Research · Cross-Industry Analysis
Does size predict AI savings?
Analysis of 109,493 US companies with joint revenue + AI-addressable savings data. Headline finding: median Tier-1 savings is 20% of annual revenue, with cross-industry Pearson r = -0.02 between revenue scale and savings-%.
Companies analyzed
109,493
of 134,368 indexed
Median savings %
20%
tier-1 ÷ annual revenue
Cross-industry r
-0.02
revenue × savings-%
Industries (NAICS-2)
23
cohorts ≥ 30 companies
Revenue × savings-% — full sample
Each dot is one company. The regression line (dashed blue) is the least-squares fit over log-transformed revenue. A flat or shallow line means savings-% is roughly constant across the size spectrum; a steep negative slope would mean larger companies see proportionally less waste.
Across the full sample the slope is mildly negative (r = -0.02). Industry-level dynamics dominate — the heatmap below shows where size effects are strongest.
Industry × size-band heatmap
Median savings-% by 2-digit NAICS sector and size band. Darker cells = larger addressable waste relative to revenue. Top 10 industries shown by sample size.
| Industry (NAICS-2) | mid-size regional | regional multi-site | national operator | enterprise |
|---|---|---|---|---|
| Healthcare(529) | 20% | 20% | 11% | 3% |
| Professional Svcs(507) | 20% | 20% | 20% | 6% |
| Manufacturing (Metal/Comp)(326) | 20% | 20% | 7% | 3% |
| Finance & Insurance(297) | 20% | 14% | 10% | 3% |
| Education(290) | 20% | 20% | 20% | 11% |
| Other Services(214) | 20% | 20% | 20% | 20% |
| Information(212) | 20% | 20% | 11% | 3% |
| Admin & Support(211) | 20% | 20% | 20% | 10% |
| Accommodation/Food(178) | 20% | 20% | 20% | 3% |
| Construction(176) | 20% | 20% | 7% | 3% |
Cell value = median savings_estimate_tier1 ÷ annual_revenue, expressed as a percent. Cells with fewer than 5 companies are blanked. Numbers in parentheses = total companies in that NAICS-2 cohort.
Top 5 industries by median savings-%
Cohorts with the highest median Tier-1 savings as a fraction of revenue. Read these as "where AI deployment is currently underexploited relative to the operating efficiency it could deliver."
- 1
Professional, Scientific & Technical Services
Median savings: 20% of revenue · 14,245 companies · within-cohort r = -0.05
- 2
Health Care & Social Assistance
Median savings: 20% of revenue · 13,770 companies · within-cohort r = -0.18
- 3
Manufacturing — Metal, Machinery, Computer
Median savings: 20% of revenue · 8,395 companies · within-cohort r = -0.07
- 4
Educational Services
Median savings: 20% of revenue · 8,213 companies · within-cohort r = -0.73
- 5
Finance & Insurance
Median savings: 20% of revenue · 7,910 companies · within-cohort r = -0.07
Methodology
Source. The dataset is content.ai_opportunity_companies, filtered to rows with both annual_revenue > 0 and savings_estimate_tier1 > 0. That yields 109,493 companies of the 134,368 indexed in the AI Adoption Index.
Savings estimate. Tier-1 savings is a per-company estimate of addressable labor waste that AI agents could eliminate within 12-24 months given current model capabilities. It's bottom-up: we identify high-confidence automatable tasks (Tier 1 = production-ready agent solutions today) using the company's industry, headcount, and tech-stack signals, then apply industry-standard cost loadings.
Statistics. Median rather than mean for cohort summaries — the distribution has long tails. Pearson r is computed in-sample on log-transformed revenue; values close to 0 indicate no monotonic relationship at the global level, even when individual cohorts show structure (Simpson's paradox is a real risk here, hence the per-industry heatmap).
Caveats. Companies without published revenue are excluded; this biases the sample toward larger, public-facing firms. Tier-1 savings only counts workloads that are already deployable today — Tier-2 (12-36 month roadmap) and Tier-3 (research-stage) workloads are excluded from this analysis. Healthcare (NAICS-62) is over-represented because Apollo enrichment coverage is denser in that vertical.
Keep exploring
Source dataset
AI Adoption Leaderboard
80,000+ companies ranked by AI adoption — the dataset behind every research page.
Annual benchmark
State of AI Adoption 2026
80,789 US companies scored across 25 sectors and 540 sub-industries — flagship annual benchmark.
Adjacent research
WARN Layoff Atlas
12,623 US mass-layoff filings cross-referenced with AI adoption scores — industry-level r=+0.33.
Visual essay
The Architecture of Intelligence
68 years of AI architecture in a scroll-driven visual journey — Perceptron through Transformer scaling.