The AI Operator & Data Ladders: Where You Are and What’s Next

About the Episode

Why AI Hasn’t Clicked for You Yet — And What Actually Fixes It

If you’ve tried AI for your business and walked away thinking “that’s it?” — you’re not behind, and you’re not doing it wrong. You’re just missing one piece most people never hear about.

On this episode of Out of the Hourglass, Chris Kiefer returns for his third appearance, joined this time by his business partner Cody Hopkins. Chris has spent the last several years building Boolean, where he and his team work as what he calls “a fractional CTO” for twenty six painting businesses around the country, helping them integrate technology in a way that actually moves the business forward. Cody brings a different background to the table — ten years as a high school teacher before making the leap into operations and, eventually, AI systems building. That teaching instinct shows up constantly in how he explains things, and it makes this conversation an easy one to follow even if you’ve never touched an AI tool beyond your phone.

Why AI Feels Like a Smarter Google — And Why That’s Not the Ceiling

Most business owners are using AI the same way Cody used it eight months ago: as a smart conversational tool, nothing more. As Cody put it, he “had a working relationship with ChatGPT on my cell phone… It made me chuckle, answered my questions, really validated my feelings. And that was it.”

Chris explained that this isn’t a knock on business owners — it’s just where the easy wins are. “It’s because that’s where you can see an ROI,” Chris said. “Like I could save myself five minutes by writing this email.” The problem is that incremental gains like faster emails are nowhere near the ceiling of what’s possible — they’re just the most visible place to start.

The real unlock, according to Chris, isn’t asking AI better questions. It’s stepping back and asking a different kind of question altogether: “Why do we send emails to begin with? … What would be a way that I could insert you into that problem to solve that problem better for me?”

The Two Ladders: Operator and Data

Cody has spent months building out a framework to help people diagnose exactly where they stand with AI — not as a single scale, but as two separate ladders that work together.

The first is the Operator Ladder — how you personally use AI, regardless of which tool you’re using. It runs from someone who simply chats back and forth with an AI tool, up through people who consolidate files and information for it to work from, up to those who orchestrate multiple AI agents working autonomously.

The second is the Data Ladder — how your information is structured and accessed. Cody walked through this using a meal-planning example with Chris’s own household: a complex set of dietary needs, allergies, and a training schedule for Chris’s wife. At the most basic level, you’re just copying and pasting information into a chat window every time. Higher up, your tools are connected — AI can go read your calendar or your files directly. Higher still, everything routes into a centralized database designed specifically so AI agents can navigate it. At the top, that data refreshes itself automatically, staying current without anyone manually updating it.

Here’s the key insight Cody offered: these two ladders move independently. “You could really level up as an operator, but have poor data and then not be as effective with AI as you could be,” he said. “On the other hand, you could have really, really excellent data and a very basic understanding of how AI works and get really similar results.”

Clean Data Is the Real Bottleneck

If there’s one phrase that anchors this entire conversation, it’s “clean data.” Chris broke down exactly what that means in practical terms — and it’s more specific than most people assume.

First, your data needs to be structured. That means your project types, customer fields, and other key data points exist consistently across every tool you use — your CRM, your project management software, your QuickBooks — using the same names and formats, not just dumped into a spreadsheet cell as a wall of text.

Second, it needs to be accurate and current. Chris used a simple example: if a customer’s email address is wrong in your CRM and you correct it on-site, does that correction make it back into the system? “It might be correct in one system and not the other,” he said. “Do you have processes built in place to keep it clean so that you’re still getting value out of that with all the tools that you’re putting onto it?”

Without that foundation, Chris explained, AI can still technically work — it’ll process a messy spreadsheet and give you an answer, the same way a human would muddle through it. But it’s working despite your data, not because of it.

A Reality Check on Cost

One of the more surprising parts of the conversation was about money — specifically, how much AI actually costs to run versus what users are currently paying.

Chris explained that we’re currently in what’s being called the “subsidy era” of AI, with roughly a trillion dollars invested into AI infrastructure over the last several years. The tools available to consumers right now are priced well below what they actually cost to run. As Chris put it, the comparison he’s heard is that it’s “the equivalent of the government just giving everyone in America brand new iPhones and brand new AirPods… and they’re saying twenty bucks a month.”

The numbers back it up: as of April 2026, a $200 a month Claude plan is roughly the same as $5,000 of compute if paid through Claude’s API directly. The takeaway isn’t to panic — it’s to plan ahead. As that gap closes, costs are likely to rise, which means it matters how you build.

Build With AI — Don’t Depend on It to Run

This is the practical advice Chris and Cody kept coming back to: use AI to build tools, but don’t make your tools dependent on AI to keep functioning.

Chris shared a real example — a scheduling tool his team vibe-coded in about three days for a painting business owner in Atlanta, pulling in real-time data from Google Maps and Airtable to help optimize drive time between job sites. The tool itself doesn’t run on a language model every time it’s used — it runs on the business’s existing data through simple API calls. “We used AI to make the app,” Chris explained. “But for [the owner] to continue using this when AI prices go up, this is going to cost next to nothing to run.”

The contrast Chris drew was sharp: a lot of what gets shown off at AI conferences is impressive-looking but disconnected from real data — what he called the equivalent of “a piece of paper” with “a house floating on a cloud… it’s got a water slide.” It looks great in a demo. It doesn’t hold up in a real business unless it’s actually wired into your systems.

Where Do You Stand — And What’s Next?

Toward the end of the conversation, Cody and Chris pointed listeners toward a free scorecard built around the Operator and Data ladders discussed throughout the episode. The idea is simple: answer a short series of questions, find out where you currently sit on each ladder, and get a clear, actionable next step — rather than trying to leap straight to the most advanced use case you’ve heard about.

As Chris put it, “It’s not useful to just jump in and start giving feedback until… if you’re gonna do math class, right? Take an assessment to know what level of math you understand, and then we’ll start from there.”

You can find the scorecard at paintos.app/ai.

Wherever you currently sit — whether you’re still chatting with AI one question at a time or you’ve already started connecting tools together — there’s a clear next move. The goal isn’t to become an AI expert overnight. It’s to take the next honest step from exactly where you are.