Enterprise AI Statistics 2026: Adoption, ROI & Deployment
Most surveyed organizations now use AI somewhere. Far fewer can show what it adds to the income statement. Stanford HAI's 2026 AI Index reports that 88% of surveyed organizations used AI in 2025, yet McKinsey's August 2026 survey finds that only 37% of respondents attribute any EBIT impact to it. This guide separates adoption, deployment, scale and value, and explains what each enterprise AI statistic for 2026 actually measures.
TL;DR
Adoption is mainstream. Stanford HAI reports 88% of surveyed organizations used AI in 2025, and McKinsey's 2026 survey finds nearly nine in ten respondents use it regularly in at least one function.
Scale is partial. 44% of McKinsey respondents say AI is scaling across their enterprise, up from 38%. In Deloitte's survey, fielded in 2025, only 25% had moved 40% or more of pilots into production.
Value is uneven, and definitions drive the numbers. McKinsey: 37% report some EBIT impact and about 6% are high performers. BCG: almost half of companies generate value. KPMG: 7% report established ROI.
Agents are scaling mostly in large firms. 40% of McKinsey respondents at $1 billion-plus organizations report scaling agents versus 22% at smaller ones, while BCG finds only 5% of companies have the full set of agent controls.
Spending data is easy to misread. Gartner forecasts $2.7 trillion in worldwide AI spending for 2026, more than half of it infrastructure. That is not an enterprise software budget.
Key 2026 enterprise AI statistics: Stanford HAI reports 88% of surveyed organizations used AI in 2025. McKinsey finds 44% are scaling AI across the enterprise, while 37% attribute some EBIT impact to it and about 6% are high performers. Gartner forecasts $2.7 trillion in worldwide AI spending for 2026.
Table of contents:
Enterprise AI Statistics 2026 at a Glance
The 2026 evidence shows a gap rather than a single adoption rate. Use is near-universal among surveyed organizations, enterprise-wide scaling covers roughly half of large-company respondents, and attributed financial impact is narrower still. The table lists headline figures and what each one measures, because they are not interchangeable.
Metric | 2026 finding | What it actually measures | Source |
|---|---|---|---|
Organizational AI use | 88% | Respondents whose organization uses AI in at least one function (2025 data, self-reported) | |
Enterprise-wide scaling | 44% | Respondents saying AI is scaling across the enterprise (38% a year earlier) | |
Any EBIT impact | 37% | Respondents attributing at least some EBIT impact to AI; flat on 2025 | |
AI high performers | About 6% | At least 5% of EBIT attributed to AI, with "significant" impact | |
Companies generating value | Almost 50% | BCG tiers: 7.5% future-built plus 41% scaling (1,330 executives) | |
Established ROI | 7% | Leaders reporting established ROI (2,145 leaders, 20 markets, April to May 2026) | |
Pilots reaching production | 25% | Respondents who moved 40% or more of pilots into production (fielded 2025) | |
Agent scaling, large firms | 40% | Respondents at $1 billion-plus organizations scaling agents in at least one function | |
Company AI spend | 3.3% of revenue | Survey-reported average; about 80% sits outside enterprise IT | |
Worldwide AI spending | $2.7 trillion (forecast) | Gartner forecast covering infrastructure, software and services, not enterprise budgets | |
AI-enabled breach cost | $6 million | Average for AI-enabled malicious breaches versus a $4.99 million global average |
How Widespread Is Enterprise AI Adoption in 2026?
Enterprise AI adoption is mainstream when it means regular use in at least one business function. Stanford HAI's 2026 AI Index reports that 88% of surveyed organizations used AI in 2025, up from 78% in 2024, drawing on McKinsey's 2025 survey. Stanford notes that such self-reported results are directional. McKinsey's 2026 survey finds nearly nine in ten respondents report regular use.
Depth is lower. McKinsey surveyed 1,719 participants in 97 nations from May 4 to June 8, 2026. It finds 44% say AI is scaling across their enterprise (38% a year earlier) and 56% use AI in three or more functions (51%). Size matters: 54% of respondents at organizations with $1 billion or more in revenue report enterprise-wide scaling, versus one-third at smaller organizations.
Deloitte's 2026 State of AI in the Enterprise surveyed 3,235 leaders at organizations on the leading edge of AI in August and September 2025. Worker access to sanctioned AI tools rose from under 40% to about 60%, yet only 25% of respondents had moved 40% or more of their pilots into production.
Reported adoption rates vary because surveys define adoption differently. Some count any use in one function, others production deployment, scaling or enterprise-wide transformation. Samples differ too: McKinsey weights by national GDP, KPMG surveys senior leaders at larger companies, and Deloitte targets AI-advanced organizations. Compare figures only within the same stage.
