The Integration Tax Just Collapsed
The biggest opportunity in AI for business isn't the model — it's that the cost of wiring software into real workflows fell by an order of magnitude. The risk is pretending the verification layer was never load-bearing.
Honestly? I think the biggest opportunity isn't AI itself. It's that the integration tax just collapsed.
For about twenty years, the bottleneck between a good idea and a system that actually runs was the boring stuff. Wiring it into an existing workflow. Surfacing it where decisions actually get made. Caching, retries, all the plumbing nobody claps for at the demo. AI knocked that cost down by an order of magnitude.
So a solo operator can now ship what used to take a team. A small business can deploy production tooling that was enterprise-only a year ago. The advantage goes to whoever can map a workflow first and ship into it. It's the same shift that happened when EHRs replaced paper, or calculators replaced slide rules: the tool changes, but the professional standards stay yours to enforce.
The risk is the same thing, flipped upside down
People treat AI output like it's the final answer when it's really a starting point.
I've watched this pattern over and over. The tool demos great in a conference room. It pilots in a real workflow. Three weeks later usage is at 12% and the project quietly disappears. The model wasn't wrong. The business just skipped the verification layer we always had for each other — the peer review, the "wait, does that actually match what we agreed?" Pull that layer out, pretend it wasn't load-bearing, and errors stack up silently until something visible breaks. In a regulated environment, that's not a bug. That's a lawsuit.
What the demo never shows
Here's a real example from my own work. I built a retrieval-augmented agent that plugs into an enterprise knowledge base — staff ask a question, it pulls grounded answers in real time. The model was the easy part. The hard part was the verification scaffolding around it: knowing when an answer is actually grounded versus when it's just pandering to the vibe. Surfacing confidence honestly. Making the failure mode graceful instead of confidently wrong.
None of that shows up in a demo. All of it shows up when a clinician is leaning on the thing at 3 AM on a busy unit.
It happens outside healthcare too
Air Canada got forced to honor a bereavement fare its own chatbot hallucinated — the tribunal basically ruled that the chatbot was speaking for the company. Law firms have been sanctioned for filing briefs full of AI-fabricated case citations. Nobody asked the model to lie. Nobody verified, either.
So that's my take. AI hands businesses superpowers, and businesses tend to use superpowers carelessly. The ones that win this decade are the ones who treat AI as a force multiplier on judgment — not a replacement for it.