Parand AkbariEconomics · Carnegie Mellon

Research data · 2011–2024

Measuring corporate AI
as it diffuses.

A disclosure-based measurement system that separates signal from deployment and traces when firms adopt, where AI spreads, what firms use it for, and how deeply it is integrated.

I use large-language models to classify corporate disclosures and build annual firm-level measures of AI adoption. The exhibits below show how adoption spreads across firms and industries, how deeply it is integrated, and what firms use it for.

187,191Corporate transcript events
22,974AI-related events analyzed
4,313Public firms
60,382Annual firm observations

Measurement framework

Disclosure → annual measures → adoption history
EVENT

Classify the disclosure

Identify AI exploration and deployment, then classify technologies, applications, integration depth, and named providers.

ANNUAL FIRM MEASURES

Construct annual measures

Combine disclosure evidence for 4,313 firms over 2011–2024.

FIRM HISTORY

Trace diffusion

Identify each firm’s first adoption disclosure and trace disclosed adoption over time and across industries.

“Adoption” on this page means disclosed adoption. A first disclosure is an observed disclosure date, not necessarily the date of internal implementation.

01 / Firm-level diffusion

Adoption is a stock.
Disclosure is a flow.

The figure distinguishes firms that have ever disclosed AI adoption, firms disclosing adoption in a given year, and firms making their first adoption disclosure in that year.

Universe: 4,313 firms
AI adoption and diffusion from 2011 to 2024.
31.0%Ever disclosed by 2024
18.9%Disclosed in 2024
3.3%First disclosure in 2024

These are disclosure-based measures: firms may use AI without discussing it publicly.

Definitions & annual counts
Broad AI; universe: 4,313 firms.
YearEver disclosedDisclosed this yearFirst disclosureEver share
20111111110.3%
2012191180.4%
2013282090.6%
20144837201.1%
20157650281.8%
2016146112703.4%
20172942481486.8%
201844736015310.4%
201957942613213.4%
202070843912916.4%
202189758718920.8%
2022102557412823.8%
2023119672817127.7%
2024133881714231.0%

02 / Across industries

Diffusion is broad—
and sharply uneven.

Within-industry adoption rate. The heatmap shows, for each industry and year, the share of firms that have disclosed AI adoption. The 2024 ranking compares those within-industry rates.

4,256 firms across industries
AI adoption rates by industry from 2011 through 2024, with a 2024 comparison.

Industry coverage includes 98.7% of firms. Rates describe disclosed adoption, not industry performance.

2024 industry rates
Broad AI; panel firms with industry classifications.
Industry groupPanel firmsEver disclosed by 20242024 cumulative rate
Software & Services28621575.2%
Commercial & Professional Services1447954.9%
Media & Entertainment1789553.4%
Semiconductors & Semiconductor Equipment874450.6%
Technology Hardware & Equipment1949247.4%
Insurance843744.0%
Transportation662943.9%
Health Care Equipment & Services32613441.1%
Consumer Discretionary Distribution & Retail1516241.1%
Consumer Services1595836.5%
Telecommunication Services281035.7%
Consumer Staples Distribution & Retail331133.3%
Real Estate Management & Development371232.4%
Financial Services2859031.6%
Capital Goods37010528.4%
Automobiles & Components531528.3%
Household & Personal Products491122.4%
Consumer Durables & Apparel1252620.8%
Utilities821619.5%
Food, Beverage & Tobacco941617.0%
Energy2313515.2%
Materials1662313.9%
Pharmaceuticals, Biotechnology & Life Sciences6909313.5%
Banks337308.9%
Equity Real Estate Investment Trusts (REITs)100.0%

Industry composition

Which industries account for adopters—and for disclosure volume?

Composition of the adopter sample. The left bars show each industry’s share of all AI-adopting firms; the right bars show its share of AI adoption disclosures. Unlike the heatmap above, these are shares of the total adopter/disclosure sample—not within-industry adoption rates.

