Every number AIMI uses — verified at its source, dated, and linked. This is the research behind the Index.
Last updated: 2026-09-20 · 38 verified sources
Every figure here was checked against the primary publication, or corroborated across at least two independent sources. Figures we could not verify are kept in a private working file and are never published — even when they would make a better headline.
The base is maintained continuously: an automated research radar scans McKinsey, BCG, Deloitte, Gartner, IDC, Stanford HAI, MIT, KPMG, the EU institutions and DACH-specific sources every few days, and every candidate is re-verified before it is added.
Sources are cited as published, in their original language. We report what the source says — including the caveats.
Figures and findings are translated; source titles, dates and samples are given as published.
AI use is now near-universal. Maturity is not.
80% of respondents say AI has improved their individual productivity — yet only 37% of organisations report any positive EBIT contribution (flat year-on-year), and AI high performers remain 6%.
Individual gains have not translated into enterprise value. High performers redesign workflows (three-quarters vs one-quarter of others) and are twice as likely to have defined processes to measure impact. "The limiting factor is increasingly the organization's ability to absorb change."
Source: McKinsey, "The State of AI in 2026: On the road to ROI" (Global Survey) · 25 Aug 2026 · Sample: 1,719 participants, 97 nations, fielded 4 May – 8 Jun 2026; GDP-weighted
44% report AI scaling across their enterprise (up from 38%); 54% at companies above $1B revenue versus one-third of smaller ones. About two in ten have reached the scaling phase with AI agents.
Scaling is real but concentrated in large enterprises — and it has not moved the share reporting financial impact.
Source: McKinsey, "The State of AI in 2026: On the road to ROI" (Global Survey) · 25 Aug 2026 · Sample: 1,719 participants, 97 nations
88% use AI in at least one function — only ~1% consider themselves fully mature; ~two-thirds have not scaled beyond isolated pilots.
Adoption is universal; maturity is rare. The question has moved from "are we using AI?" to "can we govern and scale it?"
Source: McKinsey, "The Seven Operating Truths of AI-Native Companies" (citing the 2025 Global Survey on AI) · Jun 2026 (2025 survey data)
Only 25% of organisations describe themselves as "Scaling" AI; 47% are piloting and 28% still "understanding". Scaling has risen from 7% (2023) → 10% → 14% → 25% (2026) — yet 62% of the most advanced individual users sit in organisations that have not reached Scaling.
Individual capability runs ahead of organisational capability — the gap between what people can do and what the organisation can govern is itself a maturity measure.
Source: SmarterX / Marketing AI Institute, "State of AI for Business" 2026 (report page; 10 Key Findings) and "State of Marketing AI" 2023–2025 report PDFs · 2023–2026 · Sample: 918 (2023) · 1,784 (2024) · 1,882 (2025) · 2,109 (2026); self-selected, 82% US in 2026
Only 3.1% of organisations at the "optimised" stage; 12.8% in the top two stages; 61.3% still in the two least mature. Mean maturity 2.43 vs 2.39 a year earlier.
The field is standing still. Ambition is everywhere; maturity is rare. IDC elevated governance to a distinct dimension because "trust and risk management became prerequisites for scaling AI".
Source: IDC, "MaturityScape Benchmark: AI-Fueled Organization Worldwide, 2026" · Aug 2026 · Sample: 1,900 organisations, 20 markets
Organisational AI use rose to 78% in 2024 (from 55% in 2023). Inference cost for GPT-3.5-level performance fell ~280× in two years ($20.00 → $0.07 per million tokens).
Cost and access are no longer the constraint. What differentiates organisations now is maturity.
Source: Stanford HAI, "AI Index Report 2025" · Apr 2025
More than two-thirds expect 30% or fewer of their GenAI experiments to fully scale within 3–6 months — yet ~three-quarters say their most advanced initiative meets or exceeds ROI expectations.
Value is real where initiatives are mature. The gap is scaling, not viability. The top barrier — regulation and risk — rose from 28% to 38% across four survey waves.
Source: Deloitte, "State of Generative AI in the Enterprise", Wave 4 · Jan 2025 · Sample: 2,773 director-to-C-suite respondents, 14 countries
The evidence points at integration, process and governance, not at the models.
Only 22% of organisations have scaled AI across multiple business units or adopted an AI-first approach; 11% are entirely unaware of what their function spent on AI in 2025 — while 85% plan to increase spending.
