AI Spend ROI.

The 25% potential figure is the realistic upper end of productivity lift from a well-installed AI-native OS across knowledge work, based on the engagements we've sized and the published literature on agentic workflow automation. It is not a marketing number — it's what the math supports when models, knowledge, agents, workflows, and observability are all in place. Most firms land at 18–28%; we've chosen 25% as the conservative middle.

Lift captured = (observed lift × payroll) ÷ (25% × payroll). A firm at 6% observed lift is capturing 24% of what's possible. The other 76% is the operating gap.

$ left on table = (potential lift − observed lift) × total fully- loaded payroll. This is the dollar value of the productivity not captured at your current operating layer.

ROI on AI spend = actual lift dollars ÷ total annual AI spend. This number is uncomfortable on purpose — most firms our size find it's below 1.5× even with charitable assumptions about observed lift.

Sources: McKinsey's State of AI 2025 puts AI investments producing no measurable business impact at 80%. Anthropic, OpenAI, and several independent productivity studies put well-installed AI workflow systems in the 25–40% lift range. We use the conservative end.

If your lift captured is under 30%: typical. The license stack is doing some work but the operating layer underneath isn't built for the capability. This is the band where Beachhead Build or Foundation Build pays back in under 12 months.

If your lift captured is 30–60%: above average. You probably have one or two well-instrumented workflows already and are looking for the next domain to attack. Beachhead Build on a specific domain typically the right call.

If your lift captured is 60%+: rare. You're likely already running an internal AI-ops function. The conversation we'd have is about specific workflows or a Strike Goal on a narrow KPI rather than a broad Install.