The highest-ROI use of AI in your business right now is not content creation and it is not marketing. It is analyzing your existing customers to find out who is about to leave and who is ready to buy more. If you missed Part 1, which covers prompting strategy and tool selection, you can read it here. This part focuses on where AI actually moves money.
Start Here How to Get More Out of AI in Your Business - Part 1 →Most Founders Are Using AI for the Lowest-Leverage Work First
The number one use case of AI in business today is content and marketing. That is where most founders start because it is visible, it feels productive, and the output is easy to count in terms of volume. More posts, more emails, more ads. It feels like momentum.
It is also one of the lower-leverage places to focus first.
If your business has customers, the highest-value work AI can do for you is to tell you which of those customers is at risk of leaving and which ones are ready to buy something else from you. Revenue follows the customer who stays. A customer who buys again costs a fraction of what it costs to acquire a new one, and their value compounds over time because they already trust you and they already know how to buy from you.
The question most founders never ask is: what does my customer data actually show? Not what you assume based on gut feel, but what the actual behavior patterns reveal. That is a question AI can answer in minutes, using data you already have and are not doing anything with.
How to Use AI to Analyze Your Existing Customer Base
The first move is to connect your customer data to an AI tool and ask it to do the analysis. You do not need a data science team for this. You need your customer list, whatever you know about their behavior: purchase frequency, logins, support tickets, email opens, and a clear question.
"Here is my customer list and their purchase history from the last eighteen months. Tell me which customers are showing signals that they are likely to stop buying from us. Then tell me which customers look like they are ready to buy more."
Let it work. Come back to the output.
What you get is not a generic report. It is a specific list, ranked by risk and by opportunity, with the reasoning behind each one. Your job at that point is to act on it. Reach out to the at-risk customers before they leave and make an offer to the ones who are ready to buy more. Your core competency is the judgment call on what to do with the information. The analysis is what AI handles.
This is not a complicated workflow. I have seen it set up and producing results within a single session. The complexity lived in the analysis, the part nobody had bandwidth to do manually, and AI handles that the same way it handles anything else you give it a clear outcome for.
Revenue follows the customer who stays. Before you use AI to add more customers at the top of the funnel, use it to seal the bottom. Adding customers to a leaking funnel is expensive and slow. Fixing the leak first makes every new customer worth dramatically more.
The Three Levels of AI Adoption: Why Most Businesses Are Stuck on Level One
There are three levels of how businesses use AI, and most founders are stuck on the first one.
Level one is chat. You ask it something, it answers, you copy the answer and go do something with it manually. This is better than not using AI at all, but it leaves most of the leverage on the table because the human is still doing all the connecting work.
Level two is automation. You connect AI to the tools and systems you already use: your customer list, your email platform, your accounting software, and it processes information and takes actions without you having to copy and paste between tools. This is where a meaningful amount of time gets recovered and where most growing companies should be operating.
Level three is agents. You define a complete outcome and walk away. The AI handles the entire workflow, including getting input from your team, incorporating feedback, and producing a final deliverable, without you managing any step of it.
Most business owners who say they use AI are at level one. The ones who say AI has fundamentally changed their business are at levels two and three. The path from one to two starts with a single question: what is the most time-consuming thing in my business right now that involves moving information from one place to another? That is your first automation candidate.
Find the Bottleneck Before You Find the Tool
One of the most expensive mistakes in AI adoption is choosing the tool before identifying the problem. A founder sees a new AI capability and immediately wonders how to apply it everywhere. The better sequence is the reverse.
What is the actual bottleneck in your business right now? Not what sounds interesting in a newsletter, but what specific constraint is limiting your output, your revenue, or your customer retention? Identify that first. Then ask: is AI the right solution, or is the right solution a process change, a hire, or something else entirely?
Think about this the way you would think about a production line. If you have two chassis, one engine, and sixteen wheels, you can build exactly one car. Adding more wheels solves nothing. The bottleneck is the engine, and every hour you spend anywhere else is wasted.
The same logic applies to your business. If your biggest problem is that leads are not converting, no amount of AI-generated content is going to solve that. You need to fix the conversion problem first. That might involve AI, or it might involve understanding why people are not buying and addressing that directly. The point is to identify the constraint before you reach for a tool.
Complexity kills momentum in every business I have worked with. The founders who try to implement AI across ten areas at once rarely make meaningful progress in any of them. The ones who pick one problem, solve it completely, and build from there see results fast, and that momentum carries them into the next problem with confidence and a working model.
What AI Agents Actually Do and When They Make Sense
An AI agent is fundamentally different from a chat tool. When you use Claude to draft an email, that is a single task with a single output. When you use an agentic setup, you are giving it a complete workflow and walking away while it handles the entire sequence.
A real example: you want to analyze your customer base for retention risk, draft personalized re-engagement messages for the customers who are at risk, send those messages, track responses, and report back on what happened. That is five steps. In most businesses, that sequence either never gets done because no one has the bandwidth, or it takes a team member most of a week to work through manually.
With an agentic setup, that entire workflow runs on its own. You define the outcome and the parameters once. The agent executes the sequence, sends the messages, collects the responses, and delivers a summary of what happened and what needs your attention next.
The barrier to getting here is not the technology. The tools exist and they work. The barrier is that most founders have not thought through their workflows at a level of clarity that lets them hand them off cleanly. The discipline of defining the outcome precisely. That discipline is what makes the agentic tool work and it is the same discipline that makes everything else in your business work better, so the investment compounds.
Using AI to Get Energy-Draining Work Off Your Plate
Beyond revenue and retention, there is a category of AI use that most founders undervalue: getting the parts of your business that drain your energy off your plate for good.
Your inbox is the obvious example. Most founders spend more time in their inbox than it deserves. It is reactive. It cycles you through highs and lows based on whatever arrived most recently. And the majority of what is in there does not require you specifically to handle it. It requires someone who knows your context to handle it, and that is exactly what AI is equipped to do.
Connect your email to Claude. Give it a clear instruction: process my unread emails, rank them by priority, draft a response for each one, and give me the output in a spreadsheet. The drafts will be accurate because the AI reads the full thread, understands who the sender is in the context of your business, and writes a reply based on everything it knows about your situation. Your job is five minutes of reviewing and approving, not two hours of composing and responding.
This same principle applies to any task that follows a pattern. Status reports. Research summaries. Scheduling coordination. Anything that involves taking information from one place and doing something predictable with it is a candidate for AI to own. The question to ask is not "can AI do this?" The question is "is this something only I can do?" If the answer is no, it belongs somewhere else.
Freeing your attention from the predictable work is how you redirect it toward the things that actually require your judgment, and that is where the real growth in any business comes from.
Related Insights The Business Mistake That Stops Growth Cold →