High performers — those who constantly track ROI and treat AI as a portfolio — report positive returns on 81% of initiatives; low performers do not know the return on 29% of theirs. Measurement, not spend, separates them.
Source: Gartner, press release "Only 22% of Organizations Have Successfully Scaled AI Across Multiple Business Units" · 1 Sep 2026 · Sample: 1,303 respondents, organisations ≥ US$50M revenue, fielded Jan–Apr 2026
About 5% of AI pilots achieve rapid revenue acceleration; the vast majority deliver little to no measurable P&L impact.
The root cause is not talent, infrastructure or regulation — it is the absence of learning, integration and contextual adaptation. Maturity is integration plus process plus a feedback loop, not owning a model.
Source: MIT NANDA, "The GenAI Divide: State of AI in Business 2025" · Aug 2025 · Sample: 150 leader interviews, 350 employees surveyed, 300 deployments analysed
~10% of AI value comes from algorithms, ~20% from technology and data, ~70% from people and process change.
The technology is the small part. This is why AIMI weights people, skills and process integration as heavily as tooling.
Source: BCG, the "10-20-70" rule (CEO's Guide to Maximizing Value from AI) · 2024–2026
A clear AI strategy plus workflow redesign lifts measurable business impact by ~25 points; better tools without that strategy move it by ~5.
Operating model beats tool-buying roughly five to one. A tool-heavy, process-light function scores low — and should.
Source: BCG, "AI at Work", 4th edition · Jun 2026 · Sample: 11,749 workers, 14 markets
Real AI value appeared as shifts in decision quality and coordination — not as hours saved — in a four-year fixed-window study with staffing held constant.
"Organisations that judge AI only by hours saved risk missing the real gains." Measuring the wrong thing is itself a maturity failure.
Source: MIT Sloan Management Review, "GenAI Success Metrics: Look Beyond Reduced Workload" · Jul 2026
52% feel prepared on vision and strategy, 39% on risk, security and governance — and only 5% on the business processes themselves. 70% say they cannot trust and govern AI agents.
Average those three and an organisation looks moderately mature. Cap at the weakest and the truth appears. This is why AIMI caps the level instead of averaging.
Source: Deloitte, "AI Agents are Only the Beginning" — AI readiness survey · Aug 2026 · Sample: 501 senior-manager-to-C-suite leaders, US
Named deployments and function-specific benchmarks, because value is captured by functions, not by enterprises.
Customer operations, marketing & sales, software engineering and R&D account for the bulk of an estimated $2.6–4.4 trillion in annual generative-AI value.
Value concentrates in specific functions. That is why AIMI scores at the level of the function, not the enterprise.
Source: McKinsey, "The economic potential of generative AI" · 2023
AI assistant handled 2.3 million conversations in its first month — the work of ~700 full-time agents — resolving in under 2 minutes versus 11, with 25% fewer repeat inquiries.
A named, quantified customer-operations deployment. Note the 2025 rebalancing toward human agents: scale without governance was later corrected.
Source: Klarna (with OpenAI), press release · Feb 2024
50.8% resolution rate and 96% involvement in the first month; roughly 1,700 support hours saved; humans refocused on complex cases.
Buy-versus-build in practice: a specialised vendor tool integrated into a human-led process.
Source: Intercom, "Fin at Anthropic" customer case · 2025
Finance-function AI adoption reached 59% in 2025 (58% in 2024, 37% in 2023) — momentum is slowing, and ~25% of adopters are stuck between planning and piloting.
Deployment is not value. Finance is the clearest case of activity outrunning maturity.
Source: Gartner, 2025 Finance AI survey (via CFO Dive) · Nov 2025 · Sample: 183 CFOs and senior finance leaders
Finance functions are "busier and more automated but not obviously more forward-leaning." The barrier is leadership practice, not technology.
Four practices separate the functions that transform: shared vigilance, routine experimentation, strategic futures-thinking and practice diffusion.
Source: MIT Sloan Management Review, "Why AI Isn't Transforming Finance Yet" · Jun 2026 · Sample: 300+ senior finance professionals, multi-year programme
88% report AI improving asset uptime, service cost and customer experience; 75% say it raised first-time-fix rates.
Field service shows the earliest measurable operational gains — the function where process integration is most visible.
