How to Train AI on Your Business Processes (And Why Systems Come First)

What if the processes your team already follows could become the programming for AI?

Chris Hunter, founder of RoofingSites.com, has been running a marketing agency for roofing companies since 2003. He’s also a longtime SYSTEMology supporter. When AI arrived, he didn’t start from scratch. He already had documented SOPs, checklists, and trained staff for his core processes. Those systems became the blueprint for 18 custom GPTs that now run across the business, handling everything from content creation to sales coaching.

Here’s how he learned to train AI on his business processes, and how you can do the same.

Start With What’s Driving You Most Nuts

Chris doesn’t start with tools. He starts with pain.

His agency was on track to need 50 writers and editors to hit their long-term growth targets. Managing that many people would have been a nightmare. So he asked one question: what’s driving us most nuts?

The answer was content. Blog articles were the highest-volume, most time-consuming process in the business. Chris runs his agency on EOS, and in weekly leadership meetings, the team kept surfacing the same bottleneck: content production couldn’t keep up with demand.

Because his team had been running documented SOPs for that workflow for nearly 20 years, it was also the easiest process to hand to AI.

That’s the key insight. You can’t train AI on something that lives in someone’s head. If your processes are already documented, AI has something to work with. If they’re not, that’s your first step.

Build Custom GPTs for Specific Tasks

Chris started with one custom GPT. He trained it on 15,000 articles his writer had produced over 15 years. The results were immediate.

Content costs dropped from $50 per article to roughly 15 cents. Editing time went from one to two hours per article down to about ten minutes. His writer didn’t lose her job. She was promoted to content manager.

Then he kept going. He now has 18 custom GPTs, each built to solve a specific problem. Press releases. Social media posts. A personal assistant trained on his book and his team’s Colby scores that helps him decide which tasks to delegate and to whom.

He even built one trained on a sales methodology that critiques his discovery calls. He feeds in the transcript, and the AI tells him what he should have asked, how he could have presented differently, and what his follow-ups should be. He then uses those insights to coach his sales assistant and build SOPs for the next person who fills that role.

Each GPT started the same way: identify a specific pain point, review the existing process, and build a custom AI tool that handles the heavy lifting while a human does the final polish. Chris started manually with ChatGPT’s custom GPT builder, then moved to the OpenAI API and had his developers integrate it directly into the agency’s own project management software. It’s now like having a purpose-built content engine inside the tool his team already uses every day.

How to Bring Your Team Along

He explains it to his roofing clients as the difference between a regular 16-ounce hammer and an air hammer. Same job, completely different speed. For his own team, he uses a different analogy: an exoskeleton. AI doesn’t replace you. It makes you capable of things you couldn’t do before.

The team bought in because they were part of the process from the start. In weekly meetings, Chris listened for new pain points and asked what else they could build. Every new GPT solved a problem the team had already raised. Adoption wasn’t forced. It grew because each win was visible and immediate. And the scalability shifted overnight. Two people can now handle the content workload that would have previously required a team of 50.

The team members who lean into AI and systems become more valuable, not less. And the businesses that give their people these tools early will outperform those that wait.