Copilot users finished tasks 26%–73% faster
A Microsoft meta-review of studies covering Microsoft Copilot and GitHub Copilot found users completed tasks in 26% to 73% less time than workers without AI access.
The money walked out. Generative AI didn’t get the memo.
Chapter 4 of the AI Index 2024 tracks AI through the economy in 2023, and almost every headline number points down: corporate investment fell for a second year, private investment fell with it, and AI’s share of American job postings shrank. Cutting straight across that trend is one line going vertical — generative AI funding — and a first real body of evidence that AI raises worker output. The numbers:
25.2private investment in generative AI in 2023 (US$ billions, ~9× 2022)
67.2US private AI investment (US$ billions) — 8.7× China’s 7.8
189.2global corporate AI investment (US$ billions, down ~20%)
1,812newly funded AI companies worldwide in 2023, up 40.6%
55% of organizations using AI in at least one function (2022: 50%)
394Fortune 500 earnings calls that mentioned AI (2022: 266)
4.3 — Two years of decline, and one line going straight up
Global corporate AI investment fell to $189.2 billion in 2023, roughly 20% below 2022, with mergers and acquisitions taking the worst of it at -31.2%. Private investment slipped 7.2% to $96.0 billion, its second consecutive annual decline. Generative AI went the other way entirely.
Generative AI takes more than a quarter of everything
Generative AI attracted $25.2 billion in private investment in 2023 — nearly nine times the 2022 total and about 30 times the 2019 figure. That one sector accounted for more than a quarter of all AI-related private investment for the year. The money went to the model builders: OpenAI, Anthropic, Hugging Face and Inflection all reported substantial rounds. The number of newly funded generative AI companies rose to 99, up from 56 in 2022 and 31 in 2019.
More companies, smaller cheques
- The count of newly funded AI companies worldwide jumped to 1,812, a 40.6% increase over the previous year — but the average private investment event barely moved, from $31.3 million in 2022 to $32.4 million in 2023.
- Deal counts fell in almost every size band. The only categories that grew were the biggest: rounds over $1 billion rose from 7 to 9, and rounds between $500 million and $1 billion from 6 to 7. Below that, $100–500 million fell from 187 to 120, $50–100 million from 260 to 182, and under $50 million from 2,840 to 2,641. Total events dropped from 3,994 to 3,639.
- By focus area, the leaders in 2023 were AI infrastructure, research and governance ($18.3 billion), NLP and customer support ($8.1 billion), and data management and processing ($5.5 billion). Infrastructure and data management were among the few areas that grew at all.
- Medical and healthcare and NLP/customer support both peaked in 2021 and have declined since — still substantial in absolute terms, but no longer where the new money goes.
- Zoom out and the picture reverses: over the past decade, AI-related corporate investment has grown thirteenfold.
The United States pulls further ahead
US private AI investment reached $67.2 billion in 2023, roughly 8.7 times the next highest country, China ($7.8 billion), and 17.8 times the United Kingdom ($3.8 billion). Germany ($1.91 billion) and Sweden ($1.89 billion) came next. The gap is widening rather than closing: since 2022, private AI investment fell 44.2% in China and 14.1% in the European Union plus the United Kingdom, while the United States rose 22.1%.
- Aggregated since 2013 the ranking is identical: the United States $335.2 billion, China $103.7 billion, the United Kingdom $22.3 billion, then Israel ($12.8 billion), Canada ($10.6 billion) and Germany ($10.4 billion).
- Company formation follows the same map. In 2023 the United States produced 897 newly funded AI companies, against China’s 122 and the United Kingdom’s 104.
- Over the decade the US total (5,509) is around 3.8 times China’s (1,446) and 7.6 times the United Kingdom’s (727).
4.2 — Fewer AI job postings, and one skill that appeared out of nowhere
AI-related roles fell from 2.0% of all American job postings in 2022 to 1.6% in 2023. Lightcast reads that as a hiring pullback at a handful of large employers rather than as cooling interest in AI — and underneath the total, one skill cluster grew more than tenfold.
