How to Build a Sphere of Influence System Using Claude AI

The Problem With How Most Agents Think About SOI

Ask an agent what their sphere of influence strategy is and they will describe an intention, not a system.

"I try to stay in touch." "I send a newsletter when I remember." "I check in with past clients around the holidays." These are habits they aspire to, not structures they run. There is a meaningful difference between those two things.

An intention depends on you. It requires that you remember. That you have time. That you feel like doing it today. That nothing more pressing is competing for the same hour. In real estate, something more pressing is almost always competing for the same hour.

A system does not depend on you. It runs on a schedule. It has defined inputs and outputs. It produces the same result whether you feel motivated or not, whether your pipeline is full or empty, whether you remembered it was Tuesday or not.

Most agents have neither. They have a CRM with hundreds of contacts, a general awareness that they should be in touch with those people, and an ongoing low-level guilt about the gap between the two. The guilt is not a system. The CRM is not a system. The awareness is not a system.

The agents who consistently generate referrals from their database are not more disciplined than the agents who don't. They have built something that runs whether discipline shows up or not. Claude AI is now one of the most accessible tools available for building that thing.

What a Claude-Based SOI System Actually Is

When I say "build a sphere of influence system using Claude AI," I do not mean: open Claude, type "write me a text to send my past clients," copy the output, close the tab.

That is using Claude the way most agents use it. It is marginally better than doing nothing. It does not compound. Next week you are starting from zero again.

A system built around Claude has three components that a one-off prompt does not.

Persistent context. Claude knows your contacts, your voice, your relationship history, and your communication standards. Not because you re-explain it every session, but because you have built a project with that information stored and available. When you sit down on Monday morning, Claude already knows who is in your database, how you've categorized them, and what kind of outreach is appropriate for each group.

A repeatable workflow. Each week runs the same way. You feed in the same kind of data. You get back the same kind of output. The output gets better over time as you refine the instructions, not because you have to think harder about it each week.

A schedule that is not optional. The system runs on Monday. Or Sunday night. Whenever you have decided. It does not run when you feel like it. It runs at the time you chose when you built it, and the choice does not need to be made again.

Three components. None of them require you to be a developer or a tech person. They require you to decide what the inputs are, what the outputs should look like, and when the whole thing runs. Claude handles the work in between.

Building the System: The Three Parts

Here is how to build this in practice. Start with the smallest version that actually runs. You can refine from there.

Part one: the contact database. Export your contacts from wherever they live — your CRM, your phone, a spreadsheet — into a Google Sheet or a document that has structure. For each contact, you want at minimum: name, relationship type (past client, referral source, personal connection, neighbor, etc.), last contact date, and any relevant personal notes. Kids' names. Job change. Moved to a new neighborhood. These details are what separate outreach that feels personal from outreach that feels automated.

You do not need every contact perfect before you start. Start with your top fifty. The people who would actually refer you if they thought of you. Build from there.

Part two: the Claude project. Inside Claude, create a project. Give it a name. In the project instructions, load your contact database and write out your outreach standards in detail. How you talk to past clients versus neighbors versus referral partners. What you will never say. What your voice sounds like. What a good check-in text sounds like versus a bad one. The more specific you are here, the better the outputs.

This is the work most agents skip. They want to prompt their way to good outputs without doing the setup. The setup is the system. The prompts are just the interface.

Part three: the weekly session. Every week, at the scheduled time, you open the project. You review which contacts are due for outreach based on your rotation plan. You ask Claude to draft personalized messages for each one. You review, adjust for anything Claude could not know (a recent conversation, a piece of news), and send.

That is the system. Contact database. Project with persistent context. Weekly session at a fixed time. Thirty minutes of consistent execution, every week, regardless of how busy you are or how much you feel like doing it.

What Weekly Outreach Looks Like When the System Runs It

Monday morning. You open the project. You have twelve contacts due for outreach this week based on your rotation.

Four are past clients. You ask Claude to draft a check-in text for each one. Claude pulls their notes: one just had a baby, one was asking about refinancing last spring, one recently changed jobs, one has lived in the same house for nine years and has never referred you but you still stay in touch. The drafts are different. They reference the specific things that are specific to each person.

Three are referral partners. A lender. A financial planner. An estate attorney. These get different messages. Relationship-focused. Value-adding when possible. Not just checking in for the sake of it.

Five are personal connections. Neighbors. Former colleagues. People from your church. These are usually the shortest messages. A line or two. Something personal first, a professional reminder second, or sometimes no professional reminder at all — just the personal connection that keeps you in their mind.

You review all twelve. You make small edits. You remove the line that sounds like something you would never say. You add the thing Claude couldn't know because you heard it in a conversation last Tuesday. You send.

Thirty minutes. Twelve people touched. Next week, twelve more.

In twelve weeks you have personally reached out to 144 people in your database. You have done it consistently, with messages that feel specific and real, without spending hours agonizing over what to say. The database is no longer a list of names producing guilt. It is a pipeline producing referrals on a schedule.

The Contacts That Come Back First

One thing that surprises agents when they start running a consistent SOI system: the first referrals do not come from the people they expected.

They assumed the past clients would be the source. The relationships they had done the most work with. Sometimes that is true. But often the first referral comes from someone in the middle of the list. A neighbor they had not talked to in eighteen months. A former colleague who had no idea they were still in real estate. A referral partner who assumed they had gone quiet because business was slow.

The common thread is not relationship depth. It is recency. The agent who showed up last is the agent who gets the referral. That is uncomfortable if you believe business comes from the relationships you have done the most work with. It is very useful if you understand that the work is showing up consistently, not banking on past connection.

A sphere of influence system does not guarantee the right people will refer you. It guarantees that when someone in your database is asked if they know a real estate agent, your name is the most recent one in their memory. That is the only thing you can control. The system controls it, automatically, every week, whether you thought about it or not.

The Part That Compounds Over Time

At week one, the system produces one week of outreach. Twelve contacts reached.

At month six, you have reached every person in a fifty-contact database roughly six times. You know who responds. You know who refers. You know which outreach format works for which group. You have refined the Claude project based on what you have learned. The outputs are better than they were in month one because you have made them better each time you ran the session.

At month twelve, the system has produced referral conversations that you can trace back to specific outreach. You have added new contacts to the rotation from closings, open houses, and new relationships. The database is growing and staying warm at the same time.

None of that happens with one-off prompting. It requires a system that runs on a schedule and improves each time it runs. That is the difference between using Claude and building with Claude. One session produces one week of outreach. A system produces a year of compounding referral activity from a database that most agents have left sitting quiet.

The agents ahead in five years will not be the ones who used more AI tools. They will be the ones who built structures around those tools that ran whether the agent was paying attention or not. The SOI system is the clearest place to start. The database is already there. The contacts already exist. The only thing missing is the structure that makes it move.