A few weeks ago I was at my kitchen table with Claude Code, working on a side project called Patina — an architecture for bringing AI into an interior designer's workflow without getting in the way of the work.
Tabs everywhere. RAG pipelines. Vector stores. An iOS app. A browser extension. Many pieces with so much power.
The friction I was feeling as the architect was the exact friction I was building Patina to remove.
That's when the phrase landed: artificial integration. Not artificial intelligence. Integration. The discipline of designing workflows where AI reduces cognitive friction instead of adding to it.
And it's quietly becoming the variable that separates the AI rollouts that work from the ones that get walked back in eighteen months. MIT's Project NANDA studied 300 enterprise AI deployments this year and found roughly 95% are failing to deliver measurable P&L impact — despite an estimated $30–40 billion in enterprise spend. They don't blame the models. They blame the gap between the tools and the way people actually work.
A manufacturing COO in that study put it more plainly than I could:
"The hype on LinkedIn says everything has changed, but in our operations, nothing fundamental has shifted."
That sentence has been sitting with me for weeks.
The Experience Is the Deliverable
I've spent more than a decade leading enterprise integration work. If you've ever walked into a boardroom, pressed one button, and had a global video call start — someone like me designed that experience.
The metric on that work was never uptime. It was confidence.
If a user didn't trust the system, they wouldn't use it. They'd text someone. They'd pick up a phone. The most expensive, beautifully engineered system in the world becomes a paperweight the moment someone hesitates before pressing the button.
That's the lesson the AI industry is still learning the hard way: people do not adopt systems they do not trust.
BCG's 2025 study of more than 1,250 firms put a stunning number on it. The companies actually capturing value from AI follow what's now known as the 70-20-10 rule: 70% of resources on people and processes, 20% on technology and data, 10% on algorithms. Laggards invert that ratio almost perfectly.
The companies winning at AI are spending nearly three-quarters of their investment on the human side of the equation.
You don't build for what users ask for. You don't build for what you assume they want. You build by watching how people actually operate, then bring the technology to meet them where they already live.
The equipment is a line item. The seamless human experience is the deliverable.
Everything else is wiring.
The Wrong First Question
Most AI strategy conversations start with the wrong question.
"How many people can we replace?"
I understand the gravity behind it. Some tasks will disappear. Some roles will evolve dramatically. Pretending otherwise isn't useful to anyone.
But the organizations that win in this next era are going to ask a different question — and this is where my background in data-driven operations starts to matter in a way it never has before.
I've worked inside both kinds of organizations. I co-founded Jydo. I've operated inside enterprise environments at the scale of Starbucks. The difference between how those worlds move isn't talent or ambition. It's friction.
Startups move fast because proximity reduces friction. Builder, customer, and decision-maker are close enough that an idea can move from conversation to execution in a single afternoon.
Large enterprises operate under what I've come to think of as operational gravity — the slow accumulation of approvals, reporting, handoffs, and overhead that pulls attention away from the work that actually creates value. For years, the only real tool we had to fight that gravity was data. And data was slow.
AI changes the speed of that work entirely.
For the first time, the way I think about process — observing how work actually moves, finding the gaps between intention and execution, designing systems that let humans do their best thinking — is the way executive teams are being forced to think too.
The opportunity isn't efficiency. It's leverage. It's enabling larger organizations to operate with the clarity and agility of a startup without sacrificing the scale that makes them powerful.
The companies that navigate this well won't be the ones with the most tools. They'll be the ones helping their people evolve alongside the technology instead of positioning the technology as the thing replacing them.
That's the only version that compounds.
Mapping the Flow of Thought
When I lead AI adoption, I don't start in the executive suite. I start on the floor.
I want to sit with the person doing the actual work. What does their Tuesday look like? Where are they spending attention they resent spending? Where does the system create hesitation? Where do they wait, and what are they waiting on?
Successful AI adoption isn't really about software. It's about trust. If the system creates anxiety, people route around it. If it creates confidence, they lean into it. Emotional friction and technical friction are usually the same problem wearing different clothes.
From there I work upward — through the managers shaping process, through leadership setting direction. By the time I recommend a tool, I've mapped how attention, decisions, and effort actually move through the organization.
The data tells one story. The people tell another. The integration lives where those two stories meet.
Most companies misread that gap. They optimize the wrong column. The left column — throughput metrics, standardization, compliance, capability, efficiency — is procurement. The right column — signal, flow, trust, orchestration, confidence — is integration. Tools are never the strategy. The strategy lives in the gap between your people and the outcomes they're trying to reach.
I run the same approach at home. I'm building AI agents to support the daily rhythms of our family — not as a novelty, but because you have to live inside a feedback loop to really understand it. When the system you designed affects the morning routine of a seven-year-old, you learn very quickly what works and what was a clever idea on paper.
You can't outsource that learning. You can only live inside it.
The Pacing of Change
The instinct with new technology is to centralize control — force adoption quickly, standardize behavior immediately, optimize before people have had time to understand what's changing.
But people cannot integrate change faster than they can emotionally metabolize it.
I learned that long before I applied it to systems. Sobriety taught me that sustainable change only works at the pace trust can absorb — and that pushing past that pace doesn't accelerate transformation, it breaks it. The same physics operate inside organizations. Every rushed rollout I've ever seen failed for the same reason every rushed personal change fails: the system underneath wasn't ready to hold the new behavior yet.
The best implementations don't make people feel replaced. They make people feel more capable. They give people room to participate in the evolution of their own workflow — to develop trust, form habits, and discover where the technology genuinely helps versus where it simply adds complexity.
It requires patience. It requires listening. And it requires leaders willing to admit they don't fully have the answers yet either.
What Artificial Integration Actually Is
So let me put a sharper edge on this.
Artificial integration is the discipline of designing workflows where AI reduces cognitive friction instead of increasing it. It's a different problem than building AI, and it requires a different set of principles than most organizations are operating under.
The ones I keep coming back to:
- Observe work before automating it. The map is not the territory. The job description is not the job.
- Optimize for trust before efficiency. Adoption is the rate-limiting step. Nothing else matters until people lean in.
- Reduce context switching, don't add to it. Every new tool is a tax until proven otherwise.
- Keep humans close to judgment. Automate the friction, not the decision.
- Make the system quieter, not louder. The best integration is invisible.
- Measure confidence, not just throughput. If people don't trust the output, throughput is theater.
These aren't theoretical. They're the operating principles I bring into every engagement, and they're what separates an AI rollout that compounds from one that gets quietly walked back in eighteen months.
AI is going to get embedded into everything. That part is no longer in question. What is in question is whether that integration gets directed by people who understand how work actually moves — or whether it happens the way most technology waves happen: with the tools racing ahead of the thinking and the people left to clean up after.
Because the seamless human experience is still the deliverable.
The rest is just wiring.
If you're leading an organization through this transition and want to talk about what artificial integration could look like in your context — I'd welcome the conversation.