Stage | What it means | Example figures |
|---|---|---|
Experimentation | Pilots and trials without steady operational use | Rarely reported as a standalone rate |
Use | Regular use in at least one function | 88% (Stanford, 2025 data); nearly nine in ten (McKinsey 2026) |
Production | Pilots moved into live operations | 25% moved 40% or more of pilots (Deloitte) |
Scaling | Deployed beyond isolated teams, across the enterprise | 44% (McKinsey); 64% Americas, 61% Asia Pacific, 56% EMEA at "scaling or beyond" (KPMG Q3) |
Enterprise-wide transformation | Products, processes or operating model redesigned | 34% deeply transforming (Deloitte); 7.5% future-built (BCG) |
For executives, adoption alone is no longer a meaningful maturity benchmark. A company whose staff use a chatbot has access, not a scaled deployment.
Generative AI and Agentic AI Adoption
Generative AI is widely used, while agentic AI is scaling mainly in large enterprises and is not yet widespread. Stanford's chapter reports generative AI use in at least one function at 70% of organizations in its highlights but 79% (up from 71%) in the figure text for 2025, so check which figure a source quotes. McKinsey's 2026 survey finds chatbots are the most widely scaled tool: 47% of respondents are scaling them across the enterprise.
Agents lag. In Stanford's 2025 data, scaled agent use was in the single digits for nearly all business functions, and even in IT and knowledge management about two-thirds or more of respondents reported no use. A year later, McKinsey finds 40% of respondents at organizations with more than $1 billion in revenue report scaling agents in at least one function, up from 27%, while the share at smaller organizations stayed flat at 22%. About two in ten are scaling software coding agents (31% at larger enterprises).
Other sources are not directly comparable. Deloitte, fielded in 2025, found 23% of companies using agentic AI at least moderately. KPMG's Q3 survey found 34% reporting significant employee adoption of AI agents, up from 25% in Q1. Microsoft's Work Trend Index reports active agents in Microsoft 365 up 15x year over year (18x in large enterprises), a telemetry count inside Microsoft's own ecosystem.
Agent use is not agent autonomy. BCG finds 42% of companies expect their agents to act autonomously by 2030, deciding without human approval, while just 5% have the full set of critical controls today. BCG's agentic share of AI value was 17% in its 2025 sample and 22% in 2026, and it projects 39% by 2030. Treat "using agents" and "agentic transformation" as different claims.
Enterprise AI Adoption by Business Function
AI is used most where work is digital, text-heavy and measurable. Reliable function rankings are scarce, so treat these as patterns. In McKinsey's 2026 survey, respondents most often report scaling agents in IT, knowledge management and software engineering. Technology companies lead on agents, consumer goods and retail firms use them most in marketing and sales, and advanced manufacturers in supply chain, inventory and the manufacturing process.
Stanford's chapter, using McKinsey's 2025 data, reports the highest use in knowledge management for business, legal and professional services (58%), software engineering (58%) and IT (56%) in the technology sector, and marketing and sales in consumer goods and retail (51%). Strategy and corporate finance, and risk and compliance, show low uptake in most sectors; financial services is the exception for risk and compliance.
Reported value follows a similar split. McKinsey's 2026 respondents most often report cost reductions in supply chain management, service operations and manufacturing, and revenue gains in marketing and sales, product and service development, and software engineering. McKinsey measured impact at the function level in 2026 rather than summing use cases, so do not compare these patterns directly with its 2025 figures.
Enterprise AI ROI Statistics
No defensible universal average ROI exists for enterprise AI. Studies measure different outcomes (productivity, cost, revenue, EBIT, or leaders' view of established ROI) and report very different shares, so the useful question is which kind of value a figure describes.
McKinsey finds 80% of respondents say AI improved their individual productivity and about half say it improved their decisions, yet 37% attribute any EBIT impact to AI, unchanged from 2025, and about 6% qualify as high performers. Respondents also cite organization-wide gains in innovation, competitive differentiation, customer satisfaction and employee satisfaction.
Deloitte, fielded in 2025, found two-thirds (66%) report productivity or efficiency gains, 40% report lower costs and 20% report higher revenue, while 74% hope to grow revenue through AI. KPMG's Q2 survey found 7% of leaders report established ROI. In KPMG's US Q3 pulse of 314 leaders at organizations with $1 billion or more in revenue, 58% report measurable business value.
BCG reports that almost half of surveyed companies generate value, and its future-built tier shows 2.4 times the top-line growth of companies in the bottom half of its sample. These are comparisons between groups, not proof that AI caused the outperformance.
Value type | Strongest 2026 evidence | Caveat |
|---|---|---|
Individual productivity | 80% report better personal productivity (McKinsey); 66% report productivity or efficiency gains (Deloitte) | Perception; time saved is not financial return |
Task output | +14% to 15% issues resolved per hour, +26% completed pull requests, +50% marketing output (studies summarized by Stanford) | Narrow tasks; one study found experienced developers 19% slower |
Cost | 40% report lower costs (Deloitte); cost reductions most reported in supply chain, service operations, manufacturing (McKinsey) | Reported, not audited; no common percentage |
Revenue | 20% report higher revenue versus 74% hoping to (Deloitte) | Aspiration exceeds reported results |
EBIT | 37% some EBIT impact; about 6% at 5% or more (McKinsey) | Respondent attribution, not audited |
Established ROI or value | 7% established ROI (KPMG); almost 50% generate value (BCG); 58% measurable value (KPMG US) | Definitions differ; BCG's tiers include "some value" |
Keep three things apart: reported benefit, measured ROI and material enterprise financial impact. A statistic about one is not evidence for the others.
