2, on the other hand, was a different story. It worked well because it was trained properly on my data, my tone, and my workflow before we let it loose. I spent extra time fine-tuning its tone, framing, and knowledge base. By the time we flipped the switch on this batch, I felt far more comfortable that this A.I. Assistant would represent me in a much more meaningful way, and the results proved it. It performed like a helpful assistant, not a clueless intern.
To take things one step further, I'd like to point out that if more batches were sent, and more hands-on time invested with this Assistant in analyzing the performance of each interaction, the outcome could have been improved even further, likely earning an "A" rating from me.
Batch 3 still won overall because humans understand emotion, context, and nuance better than any algorithm. But Batch 2 proved that with the right setup, AI can handle a lot of the heavy lifting and save serious time, allowing you to engage in larger-scale marketing and outreach that you'd otherwise never have the time for.
Not to mention the consistency of knowing this sort of marketing, lead follow-up, and nurture will remain steady no matter how busy you get, which is where A.I. truly makes a difference.
Where AI Actually Shines for Loan Officers
Used correctly, AI doesn't replace your personal touch or voice. It amplifies it, expanding your reach so that you can tap into the critical mass (velocity of message, otherwise known as "casting a bigger net") and numbers needed to move the needle and start getting reliable leads and conversions into loans.
Here's where I've seen A.I. do the most good for loan officers:
1. Speed-to-Lead (The First 5 Minutes)
When someone fills out a lead form online, timing is everything. AI can immediately send a warm, relevant message confirming receipt as well as interest, collect additional data for context, and then schedule an appointment or the next step.
This might seem like a small step, but in reality from my own notes and anecdotes, the average loan officer's response time to new leads is well over an hour. Somewhere between 1 hour and 1.5 hours, which is not good. The negative impact on lead contact is enormous once you're outside those first 5 minutes, so using A.I. to fill the gaps when you're too busy can truly make a difference.
My Recommendation:
If you're going to use A.I. for this purpose, set it on a timer that triggers at the 5-minute mark, checks your CRM to see if the lead has been contacted yet, and if the answer is no, it will spring into action.
I encourage all LOs to do their best to be the one following up with each lead when possible, but we all know that in the real world, you're often busy with clients, in a closing, or otherwise engaged. Let the A.I. step up when you cannot. This simple gap-fill can make a significant difference in your closing rates and cut down on lead-chasing.
Additional Note:
Arm the A.I. with multiple channels. Don't tie its hands by limiting it to email. I gave mine the ability to email, send a text (that then references the email), make a phone call, and leave a voicemail. If you're going to use voice though, make sure it's done correctly, or that's yet another area where more harm can be done than good.
2. Re-Activating Old Leads
Got a bunch of old leads that never converted just sitting somewhere?
Give A.I. access to the full spreadsheet, including all the notes from previous contacts and conversations. The more context you provide the A.I., the better your results will be. Just as with lead follow-up, I recommend letting the A.I. reach out through more than one channel, but make sure you have guardrails in place so it's not overdone.
Custom-tailored outreach and re-connect messages can easily resurrect leads that you thought were dead and gone. Besides, it's not like most LOs ever really get around to doing this with consistency anyway, so this gives you more shots at the basket than you'd otherwise get.
3. Consistent Nurture Campaigns
These can work extremely well if you map out a proper sequence and teach the A.I. how to follow up properly. If you're just going to have it send out boring, canned messages like the typical off-the-shelf LO CRM platforms, then don't bother. They won't convert any better.
But if you're going to map out a proper flow, teach the A.I. proper engagement, market shifts, credit updates, education, and trust-building strategies, then the A.I. will excel at this role. Each follow-up will be unique and personalized, yet still fall well within the framework and guidelines you've outlined.
Very powerful, but also easy to deploy the wrong way and just end up with an overpriced autoresponder sequence. Done correctly, this is yet another opportunity for a solid lift in conversions.
4. Realtor and Partner Outreach
Here's one with the potential to really punch above your weight. Personalized real estate agent outreach on a mass level can be game-changing. But I don't limit things there—let's get a bit more creative with it.
Why not have the A.I. reach out to the agent on the other end of each of your purchase transactions? Progress reports, post-closing follow-ups, etc.? Give each agent white-glove experiences like they've never enjoyed before, and earn their trust and referrals.
A week should never go by where you (or your A.I. clone) haven't spoken to at least 50–100 agents (minimum, personally I like to see much higher). If you want momentum, you have to create it. This is how you do it.
5. Post-Closing Follow-Ups and Review Requests
Here's some ultra-low-hanging fruit! Reviews, testimonials, referrals—we all know that we need them, but how many LOs are actually asking for them every single time as they should? Reviews have become an absolutely critical piece of your business, since you need them for your Google Maps (GBP) ranking among others.
But when will you take the time to go back through every single one of your closings in the past and ask? Not likely! So why not delegate the task?
Same with referrals. Most clients are happy to refer business, but most won't remember to do it, even if they tell you they will at the closing table when emotions and excitement are high. Reality? More often than not, it doesn't materialize.
But how many would happily send some referrals your way if they were reminded? Once again, become a master of delegation and let your A.I. take over these annoying, but profitable tasks and see what happens next. It pays off.
Final Thoughts
AI won't save your mortgage business if you haven't invested the time to address the core issues first. It's not a magic fix. It's a force multiplier. If you've accepted this fact and internalized what it means, you can wield A.I. to help you knock out all those little tasks that fall in the "little hinges swing big doors" category that you never seem to have time to complete with consistency.
If your process is solid with clean data, consistent follow-up, and a clear message, AI will help you do it faster, better, and to far more people, which is where the magic happens. But if your process is broken, and all you're doing is strapping an A.I. platform onto said broken processes, then AI will just help you fail faster and empty your wallet a bit quicker.
That's the truth most "AI gurus" won't tell you. Now you know.
