The $186 hidden tax your AI program is generating every month.

There is a number in McKinsey's 2025 AI research that most CxOs have not seen. $186. Per employee. Per month. That is the average hidden cost of fixing low-quality AI outputs, what the research calls "workslop."

At 1,000 employees, that is $2.2 million per year. Not in AI investment. In the cost of cleaning up after bad AI.

The mechanism is straightforward. Employees use AI to produce outputs faster. The outputs contain errors: wrong figures, hallucinated facts, style violations, tone mismatches. Other employees spend time finding and fixing those errors. That rework time is invisible in most productivity dashboards. It does not show up as "AI cost." It shows up as nothing.

Two other numbers sit alongside the $186 figure and make it worse. 40% of AI productivity gains are lost to rework before they reach the business. And 75% of AI time savings evaporate, because only 1 in 4 workers knows how to reinvest recovered time productively.

Together, these three numbers describe an AI program that is simultaneously generating cost and destroying value. The tools are running. The ROI is not arriving.

The fix is not a better model. It is two structural changes. First: deterministic AI architecture, systems that produce consistent, verifiable outputs rather than probabilistic ones that vary with every call. Second: AI literacy training that teaches people what good AI output looks like, how to evaluate it critically, and what to do with recovered time.

Both are built into every Weaver engagement. Not as optional add-ons. As go-live requirements.