Why Some Companies Get More AI Value Than Others
Studies point to the same drivers: redesigned workflows, senior ownership, measurement, shared platforms and controls. Most of the evidence is correlational. It shows what leaders have in common, not that these practices cause returns.
Workflow redesign. McKinsey: nearly three-quarters of high performers fundamentally redesign workflows because of AI (up from 55%), versus one-quarter of other respondents. They are also about twice as likely to say senior leaders show commitment and that defined processes measure impact.
Strategic focus. BCG: companies strong on both strategic clarity and applied AI generate five times as much AI value as those that are not. Among future-built companies, 61% fund a single multiyear program, versus 17% of laggards.
Measurement. BCG: 95% of future-built companies use clear KPIs or track P&L value from AI, and those tracking directly in the P&L realize three times the AI value of companies that do not formally measure it (3.6% versus 1.2%).
Accountability and cost visibility. KPMG's Q2 survey: organizations where the CEO is accountable for decisions based on AI outputs report established ROI at 14% versus 4%, and those with strong cost visibility 15% versus 3%. Only 24% of leaders say the CEO is accountable for AI-driven outcomes.
Organization over individual. Microsoft finds that culture, manager support and talent practices account for 67% of reported AI impact versus 32% for individual mindset and behavior, and labels this a statistical association from self-reported data.
Enterprise AI Spending and Budget Statistics
Worldwide AI spending is forecast at $2.7 trillion in 2026, but the total is dominated by infrastructure bought largely by technology providers and hyperscalers. It is not an enterprise software budget. Gartner's September 16, 2026 forecast puts 2026 spending at $2.67 trillion, up 49.5% from $1.79 trillion in 2025, with $3.64 trillion for 2027. It raised its May estimate of $2.59 trillion.
In Gartner's table, 2026 infrastructure is $1.48 trillion, about 56% of the total (our calculation). Services follow at $576 billion, software $462 billion, AI cybersecurity $51 billion, AI agents and assistants $29 billion (from $16 billion) and generative AI models $28 billion (up 117%). Excluding infrastructure, the remaining categories total about $1.19 trillion (our calculation from the same table). Gartner adds that enterprises mostly receive AI as embedded features from incumbent software providers, and that lock-in, data sovereignty and runaway-cost risks are not deterring buyers.
Other totals measure other things. Stanford reports global corporate AI investment of $581.69 billion in 2025 (up 129.9%), counting mergers and acquisitions, minority stakes, private investment and public offerings in AI companies, not enterprise purchasing. Google alone reported more than $150 billion of capital expenditure in 2025.
At company level, BCG reports survey-based AI spending of 3.3% of revenue, double a year earlier, with about 80% outside enterprise IT budgets. McKinsey finds 28% of respondents spend more than 10% of their enterprise ICT budget on AI and 60% expect to increase AI investment next year. KPMG's Q3 survey reports average planned AI investment of $210 million over 12 months, up from $186 million in Q1.
What Enterprise AI Really Costs
No source publishes a universal total cost of ownership (TCO) for enterprise AI, and license price is a poor proxy because usage-based costs grow with volume. Use a cost map and fill it with your own figures; few enterprises incur every category.
Visibility is the weak point. KPMG's Q2 survey found 42% of leaders have only partial visibility into AI spending, 33% struggle to understand cost structures including tokens, and 23% struggle with usage-based costs. McKinsey finds about 20% of respondents say operating costs, including tokens, constrained their AI use.
Model and API usage, including inference and tokens
Licenses and platform fees
Cloud and compute
Data preparation, retrieval and context infrastructure
Integration with existing systems
Security, identity and access controls
Observability and evaluation
Governance and compliance
Implementation, training and change management
Ongoing operations and maintenance
Then calculate operating cost per workflow: fully loaded monthly AI cost (usage, platform and human oversight) divided by completed workflows. McKinsey frames the test as comparing agentic AI against the fully loaded cost of the work it replaces.
Enterprise AI Deployment Models in 2026
No study ranks deployment architectures by value. The evidence shows several patterns coexisting, and the right mix depends on data sensitivity, control needs and the workflow.
Embedded AI in incumbent software. Gartner says generative AI is in the Trough of Disillusionment in 2026 and enterprises are using simpler embedded features from incumbent providers.
Copilots and assistants. Worker access to sanctioned tools rose from under 40% to about 60% (Deloitte), and Microsoft found 49% of sampled Copilot conversations support cognitive work such as analysis and problem-solving.
Custom applications on models and APIs. Gartner raised its 2026 growth forecast for AI application development platforms to 39% from 28%, and 32% of McKinsey respondents declined to buy software they could build with agentic coding tools.
Controlled or sovereign deployments. 72% of KPMG respondents formally consider model sovereignty in AI decisions, and 21% have an enterprise-wide strategy under regular review (KPMG Q3).
