Following the release of Anthropic's updated Economic Index with U.S.-specific data, I've refreshed the AI Labor Market Impact projections. The headline finding: projected impact improved from -7.49% to -6.84%, representing approximately 933,000 fewer jobs negatively affected.

But the real story isn't about better numbers—it's about what the improved data reveals about how AI adoption is actually unfolding in America.

Industry-by-Industry Impact Changes

The improvement wasn't uniform across sectors. Here's how major industries shifted:

The Knowledge Work Question

While current estimates improved across most sectors, five-year projections tell a more nuanced story. The Information and Professional Services sectors show projected deterioration of -0.60 to -0.66 percentage points by 2028.

This creates an interesting temporal pattern: current conditions look better than expected, but the trajectory steepens for cognitive occupations. One interpretation: we're in a phase where augmentation effects dominate because AI capabilities remain partial—but as systems become more capable, we should expect a shift toward greater automation in knowledge work.

Meanwhile, construction and manufacturing show essentially unchanged long-term projections despite improved current estimates. The physical world continues resisting wholesale automation, exactly as theory would predict.

What Changed in the Data

The improvement stems from two sources:

Better measurement: The shift from 20 generic global occupation types to 23 standardized SOC categories with country-specific data enabled mapping 82.9% of U.S. occupations. We can now see regional patterns that global averages obscured.

Actual evolution: Between April and August 2025, organizations continued learning how to integrate AI systems in ways that emphasize collaboration over replacement.

Distinguishing between measurement improvements and real changes matters for interpreting future updates. Not all shifts will represent actual labor market evolution—some will simply reflect better instruments.

What This Means

Two implications stand out:

1. The U.S.'s slightly higher augmentation rate (51% vs global 49%) represents a modest competitive advantage. While the difference is small, policies that reinforce this balance could meaningfully shape outcomes over time.
2. Technology sector concentration is real. While Information represents only 1.8% of total employment, it's experiencing the steepest displacement—creating concentrated geographic impacts in tech hubs.

Looking Forward

This establishes a pattern for quarterly updates following Anthropic Economic Index releases. The fundamental methodology remains unchanged—only the underlying data quality improves.

The 0.65 percentage point improvement represents genuinely good news. But we should be cautious about over-interpreting this as evidence that AI's effects will be mild. The technology continues advancing rapidly, and knowledge work occupations still face significant transition challenges ahead.

Better measurement enables better planning. That's the real value of these updates—not perfect predictions, but progressively clearer pictures of how transformation is actually unfolding, enabling more targeted interventions where they're needed most.

Full methodology: A Factor-Based Framework for Modeling AI's Impact on U.S. Labor Markets

Interactive dashboard: AI Labor Market Impact Projections

Data sources: Bureau of Labor Statistics (August 2025), Anthropic Economic Index (August 2025)

Next update: January 2026