Smaller Startup Teams Are Not Proof of an AI Cut
LinkedIn, on 27 October 2026, says 2024 founders employed 26 percent fewer people after one year than 2021 founders. That is not proof AI caused it.
Christian Kosiek•Published •13 min read

A smaller payroll can be a real observation and still be the wrong explanation. On 27 October 2026 LinkedIn published "LinkedIn data shows where opportunity is emerging as work changes." The page says founders who launched startups in 2024 employed 26% fewer people after their first year than founders who launched startups in 2021. The same page says the data do not show that technology caused smaller teams. A reader who turns that gap into "AI already replaced a quarter of the jobs startups would have created" is writing a story the release does not support. Rounding 26% to a quarter does not change what was measured.1
What the 26% figure measures
The comparison is employee counts after the first year, for two cohorts of founders on LinkedIn. It is not a measured 26% fall in total startup employment. It is not a finding that 26% of workers were laid off. It is not an estimate of how many jobs an AI tool removed. The public page does not say whether the 26% is a mean or a median, how many companies sit behind it, whether the founders themselves are in the count, whether contractors are in it, or whether industry and country were held constant.1
LinkedIn Pulse, the destination named on the page, draws on 1.4 billion members, 71 million companies and 42,000 skills. That is the size of the network, not a census of every registered business. Members and companies with a visible profile are not the same population as every firm that filed papers in 2021 or 2024. The page does not say how inactive startups were handled.1
| What the page supports | What it does not support |
|---|---|
| Founders who launched in 2024 employed 26% fewer people after the first year than founders who launched in 2021 | That AI caused the difference, or that every 2024 firm is a quarter smaller than a 2021 peer |
| The page says the data do not show that technology caused smaller teams | A headcount target for a new company |
| From July through September 2026, founder ranked among the 10 fastest-growing roles in the United States, the United Kingdom, Germany, France and India | A definition of "fastest-growing," or a count of companies that survived |
A firm that had ten people after year one, and a later firm that had about seven, would be a 26% gap if those were the only two numbers. That pair is arithmetic on an invented headcount. It is not a company in the release. A tighter funding market, a different product, more contractor work, or a choice to stay profitable can move a first-year payroll without a model writing the code. The page itself leaves the cause open.1
Founder titles and AI skills sit beside the gap
From July through September 2026, founder ranked among the 10 fastest-growing roles in the United States, the United Kingdom, Germany, France and India. The page does not define the growth rate: new titles, hiring into the role, or something else. A rising founder title is not, by itself, a count of companies that hired employees.1
Globally, founders are 12 times more likely than the average LinkedIn member to have an AI skill, and four in 10 professionals say AI has made them more likely to start a business. The page does not name the survey behind "four in 10," the sample, or the date of the question. These lines describe role growth, a skill on a profile, and a stated intention. They are not a verified count of businesses created because of AI, and they do not explain the 26% gap. The page's own question, "could new tools help people build businesses differently from the start?", is a question. It is not a result.1
Two nearby sentences use different verbs, and this article does not merge them. Only 1.2% of LinkedIn members list an AI skill. Founders are described as more likely to have an AI skill. Multiplying 1.2% by 12 would invent a founder share the page does not print.1
Members who list AI literacy skills are 78% more likely to get hired and 11% more likely to receive a promotion than members who do not. The page does not publish controls for occupation, education, employer or country. In the United States, workers with AI skills move into occupations that pay an average of $28,000 more than the occupations entered by workers without them. That is a comparison of occupations entered. It is not a raise for the same person in the same seat, and the sentence says "AI skills," not the "AI literacy" list used for the hire and promotion lines.1
The page also disagrees with itself on one posting statistic, so this article does not pick a single window. The key-insights list says the share of job postings requiring AI literacy skills has jumped roughly 65% in the last 12 months. The body says that share has jumped roughly 65% in each of the past two years and is now more than 10 times the 2023 rate. Those are not the same sentence. Neither is a startup headcount.1
An estimate of 2 million jobs is a different question
LinkedIn estimates that AI has helped create roughly 2 million new jobs. The page does not say over what period, in which countries, or by what method. The figure is not a net balance against the 26% staffing gap. One is an estimate of jobs the company attributes, in part, to AI. The other is a comparison of first-year startup payrolls, with an explicit statement that technology is not shown as the cause. Putting them on one line would invent a national surplus or deficit the release does not calculate.1
A new job somewhere is also not a placement for the person whose task changed. See why a created job is not a transfer plan. What happens after a job ends, including a longer search, is a separate measurement. See when the next job takes longer to find.