Shared platforms and control layers. More than two-thirds of BCG's future-built companies are committing to a single enterprise-wide AI platform with a common architecture and control plane, rather than a single-vendor stack. 55% of KPMG respondents operate a formal AI harness layer.
The primary sources reviewed here give no comparable adoption rates for retrieval-augmented generation, fine-tuning or multi-model strategies, so this article does not rank them.
Build vs Buy vs Partner for Enterprise AI
No evidence shows one route winning for every enterprise, and the balance is shifting. McKinsey finds 32% of respondents decided against buying at least one software product or feature because they could build it in-house with agentic coding tools, most often in technology and healthcare. Gartner finds enterprises mostly receiving AI through incumbent vendors. The table is editorial synthesis, not a published standard.
Dimension | Buy (embedded or SaaS) | Build | Partner |
|---|---|---|---|
Speed to first value | Fastest | Slowest | Moderate |
Differentiation | Low; rivals get the same features | Highest | Medium |
Control and data sensitivity | Vendor-set terms | Highest control | Contract-dependent |
Engineering burden | Low | High | Shared |
Governance and audit | Depends on vendor logs | Fully owned | Split; define roles |
Lock-in risk | High | Lower, but maintenance debt | Medium; check exit terms |
Cost profile | Seat or usage fees | Build plus run costs, including tokens | Fees plus services |
Best fit | Standard workflows | Core, differentiating workflows | Capability gaps, fast starts |
Before signing with any AI vendor or platform, ask:
Model choice and portability: can models be swapped without rebuilding workflows?
Data: what is retained, and is customer data used for training?
Permissions and identity: what can agents access or do, and who approves it?
Security: which certifications, data-residency options and incident history?
Auditability: are agent actions logged and exportable?
Evals: how are accuracy and hallucinations tested and monitored?
Human approval: can review thresholds be configured?
Integrations and scale: what does cost per task look like at 10x volume?
Lock-in and exit: can prompts, workflows, agents and data be exported?
Enterprise AI Deployment Maturity
This six-level model is editorial synthesis, not an official standard. Each level lists observable characteristics, with a published figure as an anchor rather than a threshold.
Level | Observable characteristics | Evidence anchor |
|---|---|---|
1. Exploration | Individual use of tools; no owner, budget or measures | Deloitte: worker access to sanctioned tools about 60% |
2. Pilot | Funded trials in isolated teams; value unmeasured | Deloitte: 25% moved 40% or more of pilots to production |
3. Production | Live workflow with an owner, evaluations and cost tracking | KPMG: 61% review costs at approval, 59% monitor in operation |
4. Functional scale | Several workflows redesigned within a function | McKinsey: 56% use AI in three or more functions |
5. Enterprise scale | Shared platform, controls and measurement across functions | McKinsey: 44% scaling enterprise-wide; KPMG: formal harness layer at 58% of scaling organizations |
6. Agentic operating model | Agents with defined authority; roles redesigned around humans and agents | BCG: 7.5% future-built; 5% have full agent controls |
Workforce and Productivity Statistics
Worker access and individual productivity gains are well documented. Aggregate productivity and employment effects are earlier, mixed and uneven. This section keeps access, productivity, skills, hiring, wages and expectations apart.
Access and use. Deloitte: worker access rose from under 40% to about 60%. Microsoft, surveying 20,000 AI users in 10 countries from February to April 2026, found 66% say AI frees time for high-value work and 58% say they produce work they could not a year ago, rising to 80% among its "Frontier Professionals" (16% of AI users). These are self-reported.
Task productivity. Stanford summarizes studies showing 14% to 15% more customer-support issues resolved per hour, 26% more completed pull requests and 50% more marketing output. A study of experienced open-source developers found them 19% slower with AI, though its authors later reported difficulty replicating the result.
Aggregate productivity. Stanford cites a 4% labor-productivity gain across 12,000 European firms, US productivity growth of 2.7% in 2025 against a 1.4% decade average, and a survey of 6,000 executives in four countries that found minimal realized productivity gains so far.
Skills and wages. PwC analyzed more than one billion job ads in 27 countries and territories: jobs requiring AI skills grew 69% versus 9% for all jobs, and the average AI-skills wage premium reached 62%, up from 57%. Productivity grew 34% from 2018 to 2025 in the most AI-exposed sectors versus 24% in the least exposed, a correlation rather than a causal estimate.
Hiring and headcount. McKinsey: 14% of respondents say AI contributed to a workforce decline in the past year, versus 32% who expected one a year earlier, and 39% expect a decline next year. Stanford reports employment of US software developers aged 22 to 25 fell nearly 20% from its 2022 peak by September 2025, while large-scale job losses have not yet shown up in overall employment data. BCG finds 89% of companies expect AI to generate new work.
Task exposure is not job elimination, and employer expectations are not observed layoffs. McKinsey's own comparison shows expectations ran well ahead of reported outcomes.
AI Governance, Security, and Regulatory Risk
Governance is lagging deployment. The strongest 2026 evidence links control maturity to value and to breach cost. This section is informational, not legal advice.
