AI and the Analyst Labor Market

Move the assumptions and watch supply and demand for analytics talent shift. Faded lines are the baseline (salary and employment indexed to 100); bold lines are your AI scenario, holding everything else constant. All parameters are illustrative assumptions, not calibrated estimates.

Salary
Analyst FTEs
Analysis output
Cost per analysisUnit cost
Demand, AI scenario Supply, AI scenario Baseline
Scenarios

Demand for analysts

Slop halves it3× output
Useful analysis per analyst. Below zero, analysts use AI to generate decks and memos that colleagues must wade through, so each useful answer costs more than before.
NoneThree-quarters
Fixed pile of workNear-infinite work
How much more analysis businesses commission as it gets cheaper. This bundles every limit on using more of it: data access, review capacity, and managers' ability to act on the answers.

Supply of analysts

Same poolPool doubles
Extra people able to fill analyst roles at today's salary, as AI lowers the bar for marketers, ops, and finance staff.
Advanced
Scarce, slow to trainEasy to enter

Results are a long-run equilibrium after hiring and pay adjust. A lower employment number does not say how much of the change comes from layoffs versus slower hiring.

How the model works

Demand is derived

Firms buy analysis, and analysts are an input. Each analyst FTE produces m units of analysis, so labor cost per analysis is W/m. Demand for analysis responds to that cost with elasticity η, and business users take over a share f of it themselves.

FTE demand = A · (1 − f) · m^(η − 1) · W^(−η)

The exponent η − 1 carries the argument. Before other effects, a productivity gain raises demand for analyst labor when η > 1 and lowers it when η < 1 (a simplified Hicks-Marshall derived-demand result). At η = 0 the work is a fixed pile, and every productivity gain cuts FTEs one for one. Slop is m below 1: AI-generated material that costs colleagues reading time lowers useful output per analyst, and the same exponent runs in reverse. The exact threshold of 1 depends on analyst labor being the only cost; adding software, data, and review costs moves it.

Supply shifts too

AI lowers the skill bar, so more people qualify for analyst roles. That shifts supply out by the "more qualified candidates" share at every salary.

FTE supply = B · (1 + entry) · W^(ε)

How a demand shift splits between pay and jobs depends on ε relative to η. A demand shift worth x% more FTEs at today's salary raises salaries by about x/(η + ε)% and FTEs by about ε·x/(η + ε)%, so a responsive talent pool turns more of the shift into jobs and less into pay.

Outputs

Analyst FTEs and salary come from the closed-form equilibrium, calibrated so the baseline clears at 100 and 100. Analysis output is FTEs × m and covers only analyst-produced work, excluding what business users do themselves. Cost per analysis is W / m and excludes AI tooling. Spending on analyst labor is W × FTEs.

Employment is in full-time equivalents, which assumes hours per analyst stay constant.