ROAI is the four-quadrant return framework for AI investment in legal functions. The traditional return-on-investment conversation, focused on time saved per task and cost per matter, captures one of the four quadrants and misses three. Functions that justify AI spend on productivity alone underweight the investment relative to its institutional value, build governance posture poorly because the funding case did not require it, and fail to position the function competitively against peers evaluating AI on a fuller frame.
ROAI is the case the board sees. Single-quadrant cases lose to fuller-frame cases at every funding cycle. A function that presents productivity savings alone signals to the board that its AI investment is a cost-reduction exercise; a function that presents the four quadrants signals an institutional posture build. Boards fund the latter at materially different scales.
The four quadrants
Q1 Productivity: hours saved per task, throughput gained, capacity freed. Typically 15 to 30 percent of total ROAI. The CFO question; the smallest of the four. Measurable in months at the first quarterly review. The category most easily verified against pre-AI baselines, and the category most easily overstated by vendor demos.
Q2 Defensibility: avoided regulatory exposure, avoided incidents, audit readiness. Typically 40 to 60 percent of total ROAI for institutional legal functions. The largest single quadrant. Visible at the first audit cycle (quarters, not months). The board question; the hardest to measure prospectively because the function avoids costs that did not occur. Boards that understand insurance accept the framing; boards that demand only positive-side returns will struggle with it.
Q3 Institutional: the function’s standing inside the organisation as a contributor to strategic capability. Typically 15 to 25 percent. The longest payoff horizon; the executive committee question; compounds across years. The quadrant that determines whether the legal function is consulted on AI questions broader than the legal department’s own use, or routed around.
Q4 Category positioning: peer standing, talent attraction, client outcomes where AI capability differentiates. Typically 10 to 20 percent. Rewards early Concept-stage commitment; first-movers capture most of this quadrant; late movers capture very little regardless of absolute investment. The most variable quadrant; the most steeply rewarded by deliberate action.
How to apply the frame
ROAI is applied per AI investment and per investment portfolio. For each candidate AI investment, the function evaluates expected return across all four quadrants, presents the four-quadrant case to the board, and tracks actual return against the projection across the investment lifecycle. The portfolio view surfaces structural gaps: a function that has not made a Defensibility-quadrant-focused investment in the last twenty-four months is exposed, regardless of how productive its other tools are.
Operationally, ROAI tracking lives in the Measurement Operating Layer. Each quadrant has measurable proxies. Productivity: cohort-controlled time-saved studies against pre-AI baselines. Defensibility: avoided-incident counts plus reduced regulator-response time plus Evidence Register completeness. Institutional: budget trajectory, board agenda inclusion, cross-function consultation frequency. Category positioning: peer-function benchmarks plus talent acquisition data plus client retention attributable to AI-capable service.
Where it sits in the framework
ROAI is the measurement output of Pillar 5 (Use Cases, Execution and Measurement) and Pillar 7 (Maturity, Benchmarking and Progression). It interacts with Defensibility (the operational system that produces Q2 return), Risk Taxonomy 2026 (the inventory against which Q2 is measured), and Vendor Index (the procurement standard that determines Q2 vendor contribution). The Annual Legal AI OS Index will report, from 2027 onward, the aggregate distribution of legal functions across ROAI performance, validating the framework’s structural predictions against industry data.