Why Okram Labs exists
There's a kind of expertise that never quite makes it into a system: the agent who can feel a comp is wrong before she can say why, the ops engineer who knows a device is about to fail from a pattern nobody wrote down, the person who can eyeball a bank statement and know something's off. That knowledge is real, and it's almost entirely illegible to any tool built to help with it.
Why expert judgment
Most software treats expertise as something to automate away — collect enough examples, train it out of the person, ship the model. That's backwards for the calls that actually matter. The interesting problem isn't replacing an expert's read on a situation; it's building something structured enough to hold that read, question it, and hand it back sharper than it arrived.
Why AI needs transparency
A model that outputs a number without a reason is asking for trust it hasn't earned. The more capable these systems get, the more that gap matters — not because the answers are usually wrong, but because nobody can tell, from the outside, when they are. Transparency isn't a compliance feature. It's the difference between a tool you can argue with and one you can only obey.
Why decision systems matter
Every domain has a version of the same failure: a decision gets made, the reasoning behind it lives in one person's head, and the next person has to reconstruct it from scratch or just trust it blindly. A decision system's job isn't merely to make the call. It's to make the call inspectable—so the next person, or the same person six months later, doesn't have to reconstruct the reasoning from scratch.
Why these experiments converge
TESSA, NVEE, and M²W² don't share a codebase, an industry, or even a user. What they share is a shape: take judgment that currently exists nowhere but in someone’s head and give it a structure a system can hold—a claim, its evidence, and its reasoning—while making explicit where human judgment remains necessary.
The twenty years that led here
Twenty years across infrastructure, customer-facing engineering, and product leadership exposed me to very different technologies, customers, and products. The industries changed. The technologies changed. The question didn't.
How do you make expert judgment visible without replacing the expert?
Okram Labs is where I continue exploring that question.
The rest of this site is a collection of experiments in answering that question.