Agent controls. BCG: 42% expect autonomous agents by 2030 while 5% have the full control set, and companies with all six AI controls generate three times the agentic value of those with one. Deloitte, fielded in 2025, found agentic adoption outpacing governance.
Management layer. KPMG's Q3 survey: 55% operate a formal AI harness layer (31% in experimentation, 58% scaling, 86% with established ROI). 53% place accountability for AI-informed decisions at C-suite level or above, and 86% are adapting their cybersecurity operating model.
Agent permissions. Microsoft advises treating agents as managed entities with identities, permissions, policy enforcement and lifecycle management. IBM recommends tightly scoped permissions enforced at runtime, human attribution and auditability.
IBM's 2026 Cost of a Data Breach study, covering 602 organizations and breaches from March 2025 to February 2026, found one in four malicious breaches was AI-enabled (up 56%), averaging $6 million against a $4.99 million global average. More than 20% of organizations reported a breach targeting AI models or applications, most often through compromised APIs, applications or plug-ins (27%) and cloud misconfigurations affecting AI workloads (27%). Organizations using AI and automation in security cut breach costs by almost $2 million, yet one in four had not adopted them.
EU AI Act. The European Commission states that the Act became applicable on 2 August 2026, with exceptions. Prohibited practices and AI literacy duties applied from 2 February 2025, and general-purpose AI obligations from 2 August 2025. Under the AI Omnibus, in force since 27 July 2026, high-risk rules for certain sensitive uses apply from 2 December 2027 and for AI embedded in regulated products from 2 August 2028. Transparency rules, including chatbot disclosure and labeling of deepfakes and certain AI-generated text, take effect from August 2026, and from 2 August 2026 the AI Office and national authorities supervise and enforce the Act. See the Commission's transparency guidelines.
NIST. The AI Risk Management Framework, released January 26, 2023, is voluntary. The Generative AI Profile (AI 600-1) followed on July 26, 2024. NIST published a concept note for a critical-infrastructure profile on April 7, 2026, and says AI RMF 1.0 is being revised as part of the White House AI Action Plan.
Data Readiness and Infrastructure
Production AI needs more than a model: trusted, permissioned data, retrieval and context, integration, observability, evaluation, security and compute. Statistics are thinner here than for adoption, but they point the same way.
KPMG's US Q3 pulse names data readiness and access as the top barrier to AI agent deployment.
BCG: 95% of future-built companies are undergoing a data transformation, and more than two-thirds are committing to one enterprise-wide AI platform.
Deloitte reports leaders feel more strategically than operationally ready: in 2025 fieldwork, 42% said their strategy was highly prepared for AI adoption and 30% said the same of risk and governance.
IBM: the leading causes of AI-targeted breaches were weaknesses in surrounding systems, not the models: compromised APIs, applications or plug-ins and cloud misconfigurations (27% each).
Microsoft argues that the more agents execute, the higher the stakes for human evaluation, so evaluation infrastructure must keep pace.
Enterprise AI Adoption by Industry and Region
Industry and regional comparisons are directional, because studies use different samples and definitions. This section describes patterns instead of a league table.
Industry. McKinsey finds technology companies most likely to report scaling agents, and respondents in pharmaceuticals and medical products, insurance, and banking most likely to expect rising AI investment. BCG found at least one future-built or scaling organization in each of its 20-plus sectors. Among ten sectors it highlights, technology pairs the highest applied-AI maturity with the strongest realized value, banking's cost gains run ahead of revenue impact, and machinery and chemicals report modest value. PwC found the AI-skills wage premium ranging from 118% in consumer markets to 16% in government and public sector work.
Region. Stanford, citing McKinsey's 2025 survey, reports organizational AI use of 91% in Europe and 90% in North America, against 88% globally, with China and Europe posting the largest year-over-year increases (13 and 11 points). KPMG's Q3 survey measures maturity instead: 64% of Americas organizations are scaling AI or beyond, versus 61% in Asia Pacific and 56% in EMEA, a spread that halved from 16 points in Q1 to eight.
Do not confuse these with population-level figures. Stanford's Microsoft-based diffusion data put the United Arab Emirates at 64% and the United States at 28.3% (24th), which measures individuals, not organizations.
The Enterprise AI ROI Gap: Why Adoption Does Not Equal Value
The ROI gap is the distance between using AI and attributing financial results to it. Four findings define it.
Use outruns value. 88% of surveyed organizations use AI (Stanford), while 37% attribute any EBIT impact and about 6% attribute 5% or more (McKinsey).
Individual gains outrun enterprise gains. 80% report better personal productivity, yet the EBIT share did not move even as enterprise-wide scaling rose from 38% to 44%.
Spending outruns measurement. Only 12% of organizations consistently assess AI value against cost, rising to 48% among those with established ROI (KPMG Q3).
Tools outrun redesign. 37% of Deloitte respondents use AI at a surface level with little process change, and about one-quarter of McKinsey's non-high-performers redesign workflows.
