Data study ·
AI in the workplace: top 10% generate 8.3 times more output
Original title, in French: “IA en entreprise : 8,3× plus de sorties pour le top 10%”
- Publication
- Brief IA (France)
- Author
- Tom Levy
- Published on
- 1 September 2026 · read in the page structured data
Summary
An OpenAI overview describes a shift from assistance to execution in how AI agents are used in business. According to OpenAI, the top 10% of user organisations now generate 8.3 times more output tokens per active user than the average, up from 2.6 times in January, a sign of a widening gap between pioneers and followers. Three cases illustrate this trend: at Basis, onboarding of new employees went from two hours to 30 minutes thanks to Codex; at Clay, per-account sub-agents save about an hour of overnight email triage for sales follow-up; at Exa Labs, Codex automates the transition from discovery to technical implementation via Slack and Notion. OpenAI also notes that junior employees send 13 more messages per week than executives six months after adoption.
Summary based on the full text of the article.
Read the article on Brief IA →Figures cited
-
8.3 x
Multiplier of output generated per active user, top 10% of companies vs average
User organisations, according to OpenAI
« Les organisations du top 10 % génèrent 8,3 fois plus de jetons de sortie par utilisateur actif que la moyenne, contre 2,6 fois en janvier, d’après OpenAI. »
Source: OpenAI
-
2.6 x
Same multiplier measured in January (change in the gap)
User organisations, January, according to OpenAI
« Les organisations du top 10 % génèrent 8,3 fois plus de jetons de sortie par utilisateur actif que la moyenne, contre 2,6 fois en janvier, d’après OpenAI. »
Source: OpenAI
-
30 minutes
Onboarding time for new hires at Basis with AI
Basis
« Chez Basis, l’intégration des nouveaux arrivants est passée à 30 minutes au lieu de deux heures. »
-
2 hours
Onboarding time for new hires at Basis before AI
Basis
« Chez Basis, l’intégration des nouveaux arrivants est passée à 30 minutes au lieu de deux heures. »
-
1 hour
Time saved per night on email triage at Clay thanks to sub-agents
Clay
« Selon l’entreprise, ce dispositif économise environ une heure de tri nocturne des e-mails et facilite le suivi des micro-actions sur des cycles de vente longs. »
Source: Clay
-
13 messages/week
Difference in messages sent per week between junior employees and executives, six months after AI adoption
According to OpenAI, six months after adoption
« Six mois après l’adoption, des salariés en début de carrière ont envoyé 13 messages de plus par semaine que des dirigeants, selon OpenAI, ce qui traduit une dynamique d’appropriation. »
Source: OpenAI
Classification
- Country
- Worldwide scope
- Company size
- All sizes
- Industries
- Software vendors, IT services companies, telecoms · All industries, or industry not specified
- Companies
- Exa Labs (deploys AI), Clay (deploys AI), Basis (deploys AI), OpenAI (study author)
- Keywords
- AI agents, enterprise adoption, productivity, process automation, human resources, sales
Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to Brief IA and its author. Entry translated from French by Claude Sonnet 5.5.