Who is posting, and where
- By share of all postings, the United States led in 2023 at 1.62%, followed by Spain (1.35%) and Sweden (1.31%). Most countries fell year over year, even though over five years the count of AI postings has risen in many of them.
- Lightcast attributes the US drop to top AI employers — Amazon, Deloitte, Capital One, Randstad and Elevance Health among them — scaling back posting counts and shifting their occupational mix. Amazon in 2023 advertised a higher share of delivery driver, packager and mailroom roles, and a lower share of software developer and data scientist roles.
- By skill cluster, machine learning stayed the most in-demand at 0.7% of US postings, ahead of artificial intelligence (0.5%) and natural language processing (0.2%). Every cluster Lightcast tracks lost market share except generative AI, which grew by more than a factor of 10.
- In raw counts, 15,410 US postings named generative AI as a desired skill, 4,669 mentioned large language modeling and 2,841 named ChatGPT. Within generative AI postings the most cited skill was generative AI itself (60.0%), then large language modeling (18.2%) and ChatGPT (11.1%).
- Nearly every sector’s AI posting share fell. Public administration was the exception, up 88.8% to 1.49%, with educational services also rising 6.0%. The leading sectors were still information (4.63%), professional, scientific and technical services (3.33%) and finance and insurance (2.94%).
- California posted 70,630 AI jobs — 15.3% of the US total — ahead of Texas (36,413) and Virginia (24,417). As a share of a state’s own postings, the District of Columbia led at 2.7%, then Delaware (2.4%) and Maryland (2.1%).
Hiring, skills, and where the talent sits
- Relative AI hiring grew fastest in Hong Kong (28.8%), Singapore (18.9%) and Luxembourg (18.9%) — meaning AI hiring rose that much faster than overall hiring in those markets.
- Across 2015–2023, the highest relative AI skill penetration was in India (2.8), the United States (2.2) and Germany (1.9). A rate of 2.2 means US LinkedIn members are 2.2 times as likely as the global average to list AI skills across the same set of occupations.
- AI talent concentration in 2023 was highest in Israel (1.1%), Singapore (0.9%) and South Korea (0.8%).
- Developer pay is a rough proxy for AI-adjacent compensation. In Stack Overflow’s survey of more than 90,000 developers, a cloud infrastructure engineer earned $105,000 globally and $185,000 in the United States. Senior executives topped the table in both, followed by engineering managers — and every surveyed role paid significantly more in the US than elsewhere.
4.4 — Adoption crept up, the savings showed up, and every boardroom started saying it out loud
55% of organizations surveyed by McKinsey used AI in at least one business unit or function in 2023, up from 50% in 2022 and 20% in 2017. Generative AI, asked about for the first time, was already at 33%.
Every surveyed region reported higher AI adoption than in 2022. Europe grew fastest at 9 percentage points, Greater China rose 7, and North America stayed the overall leader. On generative AI specifically, North America leads at 40% against 33% worldwide, with Europe and Greater China both at 31% and Asia-Pacific at 30%.
What it is actually being used for
- By function, AI shows up most in product and/or service development (24% of respondents), marketing and sales (23%) and service operations (23%). Generative AI specifically clusters in marketing and sales (14%), product and service development (13%) and service operations (10%).
- The largest year-over-year gains were in marketing and sales (+18 percentage points), product and service development (+14) and service operations (+4). Strategy and corporate finance went the other way at -12 points, risk at -9 and human resources at -2.
- Across all industries, the most embedded AI capabilities were natural language text understanding (30%), robotic process automation (30%) and virtual agents (30%). Inside financial services, robotic process automation reached 46%.
- The single heaviest industry-and-function combination was product and/or service development in tech, media and telecom at 44%, followed by service operations (36%) and marketing and sales (36%) in the same sector.
