An operating manual for organizational leaders starting from zero
Trellis
A framework for AI adoption.
Executive summary
There is no such thing as an AI strategy. There is only the business strategy — and then the question of whether AI can advance it.
Most frameworks for AI adoption get this backwards. They start with the technology, build elaborate plans for how to deploy it, and ask the organization to commit to a vision before anyone has evidence about what AI will actually do for the business. The result is months of planning, a thick deck, and very little movement. By the time the organization is ready to act, the landscape has shifted again.
Trellis is the opposite. It is a framework for organizations that want to adopt AI deliberately but without freezing in place. It assumes that the right structures, policies, and metrics cannot be designed in advance — they have to be built up from real project experience. Its job is to keep the right questions visible without forcing premature answers, and to give leaders a way to start moving while the picture is still incomplete.
The framework rests on three operating principles:
Strategy first, not technology first. Every AI adoption effort begins with a clear picture of what the business is, what makes it work, where it is going, and what threatens it. AI is evaluated against that strategy — not the other way around. A perfectly valid outcome of this work is the conclusion that AI does not make sense for parts of the business. That conclusion can only be reached honestly if the strategy is the starting point.
Hypotheses, then hardening. Each project starts with lightly-sketched answers to a small set of important questions — about readiness, about plan, about risk. Those answers are good enough to move with. They are refined as the project progresses and evidence accumulates. Nothing is figured out in advance that does not need to be figured out in advance.
Patterns become policy. When the same answer keeps showing up across multiple projects — the same data gap, the same governance question, the same incentive misalignment — that pattern gets extracted into a global policy. Policy is built up from project experience, not handed down from above before projects begin.
These principles are organized into a three-layer architecture. The Strategic Loop at the top is the organization’s abstract mental model — what it believes about itself, its market, and where it is going. It is the future org: the company leadership wants it to become. The Global Policy Layer at the bottom is the organization’s concrete existence — the actual structures, rules, functions, and capabilities it has built. It is the current org: the company that actually shows up to work each day. In between sits the Project Layer — the becoming org, where future and present meet through deliberate experimentation. Projects run constantly, generating evidence that flows in two directions: upward, to revise the mental model when something new becomes possible; and downward, to harden the concrete existence as patterns from project experience get extracted into policy. Strategy is the company we want to be. Policy is the company we are. Projects are how we become it.
The framework prioritizes momentum at the start and steering as it goes. Early on, the focus is on getting projects running and learning what AI can actually do for this specific organization. As projects accumulate, patterns emerge, and the global policy layer begins to fill in. Eventually the organization has accumulated enough evidence to revisit the strategy itself — and ask the larger question of whether real business model transformation is possible.
This document is the operating manual for that journey. It assumes an organization at zero: no AI initiatives in flight, no data infrastructure decisions made, no governance structures defined, just a desire to begin. It covers what to do first, what to do next, and how to know when it is working.