The studies are less contradictory than they look. McKinsey asks about EBIT attribution, KPMG about established ROI, and BCG's "almost half" adds its scaling tier, which creates some value, to the 7.5% of future-built companies (7.5% plus 41%, our calculation). Three different questions, three different samples.
Other contributors the evidence supports include isolated pilots, weak ownership (only 24% name the CEO as accountable for AI outcomes in KPMG's Q2 survey), cost constraints (about 20% in McKinsey's survey) and missing controls (5% in BCG's). Adoption measures access. Value measures changed work.
What Enterprise AI Leaders Do Differently
These behaviors recur across 2026 studies. Each is an association, not a guaranteed cause.
Redesign the workflow, not the task. Nearly three-quarters of McKinsey's high performers do, against one-quarter of others.
Concentrate. BCG's future-built companies fund one multiyear program (61% versus 17% of laggards) and focus on fewer, higher-value workflows.
Put an executive on the hook. KPMG: CEO accountability goes with established ROI of 14% versus 4%.
Measure on the P&L. BCG: 95% of future-built companies use KPIs or P&L tracking.
Make cost visible. KPMG: strong cost visibility goes with established ROI of 15% versus 3%.
Build the control layer before granting autonomy. BCG: three times the agentic value with all six controls. KPMG: harness layers at 86% of organizations with established ROI.
Plan the workforce. BCG: 55% of future-built companies do strategic workforce planning versus 17% of laggards. Microsoft's separate study of 1,800 workers found a 17-point lift in reported AI value when managers modeled AI use.
Fix data and platform. BCG: 95% of future-built companies are undergoing a data transformation.
Enterprise AI Benchmarks for Decision-Makers
Use these questions to place your organization against published reference points. The figures are context from different surveys, not targets, and each keeps its own definition.
Area | Question to ask | Published reference point |
|---|---|---|
Stage | Are we experimenting, deploying or scaling? | 44% scaling enterprise-wide (McKinsey) |
Redesign | What share of priority workflows has AI embedded and redesigned? | About three-quarters of high performers versus one-quarter of others (McKinsey) |
Pilot conversion | How many pilots reach production? | 25% moved 40% or more (Deloitte) |
Business metric | Which P&L line has moved? | 37% any EBIT impact; about 6% at 5% or more (McKinsey) |
Cost per workflow | What does a completed workflow cost, fully loaded? | No published benchmark |
Value versus cost | Do we assess it consistently? | 12%; 48% among established ROI (KPMG Q3) |
Cost controls | Are costs reviewed at approval and monitored? | 61% and 59% (KPMG Q3) |
Accountability | Is a named executive responsible? | 53% at C-suite level or above (KPMG Q3) |
Agent controls | Are agents permissioned, logged and monitored? | 5% have the full set (BCG) |
Retirement | Do we stop projects that miss thresholds? | No benchmark; Gartner predicts over 40% of agentic projects canceled by end of 2027 |
In our reading, a CFO view starts with value versus cost and EBIT, a CIO view with control layers, platform and cost per workflow, and a board view with accountability, agent authority and risk exposure.
Enterprise AI Outlook: What the 2026 Data Suggests for 2027
This section separates measured facts, published forecasts and editorial interpretation.
Measured now. Adoption is broad, scaling is partial, and enterprise financial attribution has not moved: McKinsey's EBIT share was 37% in both 2025 and 2026.
Published forecasts. Gartner forecasts worldwide AI spending of $3.64 trillion in 2027, and spending on AI agents and assistants rising from $29.2 billion in 2026 to $65.5 billion. Gartner's June 2025 prediction is that over 40% of agentic AI projects will be canceled by the end of 2027 because of escalating costs, unclear business value or inadequate risk controls, and that at least 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028. BCG projects agents at 39% of AI value by 2030. EU high-risk rules apply from December 2027 and August 2028, and 60% of McKinsey respondents expect to raise AI investment next year.
Our interpretation. Five themes follow from the data. None is a forecast.
Agent scaling will likely depend on control layers, because autonomy expectations are running ahead of controls.
Cost scrutiny is likely to grow, moving AI FinOps from specialist practice toward routine finance management as token costs begin to constrain use.
Build-versus-buy decisions may keep shifting as coding agents lower the cost of building.
Shared platforms and embedded vendor features will likely compete for the same budget.
Proof of value is likely to precede budget, so measurement systems may matter as much as models.
FAQ
What percentage of enterprises use AI in 2026?
Stanford HAI reports 88% of surveyed organizations used AI in at least one function in 2025, based on McKinsey data, and McKinsey's 2026 survey finds nearly nine in ten respondents use AI regularly. This measures use, not production deployment or value. McKinsey adds that 44% say AI is scaling across their enterprise.
What percentage of organizations use generative AI?
Stanford's AI Index reports generative AI use in at least one business function at 70% of organizations in its chapter highlights and 79% in the figure text for 2025, so quote the source and the figure. McKinsey's 2026 survey finds 47% of respondents scaling chatbots across the enterprise.
How many enterprises are using or scaling AI agents?