Source: Geotab, "2025 State of Field Service" · Jun 2025
Germany, Austria and Switzerland: high ambition, a lagging mid-market, and boards that are regressing on embedded digital governance.
For the first time a majority of German companies use AI: 57% (36% a year earlier, 20% two years earlier). But among users, 0% say they exploit its potential fully and 59% say "not at all".
Adoption has crossed the majority line in Germany; maturity has not started. 37% expect to fall under the EU AI Act as deployers and 89% expect high implementation effort.
Source: Bitkom, "Erstmals nutzt die Mehrheit der Unternehmen KI" (Bitkom Research) · 14 Sep 2026 · Sample: 603 companies with 20+ employees, Germany, representative; fielded weeks 28–33 2026
54.5% of German companies use AI (from 40.9% a year earlier). Large firms 67.2%, small firms 51.2% — the Mittelstand only 47.2%.
The German mid-market lags even small firms. Roughly three-quarters rely on external solutions; under a fifth build their own.
Source: ifo Institute, AI use in German companies · May 2026
AI rated "highly relevant" by 89% of German executives — the highest of any topic surveyed — yet only 33% consider their company adequately equipped.
The largest relevance-readiness gap of any topic in the study. Ambition far outpaces governed maturity.
Source: KPMG Germany, "Future Readiness Monitor 2026" · Jul 2026 · Sample: 524 executives, 15 sectors
AI strategy prevalence jumped from 31% to 98% in two years — but only 39% of companies have active top-management steering of that strategy.
Near-universal strategy, minority governance. Success is measured by productivity (65%) far more than by revenue or growth (48%) — an efficiency-metric bias.
Source: KPMG Germany, "Generative KI in der deutschen Wirtschaft 2026" · Jun 2026 · Sample: 480 decision-makers
All DAX 40 companies now treat AI as a strategic key technology — yet only 12 of 40 meet all digital-leadership criteria (down from 15), and only 16 of 40 tie digitalisation to executive compensation (down from 28).
German boards are regressing on embedded digital governance even as AI becomes universal in strategy. The first decline since the study began.
Source: DAX Digital Monitor 2025 (FOM / Univ. Duisburg-Essen / Zukunftsinstitut) · Dec 2025
Boards are accountable for a risk most do not yet understand — and board attention tracks value.
93% of audit functions use AI, but 60% have no formal AI strategy and 54% of chief audit executives have not started measuring the ROI of audit's AI use.
The board's own assurance arm cannot yet give assurance on AI value — an oversight gap inside the oversight function.
Source: Gartner, press release "93% of Audit Functions Use AI, but 60% Lack a Formal Strategy" · 10 Sep 2026 · Sample: 161 chief audit executives, May 2026 (AI-use figure from a webinar poll)
54% of finance leaders now lead enterprise AI capital allocation and 48% oversee AI spend; 95% are comfortable with agentic workflows, but only 14% support full autonomy for critical decisions.
AI oversight is migrating to the CFO — a second line of AI governance below the board; 42% say ambitions exceed current capabilities.
Source: Deloitte, "Finance Trends 2027: Finance Leaders Balance AI Ambition With Enterprise Accountability" · 9 Sep 2026 · Sample: 1,434 finance leaders, 26 countries, companies ≥ US$1B revenue, fielded spring 2026
Disclosure of AI as a risk across the S&P 500 jumped from 12% to 83% (2023→2025). Disclosure of AI expertise among directors rose only from 1.5% to 2.7%.
Boards are now accountable for a risk they largely do not understand. The competence gap is real and quantified.
Source: The Conference Board, "Governing AI" (S&P 500 disclosure analysis) · 2026
25% of boards report at least one AI expert; only 14% have integrated that expertise effectively; just 2% of individual directors are AI experts.
Two independent datasets — MSCI (2%) and The Conference Board (2.7%) — converge: director-level AI expertise is genuinely scarce worldwide.
Source: MSCI Institute, "Enhancing AI Governance on Corporate Boards" · 2025 · Sample: 14,500+ directors, listed large/mid-cap companies globally
66% say their boards have limited to no knowledge or experience with AI; 31% say AI is not on the board agenda at all.
Improving (from 79% and 45%), but a two-thirds majority of boards still self-report insufficient AI literacy.
Source: Deloitte Global Boardroom Program, "Governance of AI" · May 2025
Only 26% of boards discuss AI at every meeting. Among high-AI-ROI organisations, 63% do — versus 13% of low-ROI organisations.
