Anthropic released their fourth Economic Index report in January 2026, analyzing 2 million interactions with Claude across both consumer and enterprise deployments. The findings challenge the assumptions driving most enterprise AI investments.
One data point stopped me cold: the correlation between how sophisticated the human prompt is and how sophisticated the response from Claude is sits at 92%.
That is not a loose relationship. That is near-perfect. Your AI is only as good as the people using it.
The reality behind the productivity promises
The report challenges several assumptions that have been driving enterprise AI investments.
Success rates decline as tasks get harder. Claude succeeds roughly 70% of the time on straightforward work. For complex tasks requiring college-level expertise, that drops to 66%. One in three complex tasks fails. That changes your ROI math considerably when you factor in rework, validation, and error handling.
Autonomous AI hits a wall quickly. For API-based automation with no human involvement, the success rate of Claude crosses below 50% at around 3.5 hours of equivalent human work. The technology cannot reliably handle a half-day project on its own. When humans stay in the loop, guiding, correcting and iterating, that threshold extends dramatically. Collaboration works. Delegation does not. At least not yet.
The productivity gains are real but overstated. The Anthropic baseline estimate suggested AI could boost labor productivity by 1.8 percentage points annually over the next decade. When they adjusted for actual failure rates, that dropped to roughly 1.0 percentage point. Still meaningful, but nearly half the optimistic projection. CFOs building business cases on vendor promises should take note.
The readiness gap is the real story
What struck me most was not any single finding. It was the pattern underneath them.
The organizations extracting value from AI share a common trait: their people know how to work with it. The 92% correlation between input quality and output quality means workforce capability is not a nice-to-have in your AI strategy. It is the strategy.
I saw this firsthand with a Canadian financial services client. They had already purchased AI tools. Leadership was frustrated with slow adoption and inconsistent results. When we assessed their actual readiness, the picture clarified quickly. The tools were not the problem. The organization was not prepared to use them: processes were not designed for human-AI collaboration, success metrics did not exist, and training had been an afterthought.
After addressing those gaps, the same tools they would have written off as disappointing started delivering. The technology had not changed. The readiness had.
What this means for AI investment decisions
The Anthropic data suggests a few practical implications.
Budget for failure. If your business case assumes AI works every time, your projected returns are likely overstated: in the Anthropic report, the success-adjusted estimate of the productivity gain is roughly a third to a half lower than the unadjusted one. Build validation and rework costs into your projections from the start.
Invest in people before platforms. The gap between organizations that succeed with AI and those that do not is not the tools they buy. It is whether their workforce can collaborate effectively with those tools. Training and change management are not line items to cut; they are where the ROI actually comes from.
Assess before you deploy. Understanding where your organization stands, across skills, processes, data readiness, and governance, determines whether your AI investment pays off or becomes expensive shelfware.
The bottom line
The Anthropic report gives us the data. The technology works when organizations are ready to use it. The question is whether your organization is, and whether you will find out before or after you have made the investment.
The teams that pull ahead are the ones that measure where they actually stand, honestly, with real data, before they scale. That is the whole game.
Source: Anthropic, The Anthropic Economic Index report: Economic Primitives, 15 January 2026. The 0.92 correlation, the 70% and 66% success rates, the 3.5 hour threshold and the 1.8 percentage point estimate are from the full report (PDF linked from that page).
STRATAGEM LABS