It depends on the definition. McKinsey's 2026 survey finds 40% of respondents at organizations with over $1 billion in revenue are scaling agents in at least one function, versus 22% at smaller ones. Deloitte, fielded in 2025, found 23% using agentic AI at least moderately. Stanford found single-digit scaled agent use across nearly all functions in 2025.
Are enterprises getting ROI from AI?
Some are, but measures differ. McKinsey finds 37% attribute some EBIT impact and about 6% are high performers. BCG says almost half of companies generate value. KPMG's Q2 survey found 7% report established ROI. Individual productivity gains are far more common than enterprise financial impact.
What is the average ROI of enterprise AI?
No defensible universal average exists. Studies report productivity, cost, revenue, EBIT or perceived ROI, using different populations and definitions, and most are self-reported. Averaging them would mislead. Measure ROI per workflow against a baseline, and compare your result with the exact definition a benchmark uses.
Why do enterprise AI adoption statistics differ?
Surveys count different things: any use in one function, production deployment, scaling or enterprise-wide transformation. Samples differ too, from McKinsey's GDP-weighted global respondents to Deloitte's AI-advanced organizations and KPMG's senior leaders. Fieldwork dates matter as well; Deloitte's 2026 report was fielded in 2025.
How much are companies spending on AI?
Gartner forecasts $2.7 trillion in worldwide AI spending for 2026, more than half of it infrastructure, so it is not an enterprise budget. At company level, BCG reports 3.3% of revenue, McKinsey finds 28% of respondents spend over 10% of their ICT budget on AI, and KPMG reports $210 million in average planned investment.
What stops enterprise AI projects from scaling?
The evidence points to unclear ownership, missing workflow redesign, weak cost and value measurement, data readiness and absent controls. KPMG's US pulse names data readiness and access as the top barrier to agent deployment, McKinsey finds about 20% constrained by operating costs, and BCG finds only 5% have the full set of agent controls.
What is the difference between an AI pilot and scaled deployment?
A pilot tests AI in a limited setting, often without an owner, controls or financial measures. Scaled deployment runs AI across functions or the enterprise with redesigned workflows, a shared platform, monitoring and cost tracking. Deloitte found only 25% of respondents had moved 40% or more of pilots into production.
Is agentic AI widely deployed in enterprises yet?
Not yet widely. Scaling is concentrated in large companies: 40% of McKinsey's large-organization respondents versus 22% of smaller ones. Stanford found single-digit scaled use across nearly all functions in 2025. BCG finds 42% expect autonomous agents by 2030, but only 5% have the full control set today.
How should enterprises measure AI ROI?
Tie each initiative to a business objective, set a baseline, and track P&L effects rather than activity. BCG finds companies tracking directly in the P&L realize three times the value of those not formally measuring. Include fully loaded costs and compare value against cost consistently, which only 12% do, according to KPMG.
What are the biggest enterprise AI risks?
The evidence highlights uncontrolled agents, security weaknesses and ungoverned costs. IBM found one in four malicious breaches was AI-enabled, averaging $6 million, and over 20% of organizations reported breaches targeting AI models or applications. Regulatory exposure matters too: EU AI Act transparency rules apply from August 2026.
What separates AI leaders from laggards?
Leaders redesign workflows, concentrate on fewer high-value programs, assign executive accountability, measure on the P&L, track costs and build control layers. McKinsey's high performers redesign workflows at about three times the rate of others. These are associations in survey data, not proof of cause.
Should enterprises build or buy AI?
There is no universal answer. Buying embedded features is fastest and common, building suits differentiating workflows, and partnering fills capability gaps. McKinsey finds 32% of respondents declined to buy software they could build with coding agents. Decide by differentiation, data sensitivity, control needs, total cost and exit options.
What enterprise AI trends matter going into 2027?
Watch agent scaling alongside control layers, cost discipline for token-based usage, shifting build-versus-buy economics and EU high-risk rules from December 2027. Gartner forecasts $3.64 trillion in AI spending for 2027 and predicted in 2025 that over 40% of agentic projects will be canceled by the end of 2027.
Key Takeaways
Adoption is mainstream: 88% of surveyed organizations used AI in 2025 (Stanford, using McKinsey data).
Scale is partial: 44% report enterprise-wide scaling (McKinsey), and 25% of Deloitte respondents moved 40% or more of pilots into production.
Value claims depend on definitions: 37% any EBIT impact and about 6% high performers (McKinsey), almost half generating value (BCG), 7% established ROI (KPMG Q2).
Individual productivity gains are common. Enterprise financial impact is not.
Agents are scaling mainly in large firms, and autonomy expectations are running ahead of controls.
Gartner's $2.7 trillion is a forecast dominated by infrastructure, not enterprise software spend.
Cost visibility and executive accountability accompany better outcomes in KPMG's data (15% versus 3%, and 14% versus 4%, established ROI).
Governance is both a value and a risk issue, according to IBM, BCG and KPMG data.
Actionable Next Steps
Classify every AI initiative as experiment, pilot, production, functional scale or enterprise scale, and report them separately.
