How I Used AI to Clean Up a 103,000-Contact CRM

5 min read

When I first started hearing about AI, I was one of the people who thought I could pay someone else to handle it. I had a business to run, a team to support, and plenty of other things competing for my attention. Learning another technology felt like one more job to add to the list.

That changed when I started asking what AI could do to make my day-to-day work easier. I took a class, began experimenting with different tools, and realized I didn’t need to understand everything about AI before I could put it to work.

I just needed to know what I wanted to accomplish and keep asking questions until I found a useful answer.

One of my first major projects was cleaning up our CRM. We had 103,000 contacts in Follow Up Boss, and the database needed serious attention. I knew there were duplicates and outdated records, but going through that many contacts manually would have been a huge undertaking. AI gave me a way to start tackling the problem.

Start with a problem you already understand

I’ve been in real estate for about 25 years. I’ve run large teams, worked with different franchises, and helped build the systems behind a business with multiple offices and roughly 40 agents. I understand how much work happens behind the scenes to keep a team running.

That experience also means I know where small problems can turn into bigger ones. A CRM is only useful if the information inside it is organized well enough for agents to act on it.

We had situations where the same record appeared in our database hundreds of times. One showed up 333 times. That created confusion in our lead flow because the system had trouble distinguishing between people and records. We also had duplicate contacts and records that contained little more than a name and an invalid phone number.

I knew what needed attention. The challenge was finding a practical way to work through it.

My advice to agents who are new to AI is to start with a task you already know needs to get done. Maybe your database has duplicates. Maybe you have a pile of mailing labels to organize or templates that need reviewing.

Those are good places to begin because you understand the desired result, and you can check whether the work is correct.

Talk to AI like a person

One of the biggest things I’ve learned is to talk to AI the way I would talk to another person. Explain what you’re trying to do, give it context, and ask follow-up questions.

You also need to learn which tool makes sense for the job. I’ve been using Perplexity as a starting point for this work, and I’ve experimented with other platforms, too.

Each one has its own strengths, so it helps to develop a basic understanding of what you’re asking it to do.

I didn’t sit down with a perfect plan for cleaning up the CRM. I started asking questions and working through the answers. When I found something useful, I followed that thread. Then I asked what else we could improve.

That approach helped me identify more than 4,000 duplicate records. We also removed roughly 6,000 contacts that were otherwise unusable. The initial cleanup took about a week, compared to each individual task that would have taken me days to complete manually.

Those numbers made the value clear to me. I could spend that time working on the business instead of sorting through records one at a time.

Keep looking for the next improvement

The cleanup was a starting point. I’m now looking at how we can improve the way our systems work together, including Luxury Presence and Follow Up Boss. I’ve also been reviewing templates and looking more closely at where our contacts come from.

I keep asking AI what else we could do to improve the database. It will suggest a few ideas, and I decide which ones are worth exploring. Then we work through them and see what makes sense for our business.

That process requires judgment. AI can help identify problems and handle repetitive work, but I still need to understand what it recommends and check that the results are right.

I wouldn’t make major changes to a database simply because a tool suggested them.

The same principle applies to agents who want to start using AI. You don’t have to master every platform or find an application for every part of your business.

Pick one task that takes too much time, learn enough about the tool to use it responsibly, and see what happens.

I used to think I could avoid learning AI by hiring people to handle the work for me. Now I see that understanding the tools gives me more ways to improve the business. I can still bring in help when it makes sense, but I can also spot opportunities myself and ask better questions about what we should do next.

Before you add another task to your to-do list, look at the ones you keep putting off. One of them may be a good place to start with AI.