- The most common generative AI use cases are narrow: drafting text documents (9%), personalized marketing (8%), summarizing text documents (8%) and creating images or video (8%). Most of the leading cases sit inside marketing and sales.
The returns showed up in the survey
Averaged across all activities, 42% of respondents reported cost decreases from AI and 59% reported revenue increases — a 10-point rise in cost decreases and a 4-point fall in revenue increases against the previous year. Cost savings were reported most often in manufacturing (55%), service operations (54%) and risk (44%); revenue gains most often in manufacturing (66%), marketing and sales (65%) and risk (64%). The magnitudes stayed modest: most reported cost decreases were under 10%, and the most common revenue increase was 5% or less.
Developers were already living in it
Stack Overflow surveyed more than 90,000 developers in May 2023. GitHub Copilot was the most used AI developer tool at 56.0%, far ahead of Tabnine (11.7%) and AWS CodeWhisperer (4.9%); ChatGPT was the most used AI search tool at 83.3%, ahead of Bing AI (18.8%) and WolframAlpha (11.2%). 82.6% of developers regularly used AI to write code, 48.9% for debugging and getting help, and 34.4% for documentation. Testing was the visible gap: only 23.9% used AI for it, while 55.2% said they wanted to. Asked what they got out of the tools, developers named increased productivity (32.8%), faster learning (25.2%) and greater efficiency (25.0%).
And the boardrooms started saying it out loud
Quid analyzed every 2023 Fortune 500 earnings call. AI was mentioned in 394 of them — nearly 80% of the Fortune 500 — up from 266 in 2022 and roughly double the 2018 level. The dominant theme was generative AI, appearing in 19.7% of all earnings calls, against 0.31% a year earlier. Next came investments in AI, expansion of AI capabilities and AI growth initiatives (15.2%), then company and brand AIs (7.6%).
The year the productivity question finally got answers
For years the evidence on AI and worker output was thin, simply because so few people used AI at work. In 2023 several rigorous studies landed at once — and they largely agree, including on who benefits most and where the trap is.
Consultants with GPT-4 did better work, faster
A Harvard Business School study found consultants with GPT-4 access beat a control group by 12.2% on productivity, 25.1% on speed and 40.0% on quality.
Support agents handled 14.2% more calls an hour
National Bureau of Economic Research work found call-center agents using AI handled 14.2% more calls per hour than agents without it.
The gains land hardest on the least experienced
In the same consulting experiment, bottom-half-skilled participants improved 43.0% with AI while top-half participants improved 16.5%. AI narrowed the distance between them.
Trusting the good AI too much made recruiters worse
Recruiters given any AI assistance gained 0.6 accuracy points — but those given a high-performing AI scored 1.08 points below those given an AI known to make errors.
Six questions the numbers answer
What Chapter 4 can and cannot tell you about AI and the economy in 2023.
Did AI investment really fall in 2023?
Yes, on two different lines — and it depends which one you read. Total global corporate AI investment fell to $189.2 billion, about 20% below 2022, with mergers and acquisitions down 31.2%. Private investment fell 7.2% to $96.0 billion, a second consecutive annual decline, though a shallower one than the drop from 2021 to 2022. But the number of newly funded AI companies rose 40.6% to 1,812, and generative AI alone took in $25.2 billion, more than a quarter of all AI-related private investment. Over the decade, AI-related corporate investment has still grown thirteenfold.
Why did AI job postings drop?
Not because employers stopped wanting AI. Lightcast attributes the fall from 2.0% to 1.6% of US postings to top AI employers — Amazon, Deloitte, Capital One, Randstad and Elevance Health among them — cutting their overall posting counts, and to a shift in what they posted. Amazon in 2023 advertised a higher share of sales delivery driver, packager and mailroom roles and a lower share of software developer and data scientist roles. Underneath the aggregate, generative AI was the only skill cluster Lightcast tracks that gained market share, growing by more than a factor of 10, with 15,410 US postings naming it explicitly. Nearly every sector fell; public administration was the exception, up 88.8%.
