10 metrics to measure AI’s impact on society · en · Blog
Published on August 4, 2026
Introduction
Artificial intelligence has, in just a few years, moved from a technological promise to core infrastructure for firms, knowledge workers, and public institutions. Yet discussions about “AI’s impact on society” often remain abstract. This article proposes a simple framework: ten concrete metrics to measure that impact, drawing on recent work from the OECD, the IMF, central banks, and major adoption reports.
1. AI adoption rate within organizations
The first metric is straightforward: what share of organizations already use AI in at least one business function (marketing, support, finance, production, etc.)? Major international surveys show that in 2025, about 88% of organizations reported using AI somewhere in their operations, up from roughly 78% in 2024. This rapid growth indicates that we have moved beyond isolated proofs of concept into broad diffusion, even though maturity levels remain highly uneven across sectors.
2. Individual adoption by knowledge workers
Beyond organizations, it is essential to measure actual usage by individuals. Recent reports on knowledge work indicate that about 75% of knowledge workers say they use at least one AI tool (writing assistant, coding copilot, summarization tool) in their daily work. This rate is particularly high in tech, marketing, and customer support roles, and rose sharply after large language models became mainstream in 2023–2024.
3. Task-level productivity gains
To move beyond general discourse, controlled studies measure productivity gains on specific tasks: customer response drafting, report generation, code generation, translation, and more. On these types of tasks, observed gains typically range from 15% to over 50% depending on context, with simultaneous improvements in output quality (fewer errors, more standardization). This metric helps identify AI “sweet spots” within a given occupation: tasks where AI acts as a clear multiplier rather than a gadget.
4. Aggregate workforce productivity
Economies want to know whether these local gains translate into macro-level improvements. A US survey found that generative AI users reported saving an average of 5.4% of their work hours in the previous week, translating into an estimated 1.1% increase in productivity for the entire workforce. Other reviews conclude that “classic” AI (recommendation, vision, optimization) tends to raise firm productivity by 0% to 11%, a range still difficult to detect in aggregate macro figures.
5. Impact on aggregate employment
One of the major societal debates concerns potential job destruction. Empirical studies available so far show no clear signal of massive job destruction at the aggregate level. One panel of US firms estimates that in 2026, AI will reduce total employment by less than 0.4% relative to a no‑AI scenario. The observed reality is more about task and role transformation than abrupt job disappearance.
6. Job exposure to AI
Rather than only counting jobs destroyed or created, another metric analyzes the share of jobs “exposed” to AI: those where a significant portion of tasks could be automated or heavily assisted. The IMF estimates that about 40% of jobs worldwide are exposed to AI, and nearly 60% in advanced economies where non‑routine cognitive tasks are more common. This metric offers insight into reskilling challenges and the sectors where public policy needs to anticipate change.
7. Polarization and job structure
AI does not affect all jobs equally. Several studies suggest increasing polarization: pressure on middle‑skill jobs, while some highly skilled roles and some low‑skill service jobs may be augmented rather than replaced. An informative metric is to track trends in employment and wages by skill quintile and AI exposure level, to see whether AI widens or narrows gaps within a country or sector.
8. Wage premium for AI skills
Another way to measure impact on individuals is to examine the wage premium for workers with AI skills. A global AI jobs barometer indicates that in 2025, workers with AI skills earn on average a 56% wage premium over similar roles without those skills, up from 25% a year earlier. This metric shows that AI is becoming a strong differentiator in the labor market, with benefits distributed unevenly.
9. Performance and revenue of AI‑adopting firms
From a business perspective, attention focuses on impact on innovation, customer satisfaction, revenue, and margins. Surveys show that over 60% of companies deploying AI solutions report improved innovation capacity and market differentiation. Among generative AI early adopters in certain B2C sectors, average revenue increases in the 10–20% range have been observed, with one study citing an average uplift of 15.2%.
10. Well‑being and quality‑of‑work indicators
Finally, societal impact is not limited to GDP or productivity. Recent research highlights that AI can simultaneously boost productivity while creating tensions around job quality, economic security, and mental health. An interesting metric is to track well‑being indicators within AI‑exposed organizations: perceived job security, mental workload, job satisfaction, and perceived ability to train and grow. Over time, these metrics may become as important as strictly economic ones for assessing whether AI genuinely improves workers’ lives.
How to use these metrics in an organization
For a company or SaaS platform, these ten metrics can form the basis of an internal AI impact dashboard.
- Guiding rollouts: measuring time saved per task type (support, appointment booking, email handling) and associated revenue gains helps prioritize the most promising use cases.
- Tracking effects on teams: monitoring changes in role structures, internal wage premiums for AI‑skilled profiles, and well‑being indicators avoids managing purely by productivity numbers.
- Engaging partners and regulators: having clear metrics on job exposure, reskilling, and quality of work facilitates constructive dialogue about AI, away from purely alarmist or overly optimistic narratives.
Conclusion
AI’s impact on society is profound, but neither monolithic nor instantaneous. The metrics presented here show a nuanced reality: significant productivity gains at the task level, still modest macro effects, transformation of job structures rather than mass destruction, strong premiums for AI‑skilled profiles, and real well‑being challenges at work. Adopting a small set of robust metrics helps move beyond ideological debate toward concretely steering AI in the service of organizations and people.