From July through September 2026, AI Engineer ranked among the 10 fastest-growing occupations in seven economies: Canada, Germany, India, France, Brazil, the United Kingdom and the United States. Entry-level hiring for AI Engineers grew 68% year over year in the United States and 55% in India. That is one occupation, on LinkedIn. It is not the startup comparison, and it is not a rate for every junior role.1
Access inside those AI roles is uneven on the same page. Women account for only about one quarter of new AI hires. In 2025, more than 90% of AI roles went to workers with at least a bachelor's degree. As of September 2026, just five countries or territories ranked in the top 10 for both AI talent and employer demand for that talent. The page does not name the five. A different dataset, on US occupations and gender, is already discussed in where the AI jobs gender gap is widest without a degree. The percentages there are not these LinkedIn figures, and they should not be averaged.1
The counts a payroll gap still needs
The list below is an editorial checklist. LinkedIn did not publish it as a method, and this article does not treat the missing lines as hidden findings.
A useful reading of "fewer people per young company" needs at least four counts the page does not give. How many firms were created in each cohort? What share survived the first year? What revenue or service volume did each employee support? What work moved off payroll to freelancers, agencies or software? A company with fewer employees can still buy more outside help. A rise in founder titles can sit next to fewer stable employee posts. The release cannot separate "more firms, each a bit smaller" from "the same firms, permanently leaner," because it does not print the number of firms.
Work that leaves the payroll can show up as a different purchase. A marketplace posting is still not a job count. See how freelance briefs are being rebundled. That study is Upwork and OpenAI, not this LinkedIn comparison.
Customer-service headlines have their own limit: a smaller published headcount cut is not the same as a task that disappeared. See why customer-service layoff claims need the actual headcount.
A 60-day check is local. It is not a national result.
The steps below are an editorial suggestion. LinkedIn did not test them. They are not a research result, and they do not establish an economy-wide effect.
The operational question is not whether a five-person team should become a three-person team because a platform cohort was smaller. It is which tasks got shorter, which got a new check, and which moved to a contractor. For a small business, a model can draft a sales proposal faster and still leave someone to verify the promise, protect customer data and handle the replies the draft created. Faster production can add coordination. It does not only delete hours.
A hypothetical nonprofit that prepares plain-language advice sheets with a model should, before it removes an administrative role, count cases handled, corrections, review time and the questions that still need a person. If demand rises, the organization may need more caseworkers while the drafting gets faster. That scene is an illustration. LinkedIn did not observe it.
A modest test is to pick one workflow, record hours and quality, name a reviewer, and measure the whole path including rework. Then decide. A pilot inside one organization cannot stand in for the 26% figure, and the 26% figure cannot stand in for the pilot.
Smaller founding teams can be comfortable for owners and thin for people who need a first job. Junior employees often learn on research, drafts, records and routine questions. If those tasks are automated and nothing replaces the practice, the organization saves time now and has less internal skill later. One response is to keep the check: a new hire can use a model for a market scan and still has to verify sources, explain the trade-off and defend a recommendation. Being able to attempt a task is not mastery. See who teaches judgment after the first draft, why a wider task is not mastery, and entry-level ads that already ask for senior judgment. Who had paid time to practise is a related problem. See why a training gap becomes a career gap.
