The Illusion of Data-First AI: Why Process is the Real Power

Justin Lerma

16 Jun 2025 — 3 min read

Bad processes, but now with AI 😂

We’re past the AI hype cycle. Most teams know they need to “use AI,” but I have a hot take after working with a few groups now. Some may be starting in the wrong place: with their data.

That’s like starting a fire with wet wood. Technically possible. Painfully inefficient.

Here’s the truth: data isn’t the beginning of an AI journey—it’s the byproduct of a well-understood process. If you don’t understand how your organization actually works, no amount of dashboards or models will make you smarter. You’ll just automate confusion faster.

We’ve Confused the Hammer for the Blueprint

There’s a saying: when all you have is a hammer, everything looks like a nail. In AI? The hammer is data.

People believe that if they have data, they can do AI. But that’s like saying if you have bricks, you can build a cathedral. Maybe. But only if you have a blueprint, a skilled team, and an understanding of architecture.

That blueprint? It’s your process.

AI thrives when it has something to model—a process with repeatable steps, clear outcomes, and enough variation to warrant optimization. Most companies skip this. They don’t map how work actually gets done. They don’t clarify where judgment lives or where improvisation thrives. They just feed the machine and hope for miracles.

This Might Be a Hot Take

But after dozens of conversations with customers and experts in the field, I’ve come to a blunt conclusion: if you haven’t invested in process clarity, AI will mostly create rework.

It’s not that your data is bad. It’s that your systems reflect a reality no one fully understands. This includes teams who have job security through obscurity, ill defined swim lanes and unclear job expectations.

Before you train a model, train your team to see the work. That means building out your playbooks—not as documentation theater, but as a survival tactic. Map what’s happening. Document the edge cases. Surface the inconsistencies.

This is the foundational layer where AI starts to make sense—and where human creativity starts to get space again.

Process Mapping Is Human Work—In Service of Better Human Work

When you map a process well, you don’t just prepare it for AI. You liberate your people.

You remove the non-differentiated heavy lifting—the approval pings, the redundant steps, the “just-in-case” spreadsheets. That’s where AI shines: not in replacing people, but in absorbing the work that never needed to be done by them in the first place.

What you get back is time. Attention. Cognitive slack.

And that space gets reinvested into the work only humans can do:

Designing better customer experiences
Investigating root causes instead of fighting fires
Prototyping new internal tools that people actually want to use
Thinking deeply about how to grow the business—not just keep it afloat

Even the Best AI Needs a Playbook

Let’s talk about the bleeding edge: autonomous AI agents. They’re exciting. They chain tasks, take action, and adapt in real time.

But they’re directionless without process. An agent without a clear playbook is just a toddler with access to your systems.

Want your agents to work? Build the structure first. Define roles. Assign rules. Create clarity.

AI may be autonomous—but it still needs purpose. That purpose comes from humans. From leaders. From teams that take the time to think through what excellence really looks like.

This Is How You Scale Trust

You don’t earn trust in AI by promising miracles. You earn it by showing people: “We’ve made the process clearer. We’ve given you guardrails. We’ve taken the boring work off your plate so you can finally get to the things you care about.”

You earn trust when AI isn’t just another tool—but a lever that lifts the human side of the business higher.

From Data Exhaust to Process Intelligence

Here’s the kicker: when you do this well, your data gets better.

Because now it’s not just capturing activity—it’s capturing intent. You know why something happened. You can test how a change affected it. You can simulate outcomes.

Process-first AI isn’t anti-data. It’s pro-meaning.

It’s the difference between having GPS coordinates and knowing how to navigate the terrain. One gives you a location. The other gives you the confidence to move.