This Anthropic report is way more important than MIT’s « 95% of AI pilots fail » study.
Must-read for anyone who’s CEO said « we’re going to be AI first »: https://lnkd.in/g-3BEy9Z. Amazing job to the folks who wrote/contributed to this (Ruth E. Appel, Maxim Massenkoff, Peter McCrory, R. Miles McCain, Ryan H., Tyler Neylon, Alex Tamkin)
One of many jaw-dropper charts (p 41): AI agents just crossed an inflection point. Claude maintains 65%+ success on 8-hour intellectual tasks.
THE HOLY S*** IMPLICATION: Even with 1 hour of checking, you save 7 hours.
What the chart shows: For short tasks (under 2 hours), Claude and generic APIs both hit ~60% success. But as tasks get longer, Claude stays above 65% while traditional approaches fall below 50%. AI just got good enough at full-day work to fundamentally change how we approach complex tasks.
My question: Can you actually see where this is happening in your org and what tasks are shifting?
If not, this is what Lanai does. We see AI usage across all platforms. I’m passionate about this because: if you don’t see it, you cannot manage it. You’ll agent-shore anything you can and get short-term labor cost wins that are very real.
But here is the sticky wicket: entry-level roles where future leaders learn judgment are disappearing first. But these are the exact roles that build the business intuition necessary for managing AI effectively.
Right now, as an example, we use GenAI assistants and agents at Lanai that do the work of at least 3 – 6 FTEs, but we absolutely need humans in the loop as guides, gurus, and guardrails. Those humans have 10-15 years of experience doing what AI mediated support in co-pilots and agents do now do. What happens in 4, 6, 10 years? Who trains the next generation of leaders to know when AI is right, when it’s close, and when it’s dangerously wrong? How do you ensure the short term cost bump is not compromising the core fundamentals of prospect and customer health/loyalty?
Read the report in full. Then ask: do we actually know what’s happening, or are we just really winging it? And – seriously – if we can help, we’d love to. This is what we spend all our time on.