AI Automation Exposure: Geographic and CRE Implications

✓ Knowledge-work occupations · 50 states + DC · BLS OEWS May 2025 · HBS automation scores

AI automation exposure tracks where white-collar work concentrates, but the most immediate clerical-automation risk sits in a different, more dispersed set of states.

Each occupation carries a single automation-exposure score; a state's exposure is a function of its occupation mix, since the scores are identical everywhere and only the mix varies. The map below covers knowledge-work occupations only (management, business and finance, computer and math, engineering, legal, the sciences, and office support).

Weighted automation score by state

Sum of (occupation % of state jobs × automation score) over knowledge-work occupations only. Color scale capped near the extremes to show variation across the middle. Scale: 12.4 (Mississippi) to 27.1 (District of Columbia).

US choropleth of weighted AI automation score by state, knowledge-work occupations

Key Observations

Certain back office functions are concentrated in low wage states.

Customer Service Representatives: weighted automation score by state

Weighted automation score is each state's share of jobs in this occupation times its 0.57 automation score; location quotient shown beside each bar. 2.6M US jobs.

Bar chart of Customer Service Representative weighted automation score by state

Explore the data

Rank states two ways, drill into one state's occupations, or trace where a single occupation concentrates. Filter by job family and automation score. The map and charts update live.

Controls

Rank states by

Job families

Automation score

min 0.00max 1.00

Weighted automation score by state

How two of the measures are calculated

Location quotient (LQ): a state's share of its own jobs in an occupation divided by that occupation's share of national employment. An LQ above 1.0 means the state is over-indexed in that work; 2.0 means twice the national concentration.

Weighted automation score: an occupation's share of a state's jobs multiplied by its automation score. An occupation that is 2.0% of a state's jobs at a 0.60 score gives 2.0 × 0.60 = 1.20. Summed across occupations, this recovers that state's overall automation index.

Methodology and how to read this

Exposure here is structural, not realized. Each occupation carries one Harvard Business School automation score; a place's exposure is its occupation mix weighted by those fixed scores. The honest reading is which regions are most exposed, not which have been most affected. At the national level the correlation between automation score and 2022 to 2025 employment change is near zero, so this measures the shape of the workforce, not job loss to date.

The weighted automation score is the sum of each occupation's share of a state's jobs times its automation score; for a single occupation it is that occupation's share times its score. A location quotient (LQ) above 1.0 means a state is over-indexed in an occupation relative to the national rate.

Employment comes from the BLS OEWS May 2025 state cross-industry file; scores cover roughly 98 percent of state employment, with suppressed cells excluded and territories omitted from the map. OEWS is a point-in-time estimate and is not designed for year-over-year comparison.

Sources: automation scores from Harvard Business School (Srinivasan, Chen & Zakerinia, Working Paper 25-039), one score per occupation, identical across states. Employment from BLS OEWS, May 2025, state cross-industry file. Coverage ~98% of state employment; suppressed cells excluded; territories omitted from the map. OEWS is a point-in-time estimate not designed for year-over-year comparison. Exposure is structural, not a measure of realized job loss.