90% Use AI. Only a Handful Have Scaled It.

Over the summer, a headline claiming that 95% of AI pilots fail went viral. It didn’t really make sense and has since been criticized. The reality is that Gen AI usage has become mainstream.

McKinsey’s State of AI in 2025 report provides data to support this narrative.

Nearly nine in ten organizations now use AI in at least one business function, up from 78% last year. One in three are actively scaling their programs across the enterprise. While only a small share have reached “fully scaled” maturity, the trend is unmistakable: AI is no longer an experiment. The data tells a clear story.

The share of companies using generative AI alone jumped from 33% in 2023 to nearly 80% in just two years. The real question isn’t whether firms are using AI. It’s whether they’re learning fast enough to turn pilots into practice.

AI agents continue to be the talk of the town, notably gaining traction in IT and knowledge management. Professional services are still early. Only a small subset of respondents in that category report agentic AI use that has reached the scaling phase. If you're in the AI bubble, it may feel like you’re constantly behind (I know I feel that way sometimes), but the data shows there is still plenty of opportunity ahead.

Lastly, it was interesting to note the unevenness across industries and functional areas. AI use is strongest in media, telecom, and insurance. Professional services sit further down the curve, though the patterns are familiar. The focus areas are knowledge management, marketing, and risk. No surprises, it’s the same places where firms have been experimenting with process and technology for years.

What stands out in the data isn’t how many are using AI, but how it's yet to become part of daily work for most. Many organizations are running pilots or small-scale projects. Scaling means reworking how teams share, decide, and deliver work, and rethinking processes from the ground up.

I believe we're well on our path to where AI becomes an ordinary part of how work is done.

Finds

Team dynamics after AI

AI tools push teams toward what is measurable and easy to automate, at the cost of nuance, niche skills, and messy human judgment. This risks flattening roles, weakening cross-discipline translation, and breaking feedback loops that build expertise. Instead, prioritize people who focus on sense-making and deep craft, rather than outsourcing decision work to clear systems.

Law firm AI use cases and in-house legal’s needs don’t match

Law firms often promote large AI projects that don't align with the everyday needs of their in-house teams. In-house teams need practical help with single contracts, process reviews, risk training, and clear policies. To benefit, in-house lawyers must upskill and law firms must offer targeted, evidence-backed support.

Cost of intelligence continues to drop

Frontier AI costs have fallen 64-fold with the release of Grok 4 fast. New models are expected from Google, and I look forward to seeing how the economics are impacted.

Until next time.