Is AI actually making workers more productive?
2023 was the first year with enough rigorous studies to answer. Copilot users completed tasks in 26% to 73% less time. Consultants with GPT-4 gained 12.2% in productivity, 25.1% in speed and 40.0% in quality. Call-center agents using AI handled 14.2% more calls per hour, 2.97 hourly chats against 2.60. Law students with GPT-4 improved on both work quality and time efficiency across a range of legal tasks, contract drafting most of all — although the report notes that LLM hallucinations are reported as especially pervasive in legal work. The consistent finding across studies is that lower-skilled workers gain most: 43.0% against 16.5% in the consulting experiment. The consistent caveat is oversight.
What do the forecasts say?
McKinsey projected that generative AI could raise high-tech industry revenue by 4.8% to 9.3% — an extra $240 billion to $460 billion — with banking next at 2.8% to 4.7% ($200 billion to $340 billion), then pharmaceuticals and medical products, and education. Goldman Sachs projected global productivity growth of 1.0% to 1.5% over 10-year periods, with Hong Kong, Israel and Japan especially well-positioned. On headcount, executives expect the sharpest reductions in service operations (54% anticipate a decrease), supply chain management (45%) and human resources (41%). Asked about their own employee numbers, 30% of respondents expect little or no change, 25% expect a decrease of 3%–10%, and 12% expect an increase of 3% or more — while 38% expect to reskill more than a fifth of their workforce.
What are companies telling investors?
A great deal more than a year earlier. AI came up in 394 Fortune 500 earnings calls in 2023 — nearly 80% of the list — up from 266 in 2022 and roughly double the 2018 count. The dominant theme was generative AI, mentioned in 19.7% of all calls against 0.31% the year before. Next came investments in AI, expansion of AI capabilities and growth initiatives (15.2%), then company and brand AIs (7.6%). The survey data suggests the operational reality is more modest: 42% of organizations reported cost decreases and 59% revenue increases, most of them under 10% and 5% respectively.
Are robots part of this story?
Yes, though the robot data lags a year behind the rest of the chapter. In 2022, 553,000 industrial robots were installed worldwide, a 5.1% increase and more than triple the 2012 figure, while the operational stock grew to 3,904,000 from 3,479,000. China installed 290,300 of them — 52.4% of the global total, 5.8 times Japan’s 50,400 and 7.4 times the United States’ 39,500. The fastest growth was in Singapore (68%), Turkey (22%) and Mexico (13%); Canada (-24%), Taiwan (-21%) and Thailand (-18%) fell hardest. The mix is shifting too: collaborative robots, built to work next to people rather than instead of them, rose from 2.8% of new installations in 2017 to 9.9% in 2022, and service robot installations rose in every application category except medical robotics. By sector, electrical and electronics led with 157,000 installations and automotive followed with 136,000; by application, handling dominated at 266,000, 3.1 times welding and 4.4 times assembly.
The chapter in five lines
Headline findings from Chapter 4 · Economy.
Despite a decline in overall AI private investment last year, funding for generative AI surged to reach $25.2 billion. Major players including OpenAI, Anthropic, Hugging Face and Inflection reported substantial fundraising rounds.
In 2023, the United States saw AI investments reach $67.2 billion, nearly 8.7 times more than China, the next highest investor.
In 2022, AI-related positions made up 2.0% of all job postings in America, a figure that decreased to 1.6% in 2023.
Several studies assessed AI’s impact on labor, suggesting that AI enables workers to complete tasks more quickly and to improve the quality of their output. These studies also demonstrated AI’s potential to bridge the skill gap between low- and high-skilled workers.
In 2013, China’s installations accounted for 20.8% of the global total, a share that rose to 52.4% by 2022.
Read Chapter 4 in full
Chapter 4 (sections 4.1–4.5) — the 2023 timeline, jobs, investment, corporate activity and robot installations — with every figure and citation is free from Stanford HAI.
Open the AI Index 2024 →