The thesis

Every company is under pressure to adopt AI, and nearly all of them are asking the same question: how do we use it? It sounds reasonable, but it is the wrong question. “How” produces pilots: a chatbot here, a copilot there, chosen by intuition and measured by feel. Most stall quietly after the demo, and nobody can say what was gained.

The better question is where does AI belong. Which work, done by which people, is actually worth transforming? That question is harder, because answering it requires something almost no company has: an accurate picture of how its work actually happens.

The real structure of a company is not its org chart. It is its workflows, and they are written down nowhere. Org charts say who reports to whom. Process documentation describes the process as it was designed, not as it runs today. Surveys collect opinions, and people are famously unreliable narrators of their own work. No one would renovate a building without blueprints. Companies transform how they work without them every day.

unpak exists to draw those blueprints. We reconstruct how work actually happens from observed activity: the real workflows, broken into their real steps, with honest time counts and recurrence measured across the workforce.

Because the picture is built from evidence, everything downstream of it can be checked. When we recommend where AI belongs, the recommendation traces back to the observed moments it came from. Anyone can follow the reasoning, step by step, down to the raw record, and decide for themselves that it holds.

We call this discipline AI transformation intelligence: a living picture of where AI belongs in an organization, built before you decide, used while you build, and kept current after you deploy. It runs through everything we make: the strategy that decides what is worth building, the platform that holds the picture, and the operations that keep it true once AI is in the work.

Transformation is usually treated as a bet. With an accurate picture of the work, it becomes a design problem. That is the thesis.