A general contractor knew its people were already using AI but had no operating model for it. We ran an organization-wide survey and 15 interviews across 8 departments, measured the gap between AI appetite and readiness, and turned 109 named use cases into a sequenced operating strategy: 17 solutions across one governing foundation and three phases, each tied to business goals and to a governance design that keeps people in the decision.
The problem
The company did not have an AI problem so much as an operating-model gap. Executive appetite was high and employees were already using AI to move work forward, while policy, approved tools, data controls, and repeatable workflows had not caught up. The readiness was there. The governance was not. Buying more tools would not close that gap.
The finding
The most telling discovery was shadow AI. Across every function and seniority level, people were already using personal ChatGPT and Claude accounts, putting contract language, spreadsheets, and project information into tools the company could not see or control, with no data-loss-prevention layer behind them. That made it the single largest untracked risk in the business. It was also the clearest proof the workforce was ready. The strategic response is not prohibition, which never holds. It is a sanctioned default good enough to replace the workaround.
What we did
We started with the business and the people, not a tool:
- Surveyed the whole organization and interviewed 15 people across 8 departments, from leadership to the daily work, surfacing use cases in the company's own words.
- Measured the operating gap, high appetite against low readiness, turning a vague "we should do something with AI" into a quantified case leadership could act on.
- Quantified where the time and risk actually go, hundreds of hours a week lost to document search and email, multi-month RFI cycles, a seven-figure design-drift event, presented as directional signals, deliberately not inflated into headline savings.
- Distilled 109 named use cases into 17 sequenced solutions. The hard part was not finding ideas. It was grouping them into reusable capabilities and choosing the order of operations.
What made it more than a survey
The rigor is in the design, not the deck:
- A sequenced portfolio, not a wish list. One governing foundation, policy, sanctioned tools, training, decision gates, plus three phases that move from control to advantage, with stage gates that decide whether each phase advances, gets redesigned, or stops.
- Sequencing tied to the business's reality. The order accounts for readiness and for the company's upcoming enterprise-systems transition, so early work builds portable capability instead of throwaway pilots.
- Governance built in, not bolted on. One principle runs through every solution. AI prepares, people decide. Each one has to expose its source, name an accountable owner, and preserve the human decision right, with explicit guardrails against autonomous action on consequential work.
The result
Instead of chasing tools, leadership has an evidence-backed operating strategy: where AI pays, in what order, on what governance foundation, and who owns each decision. It is the difference between AI activity and an operating system for AI.