Industry composition of AI-adopting firms and AI adoption disclosures.
Industry-composition counts
AI-adopting firms and AI adoption disclosures by industry.
Industry groupAdopter firmsFirm shareAdoption disclosuresDisclosure share
Software & Services21416.2%417133.2%
Health Care Equipment & Services13310.0%9998.0%
Semiconductors & Semiconductor Equipment443.3%9787.8%
Technology Hardware & Equipment916.9%9727.7%
Financial Services926.9%8306.6%
Commercial & Professional Services785.9%8196.5%
Media & Entertainment926.9%7836.2%
Capital Goods1057.9%5044.0%
Pharmaceuticals, Biotechnology & Life Sciences916.9%4763.8%
Consumer Discretionary Distribution & Retail614.6%3692.9%
Consumer Services584.4%3062.4%
Automobiles & Components131.0%1781.4%
Transportation292.2%1761.4%
Insurance372.8%1631.3%
Energy332.5%1571.3%
Banks302.3%1561.2%
Consumer Durables & Apparel251.9%990.8%
Consumer Staples Distribution & Retail110.8%720.6%
Materials231.7%690.5%
Telecommunication Services100.8%650.5%
Real Estate Management & Development120.9%620.5%
Food, Beverage & Tobacco161.2%550.4%
Utilities161.2%460.4%
Household & Personal Products110.8%420.3%

03 / Integration depth

From initial use
to advanced integration.

The underlying classification distinguishes initial, intermediate, and advanced integration among substantive AI disclosures.

AI disclosures by initial, intermediate, and advanced integration from 2011 to 2024.
22.1%Initial · 2024
38.7%Intermediate · 2024
39.1%Advanced · 2024

Across 2011–2024, the shares are 24.0% initial, 44.3% intermediate, and 31.7% advanced.

Integration shares by year
Integration levels among substantive AI disclosures.
YearInitialIntermediateAdvanced
201116.7%33.3%50.0%
201212.5%29.2%58.3%
201327.5%37.5%35.0%
201420.0%36.0%44.0%
201518.9%50.0%31.1%
201619.1%47.2%33.8%
201728.8%45.8%25.5%
201829.1%48.9%21.9%
201928.2%49.2%22.6%
202024.0%51.3%24.6%
202122.4%48.0%29.6%
202220.5%46.7%32.8%
202324.5%39.2%36.3%
202422.1%38.7%39.1%

04 / Applications

What firms disclose using AI for.

A single AI disclosure can describe more than one application. The figure shows how the disclosed application mix changes over time, including the sharp rise of generative and interactive applications after 2022.

Application mix of AI-related corporate disclosures over time.
Analytic & predictive 83%Generative & creative 40%Interactive 32%Autonomous systems 9%

Shares are among AI-related disclosures; categories can overlap and need not sum to 100%.

Latest-year application counts
Latest-year application classifications. Categories are nonexclusive.
Application typeEventsDenominatorShare
Analytic & predictive2902349683.0%
Generative & creative1397349640.0%
Interactive1133349632.4%
Autonomous systems31234968.9%

05 / Named providers across industries

A disclosed view of the AI vendor ecosystem.

Providers named in AI adoption disclosures, connected to the sectors of the disclosing firms. Ribbon widths count distinct firm–provider pairs over 2011–2024.

Flow diagram linking named AI providers to the industry sectors of disclosing firms

Six leading named providers are shown. A firm can name more than one provider; ribbon widths are counts, not market shares, contract values, or verified vendor relationships.

Scope, aggregation & accessible counts

Each firm–provider pair is counted once over the full period. Related sectors are grouped for readability, and only larger links are displayed.

Displayed links only; a firm can contribute a pair to more than one provider.
ProviderSectorFirm–provider pairs
OpenAICommunication services22
OpenAIConsumer sectors16
OpenAIFinancials14
OpenAIHealth care5
OpenAIIndustrials13
OpenAIInformation technology50
OpenAIUtilities & real estate5
MicrosoftCommunication services8
MicrosoftConsumer sectors10
MicrosoftFinancials8
MicrosoftHealth care9
MicrosoftIndustrials15
MicrosoftInformation technology43
GoogleCommunication services12
GoogleConsumer sectors18
GoogleFinancials5
GoogleHealth care11
GoogleIndustrials9
GoogleInformation technology32
AWSCommunication services6
AWSFinancials5
AWSIndustrials5
AWSInformation technology29
NVIDIACommunication services5
NVIDIAInformation technology26
IBMHealth care6
IBMInformation technology13

Built from disclosures.
Designed for economic research.

The public page intentionally stops at descriptive measurement. It does not report valuation estimates, causal effects, portfolio returns, labor-market treatment effects, or other paper-specific results.

Research inquiries