Board attention tracks value. Governance engagement is not just oversight — it correlates with return.
Source: Protiviti & BoardProspects, 3rd Global Board Governance Survey · Q4 2025 · Sample: 772 board members and executives worldwide
84% of boards are debating which decisions stay human-led versus AI-led; 86% say board processes caused a delayed, rushed or poor decision in the last six months.
Boards describing, in their own words, the decision-rights problem AIMI is built to measure.
Source: Board Intelligence, "Board Value Index", Summer 2026 · Jun 2026 · Sample: 405 directors, CEOs and CFOs; UK, US, Nordics, Middle East
Nearly half of public-company boards have not formally enabled or standardised AI for their own board work; most lack board-specific AI policies.
A sharper angle than oversight: boards cannot credibly govern a capability they do not use or understand themselves.
Source: Deloitte Center for Board Effectiveness, "How Boards Are Using AI Today" · Jul 2026 · Sample: 92 public and 14 private companies
Only 13% of organisations have all four AI-governance foundations (roadmap, AI council, GenAI policy, ethics policy); 32% have none.
Most organisations lack the governance scaffolding — and the presence of that scaffolding tracks the ability to scale.
Source: SmarterX / Marketing AI Institute, "2026 State of AI for Business" · Jun 2026 (survey Feb–Apr 2026) · Sample: 2,109 respondents (82% US; self-selected, AI-predisposed audience — stated by the publisher)
Organisations at the "Scaling" stage are 8.6× more likely than "Understanding"-stage organisations to have all four governance foundations; average foundations in place: 0.9 (Understanding) → 1.7 (Piloting) → 2.3 (Scaling).
In the publisher's own data governance rises with every stage of scaling — the empirical basis for letting governance cap the level rather than average it away.
Source: SmarterX, "The Artificial Intelligence Show" ep. 220 (Paul Roetzer & Taylor Radey, Dir. of Research) — publisher transcript, discussing the 2026 State of AI for Business survey · 18 Jun 2026 · Sample: 2,109 respondents
Where a dedicated AI leader owns AI, 64% report positive AI momentum; where IT/technology leadership owns it, 47%. A year earlier, 17% of respondents said no one owned AI at all.
Ownership is a maturity variable, not an org-chart detail: named, business-side accountability tracks momentum; IT-only ownership tracks siloed adoption.
Source: SmarterX, "The Artificial Intelligence Show" ep. 220 (publisher transcript, 2026 survey) · Marketing AI Institute, "2025 State of Marketing AI Report" p. 23 (n=1,842) · Jun 2026 / 2025 · Sample: 2,109 (2026) · 1,842 (2025)
What is in force, what has been deferred, and what it costs to get wrong.
Since 2 August 2026, Art. 50 transparency obligations are in force: chatbot disclosure, marking of deepfakes, emotion-recognition and biometric categorisation, and unreviewed AI-generated public-interest text. Fines up to €15M or 3% of global turnover.
Enforced by national market-surveillance authorities together with the European AI Office and the EDPS. Commission guidelines and a Code of Practice are published.
Source: European Commission, "Safer and more transparent AI" · 2 Aug 2026
The Digital Omnibus on AI — Regulation (EU) 2026/1744, in force since 27 July 2026 — deferred high-risk conformity: standalone Annex III systems to 2 December 2027, AI embedded in regulated products to 2 August 2028.
It did not touch Art. 50. A deferral is not permission to wait: it is the window in which prepared organisations pull ahead.
Source: Official Journal of the EU, Regulation (EU) 2026/1744 · 24 Jul 2026
Penalty tiers (Art. 99): €35M or 7% of global turnover for prohibited practices; €15M or 3% for other provider and deployer obligations; €7.5M or 1% for misleading information to authorities. For SMEs, the lower of the two applies.
The ceiling exceeds the GDPR's €20M / 4%. For a DACH enterprise, AI maturity and AI-Act readiness have become the same conversation.
Source: Regulation (EU) 2024/1689 (EU AI Act), Art. 99 · in force since Aug 2025
Vendor-commissioned surveys with undisclosed samples; figures that exist only in secondary coverage; self-reported "competence" without evidence; and any number we could not trace to its origin. Several widely-quoted statistics are absent from this page for exactly that reason.
See where your function stands against this evidence.
Assess a function