AI-First Is Not a Product Decision. It's a Process Decision.

The Question Nobody Is Asking

Every executive I’ve worked with in the last two years has eventually asked some version of the same question: Which AI should we use? Almost none of them have asked the better one: Which of our processes is so broken that AI is the only lever left?

That distinction is the difference between adding a layer of automation on top of a dysfunctional system—and actually transforming how a business operates.

This week, two things I’m building converged in a way that made this clearer to me than ever. One is Pheraxia, the AI-first company I co-founded to redesign how bioactive molecules get delivered to cells. The other is a full-day AI sprint I ran with thirty leaders from a Mittelstand group—E.Gruppe—where we skipped the strategy decks and went straight to building. Different industries. Different stakes. Same core insight.

AI-first is not a product decision. It’s a process decision.

The Delivery Problem Is Everywhere

Let me start with Pheraxia, because it illustrates the principle at its sharpest.

Most premium supplements fail a basic test: the bioactive never reaches the cells it was designed to act on. Not because the molecule is wrong. Not because the science is weak. But because the delivery architecture is broken. For decades, the industry’s answer has been blunt-force: increase the dosage. More of the same thing, hoping more gets through. Expensive. Slow. Unscalable. And it treats a process failure as if it were a formulation problem.

We’re not inventing new molecules at Pheraxia. We’re redesigning the carrier architecture—the method by which a molecule travels to the cell—using AI simulation instead of trial-and-error lab work. The result: thousands of potential carrier configurations tested digitally before a single lab run. Fewer experiments. Less capital consumed. Faster proof.

Here’s what I keep coming back to: this isn’t just a biotech story. It’s a business story. Because almost every company I’ve worked with has a version of this same broken delivery problem—they have good raw material (talent, data, product) and a method that wastes most of it before it reaches the customer.

What Happened in One Day with Thirty Mittelstand Leaders

A few weeks ago, I ran a full-day AI sprint with a group of thirty operators—managers, team leads, functional heads—from a mid-sized German group. No consultants. No keynotes. No waiting for alignment.

By noon, a reporting process that had consumed fifteen hours a week every week had been automated. Before the first coffee break, an inventory decision that had been made on gut feel—and had cost real money—was replaced by a data-driven system. By end of day, one manager who had never written a line of code had deployed a workflow that saves his team two full days per month.

I’ve been in rooms like this before. The energy was different. It wasn’t the excitement of a demo—it was the realization that they could do this. Not someday. Not after the next budget cycle. That day.

The challenge we set before everyone left: one AI win per week for the next sixty days. Measured in euros saved or hours reclaimed. No buzzwords, no pilots. Outcomes.

That’s the Mittelstand spirit at its best. Pragmatic, direct, unwilling to let complexity become an excuse.

The Framework: Manual → Standard → Automate

After two years of doing this work—inside other people’s companies and now in my own—here’s the model I keep returning to. I call it MSA: Manual → Standard → Automate.

Manual first. Before you automate anything, run the process by hand long enough to understand where the actual waste is. Most companies skip this step and automate chaos.

Standardize second. Document the process until a new hire could follow it without asking questions. If you can’t describe it clearly, you can’t improve it reliably.

Automate third. Now introduce AI or tooling. Not to replace thinking, but to remove friction from a process that already works.

This sounds slow. It isn’t. The E.Gruppe team built working automations in a single day because the leaders in that room knew their processes. They’d run them manually. They understood the failure points. The AI just gave them the means to act on that knowledge faster.

At Pheraxia, we’re applying the same logic: we don’t start in the lab. We start in simulation—understanding the delivery problem digitally before we commit to physical experiments. Manual insight first. Then structured testing. Then scaled method.

The Human Moment Nobody Talks About

There’s something I’ve noticed that rarely gets said out loud in transformation conversations.

I sat with a senior leader once—mid-fifties, deep operational experience, the kind of person who had held his function together through multiple ownership changes—after a workshop on AI process design. The room had cleared. He asked me quietly whether what we’d been discussing would make him obsolete.

I told him the honest truth: the people who get replaced by AI are rarely the ones who understand the work deeply. They’re the ones who held on to a method long after the method stopped serving the work. He nodded. He was the one who had flagged the broken reporting process before anyone else in the room.

That’s the EQ side of this that gets buried under the tooling conversation. AI doesn’t replace judgment. It exposes which processes were never really about judgment to begin with—and frees the people who have it to use it on things that matter.

What This Means for Your Business Next Week

Here are five things you can do—no budget required, no consultants needed:

Audit your highest-friction process. Pick the one thing your team complains about most consistently. Don’t fix it yet. Just map every step, in writing.
Ask the delivery question. Whatever your product or service is: where does it lose value between creation and the customer? That gap is your process problem.
Run one manual experiment. Before automating anything, do the next iteration by hand. Learn what breaks before you lock in a workflow.
Set a sixty-day outcome target. Not a pilot. Not an exploration. One specific result—measured in time saved or cost reduced—that you will achieve using an AI-assisted process change.
Have the quiet conversation. Ask your most experienced team members where they feel underused. That’s where your process waste usually lives.

The Connecting Question

AI-first transformation isn’t about which tool you choose. It’s about having the courage to look at a broken process and decide that this time, you’re going to fix the method, not just increase the dose.

The companies that will win the next decade in the Mittelstand aren’t the ones with the biggest AI budgets. They’re the ones with the clearest understanding of where their delivery is broken—and the discipline to rebuild it properly.

Where in your business is the method broken—and are you fixing it, or just turning up the dosage?