For a European organization, a smaller first-year payroll is not a finding that a system is high-risk. This is not legal advice. Article 4, in the Commission service-desk text of the consolidated AI Act as at 27 July 2026, says providers and deployers shall take measures to support the AI literacy of their staff and of other persons dealing with the operation and use of systems on their behalf. The same paragraph says this does not require a guaranteed level for any individual. The service desk says the summary is not legally binding. A literacy measure is not an instruction to employ fewer people.2
Annex III is headed "Employment, workers' management and access to self-employment." Point 4(a) lists systems intended for recruitment or selection, including targeted job advertisements, filtering applications and evaluating candidates. Point 4(b) lists systems intended to decide terms of a work relationship, promotion or termination, to allocate tasks based on individual behaviour or personal traits, or to monitor and evaluate performance and behaviour. A first-year headcount gap is not on that list. A tool intended to score candidates, or to end contracts, is a different question. Classification depends on the intended purpose. The service desk says Annex III rules apply from 2 December 2027. This article does not decide whether any hiring product is in scope.3
How hilighter can help
hilighter does not set a startup's headcount, explain LinkedIn's 26% gap, or decide whether a hiring tool is high-risk under the AI Act. It can make the skills inside a role visible when a small team is deciding what still needs a person. Teams can map skills by role, compare people with those requirements in skills and talent insights, record a check or a correction in skills evidence and recognition, and set personalized growth paths. A manager's review of that evidence sits in manager assessments. The AI job impact view is indicative. It is not a LinkedIn forecast, and it is not a prediction that a role will disappear. The Free plan covers up to three job roles. The number of employees is not capped.
Sources
- LinkedIn Corporate Communications, LinkedIn data shows where opportunity is emerging as work changes, 27 October 2026. Founders who launched startups in 2024 employed 26% fewer people after their first year than founders who launched in 2021. The page says the data do not show that technology caused smaller teams. The FAQ says LinkedIn data do not establish that AI is responsible for smaller startup teams. Founder was among the 10 fastest-growing roles in the United States, the United Kingdom, Germany, France and India from July through September 2026. Founders are 12 times more likely than the average member to have an AI skill. Four in 10 professionals say AI has made them more likely to start a business; the page does not name the sample. LinkedIn estimates that AI has helped create roughly 2 million new jobs, without a published method or period. AI Engineer was among the 10 fastest-growing occupations in Canada, Germany, India, France, Brazil, the United Kingdom and the United States in that same July-September window. Entry-level hiring for AI Engineers grew 68% year over year in the United States and 55% in India. Members who list AI literacy skills are 78% more likely to get hired and 11% more likely to receive a promotion. In the United States, workers with AI skills move into occupations paying an average of $28,000 more than occupations entered by workers without them. Only 1.2% of members list an AI skill. The key-insights list says AI-literacy postings jumped roughly 65% in the last 12 months; the body says roughly 65% in each of the past two years and more than 10 times the 2023 rate. Women account for about one quarter of new AI hires. In 2025, more than 90% of AI roles went to workers with at least a bachelor's degree. Five countries or territories, not named, ranked in the top 10 for both AI talent and demand as of September 2026. Pulse draws on 1.4 billion members, 71 million companies and 42,000 skills. Network data, not a census of registered firms, and not a causal study.
- European Commission, AI Act Service Desk, Article 4: AI literacy. Wording reflects the EUR-Lex consolidated version as at 27 July 2026, including Digital Omnibus amendments. The summary is not legally binding. Paragraph 1 does not require a guaranteed level of literacy for any individual.
- European Commission, AI Act Service Desk, Annex III, heading and points 4(a) and 4(b). When does enforcement start?: Annex III rules from 2 December 2027. Not a decision about any hiring product.
Sources last reviewed: 16 November 2026. The 26% line is not treated as a layoff rate, a causal effect of AI, or a headcount target. The key-insights "65% in the last 12 months" line is not merged with the body's "65% in each of the past two years." The 2 million estimate is not netted against the 26%. The $28,000 figure is a gap between occupations entered in the United States, not a raise. "Have an AI skill" is not multiplied by the 1.2% who list an AI skill. Averages such as 4.6 and 3.4 employees, and a later-year headcount comparison, appear in other public posts and are not in this 27 October page, so they are not used. PwC's Global Workforce Hopes and Fears Survey 2026 (29 September 2026; 49,364 workers in 48 countries and regions, May and June 2026) was listed in the inbox draft and was opened. It describes employees, not this startup comparison, and its statistics are not used. EUR-Lex was not opened separately. EU wording follows the Commission service desk and is not legal advice. Neither an AI-detector score nor the absence of Google penalties can be guaranteed.