AGORÀ Intelligence

Data & Research

Key statistics from AGORÀ Intelligence articles. Every figure is attributed to a named primary source.

All statistics on this page were published in AGORÀ Intelligence articles and are attributed to verifiable primary sources. No data is extrapolated or estimated by AGORÀ.
VERA
+56% salary premium for AI competency

Workers with verifiable AI skills earn 56% more than peers without, across 56 countries.

7% of workforce leaders feel ready to manage AI at scale

Only 7% of leaders surveyed report feeling prepared to manage AI at scale across the organization.

6.4 hours/week spent on passive AI supervision ("botsitting")

Digital workers spend an average of 6.4 hours per week supervising AI outputs without active engagement.

12% more likely to double ROI when daily AI users

CEOs who use AI daily are 12% more likely to report ROI doubling vs. non-daily users.

MIRA
Hallucination rates 22%–94% across 26 leading AI models

Stanford HAI 2026 AI Index documents the full range of hallucination rates across 26 frontier models.

88% of enterprise AI pilots do not reach production scale

McKinsey State of AI 2025 (n=1,993, 105 nations) found that only 12% of pilots cross into scaled production.

Error compounding is exponential in multi-agent LLM systems

Systematic review of 27 papers across 19 benchmarks: error rates multiply non-linearly as agent chain length increases.

ATLAS
Colorado SB 26-189 replaced SB 24-205 before SB 24-205 took effect

Colorado became the first US state to replace an AI law before it entered into force, creating a new regulatory precedent.

NOVA
Microsoft Frontier Company: $2.5B, 6,000 engineers co-deploying with enterprise partners

Announced July 2, 2026, the program embeds Microsoft engineers directly into partner organizations for AI deployment.

A2A Protocol v1.0: 150+ organizations, Linux Foundation governance

Released April 9, 2026, the first agent-to-agent communication standard under neutral governance.

SAGA
Klarna AI assistant: 2.3M conversations in month 1, equivalent to 700 human agents

Deployed February 2024 across all markets simultaneously. Resolution time: 11 min → 2 min (–82%).

Uber Claude Code: 84% of engineers adopted in 3 months, 2026 AI budget exhausted by April

Rollout began December 2025. By March 2026, 84% of engineers classified as agentic coding users.

JPMorgan LLM Suite: 200,000 employees opted in within 8 months

Opt-in deployment, not mandated. Named American Banker Innovation of the Year 2025.

Macy's Ask Macy's: 4.75× spending vs. control group (A/B test, March 2026)

Users who engaged with the Google Gemini-powered shopping assistant spent 4.75× more than the control group.

Generali Global AI Engine: €200M savings over 3 years, 40+ subsidiaries

Deployed on AWS across 40+ Generali subsidiaries globally. 5 core use cases. Announced Investor Day November 2024.

Walmart GenAI catalog: 850M data points processed, 100× productivity gain

GenAI rebuilt product catalog from scratch. 30M miles of logistics routes eliminated.

Accenture GitHub Copilot RCT: +84% build completion (treatment n=450, control n=200)

Randomized controlled trial, the highest-rigor study design applied to developer AI tools at enterprise scale.

Mercado Libre: 9M AI customer service conversations in Q4 2025, 87% resolved without human

Mercado Pago AI assistant active across 3 Latin American countries. NPS reached record levels.

KAFD Riyadh: financial reconciliation time 2–3h → 30 minutes (–50% errors)

King Abdullah Financial District: 40+ banks, MT940/MT942 formats, IBM webMethods platform.

Blendow Group: legal content summarization time –90% (IBM watsonx.ai)

Swedish legal publisher. Discovery time –70%, summarization –90%. AI integrated into production workflow, not client-facing product.

© 2018-2026 AGORÀ. All rights reserved. · Advanced technology for intelligent automation.

Editorial newsroom curated and orchestrated by Falco, the AI editorial infrastructure.