It was supposed to be a yearly physical. Fifteen minutes, in and out.
I walked into a clinic in Montreal and sat down across from a doctor I'd never met. Before she asked about my sleep or my last blood test, she said this:
"I'll be using AI during this visit. Is that OK?"
That was the first thing out of her mouth. Not a form I signed on a clipboard. Not fine print on a portal. A direct question, from a person, with a pause afterward that meant she was actually waiting for my answer.
I said yes. The visit started.
What the setup looked like
Two screens on her desk.
- Left screen, my file. History, past results, notes. The usual.
- Right screen, live AI output. At the bottom left, a recorder showing it was listening to our conversation. On the right, search results pulling up as we spoke. When she said something, relevant information appeared. When I described a symptom, the AI surfaced what it thought was relevant. She scrolled through, picked what was useful, ignored what wasn't.
Chair on the left. Bed on the right. Two chairs next to the desk. A very normal doctor's office, with a very not-normal workflow running quietly in the corner.
The whole thing was unremarkable in the best way. The AI didn't feel like a demo. It felt like a stethoscope.
What she didn't do
She didn't let the AI decide anything.
Every suggestion that appeared on that right-hand screen was read, weighed, and either used or dismissed. When the AI surfaced something she disagreed with, she said so, out loud, to me. When it surfaced something useful, she told me what it was and why she was using it.
She was the doctor. The AI was a tool. I was in the loop the entire time.
That's it. That's the whole model. Consent at the door. Transparency throughout. A human deciding, every single step. Nothing about it was complicated. Nothing about it was expensive.
Now let me tell you what I see in the SAP world
I spent my career inside enterprise systems. I now build AI products for companies running SAP at scale, Fortune 500s sitting on decades of master data, transactional history, supply chain signals, customer records. The raw material for genuinely useful AI is already there. It has been there for years.
And most of them are still stuck.
Not stuck on the technology. The technology is mature. SAP's own AI stack, the hyperscalers, the open-source models, the integration patterns. All of it is available today, and most of it has been available for at least two years. You can run a responsible AI pilot on an SAP landscape in under a quarter. I've done it.
They're stuck on trust.
The same executives who will sign off on a 14-month S/4HANA migration won't approve an AI pilot that could pay for itself in ninety days. The same teams that accept "the system says so" as a reason to reroute a supply chain will reject "the AI suggests" as too risky to act on. The same companies that have been handing their most sensitive data to consultants for twenty years are suddenly worried about where the data goes.
None of those concerns are wrong. All of them are answerable. That's the whole point.
The gap is the lesson
Here's what struck me in that Montreal exam room.
My doctor is not a technologist. She's not a founder. She's not on any vendor's reference call. She just figured out (or someone in her clinic figured out) that if you want people to accept AI in a sensitive, high-stakes, personal context, you do three things:
- You ask.
- You show.
- You stay in charge.
That's the template. It works for a body and it works for a balance sheet. There is nothing stopping a supply chain director, a finance controller, or a procurement lead from running the same playbook on an SAP process tomorrow morning. Ask the stakeholders. Show them what the AI is doing. Keep a human in charge of every decision that matters.
The clinic isn't ahead because it has better technology. It's ahead because it made a choice about how to use the technology it already had.
What's actually stopping you
So when I watch enterprise after enterprise circle the runway on AI (pilots that never ship, committees that never decide, roadmaps that never leave the slide), I think about that doctor.
She didn't wait for a framework. She didn't commission a working group. She asked a patient a question and pressed record.
If a Montreal clinic can integrate AI responsibly into something as sensitive as your health, a mid-sized enterprise can do it with a forecast, a supplier contract, or a customer ticket. The tools are there. The data is there. The patterns are known.
What's stopping you, the tech or the trust?