Name an accountable executive for AI outcomes and a business owner for each priority workflow.
Pick three to five high-value workflows and redesign them end to end instead of adding tools.
Set baselines and measure effects on P&L lines (revenue, cost, EBIT). Do not report time saved as ROI.
Build an AI cost map and calculate operating cost per workflow, including token and oversight costs.
Put a control layer in place before granting autonomy: agent identity, scoped permissions, audit logs, evals and human approval thresholds.
Fix data access and permissions for priority workflows before scaling retrieval and agents.
Review EU AI Act transparency duties with counsel, and set kill criteria for projects that miss value thresholds.
Glossary
Enterprise AI: AI used inside business workflows with data access, controls and owners, rather than by individuals on their own.
Generative AI: AI that produces text, images, code or other content from prompts.
Foundation model: a large model trained on broad data and adapted to many tasks.
Large language model (LLM): a foundation model trained on text to understand and generate language.
AI agent: software that uses a model to plan and take actions toward a goal using tools, with varying human oversight.
Agentic AI: systems of AI agents that carry out multi-step work with some autonomy.
Retrieval-augmented generation (RAG): retrieving relevant company data and giving it to a model so answers draw on it.
Fine-tuning: further training a model on specific data to adapt its behavior.
Inference: running a trained model to produce an output; a main source of usage cost.
Token: a unit of text a model reads or writes; usage pricing often counts tokens.
AI ROI: the financial return from AI relative to its full cost, measured against a baseline.
Total cost of ownership (TCO): all costs of an AI system over its life, not only licenses.
EBIT: earnings before interest and taxes.
AI governance: the policies, roles and controls that manage AI risk and accountability.
Evals: structured tests of whether a model or agent meets quality and safety requirements.
Observability: monitoring that shows what an AI system did, why, and at what cost.
Human-in-the-loop: a design in which a person reviews or approves AI outputs or actions.
Model orchestration: coordinating several models, tools and agents inside one workflow.
AI harness layer: KPMG's term for the controls and tooling between AI models and business use.
AI FinOps: financial management of AI usage, including cost allocation, budgets and unit economics.
Shadow AI: employee use of AI tools without organizational approval or oversight.
Sources & References
Stanford Institute for Human-Centered AI. The 2026 AI Index Report, Chapter 4: Economy. April 2026. https://hai.stanford.edu/ai-index/2026-ai-index-report/economy
McKinsey & Company (QuantumBlack). The state of AI in 2026: On the road to ROI. Dan Tinkoff, Lieven Van der Veken, Michael Chui, Tara Balakrishnan. August 25, 2026. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
Boston Consulting Group. The Formula for Agentic AI Value: The Applied AI Index 2026. September 30, 2026. https://www.bcg.com/publications/2026/the-formula-for-agentic-ai-value
KPMG International. New KPMG AI Pulse Survey: As AI maturity converges, leading organizations show what AI at scale requires (Global AI Pulse Q3 2026). September 24, 2026. https://kpmg.com/xx/en/media/press-releases/2026/09/new-kpmg-ai-pulse-survey-as-ai-maturity-converges-leading-organizations-show-what-ai-at-scale-requires.html
KPMG International. Growing adoption signals progress as cost visibility and accountability drive AI value (Global AI Pulse Q2 2026). June 24, 2026. https://kpmg.com/xx/en/media/press-releases/2026/06/growing-adoption-signals-progress-as-cost-visibility-and-accountability-drive-ai-value.html
KPMG LLP (US). AI Quarterly Pulse Survey, Q3 2026 (fieldwork July 24 to August 25, 2026). September 2026. https://kpmg.com/kpmg-us/content/dam/kpmg/pdf/2026/aipulsesurvey-q3.pdf
Deloitte AI Institute. The State of AI in the Enterprise 2026: The Untapped Edge (survey fielded August to September 2025). 2026. https://www.deloitte.com/global/en/what-we-do/capabilities/ai/research/state-of-ai-in-enterprise.html
Deloitte. The State of AI in the Enterprise 2026, US press release. 2026. https://www.deloitte.com/us/en/about/press-room/state-of-ai-report-2026.html
Deloitte. The State of AI in the Enterprise, Czech Republic and Slovakia page. 2026. https://www.deloitte.com/cz-sk/en/services/consulting/research/the-state-of-ai-in-the-enterprise.html
Gartner. Gartner Forecasts Worldwide AI Spending to Grow 49.5% in 2026. September 16, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-09-16-gartner-forecasts-worldwide-ai-spending-to-grow-49-point-5-percent-in-2026
Gartner. Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026. May 19, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-05-19-gartner-forecasts-worldwide-ai-spending-to-grow-47-percent-in-2026
Gartner. Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027. June 25, 2025. https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027
Microsoft. 2026 Work Trend Index Annual Report: Agents, human agency, and the opportunity for every organization. May 5, 2026. https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization
PwC. AI reshapes global labour market into two distinct paths, rewarding human skills: PwC 2026 Global AI Jobs Barometer. June 15, 2026. https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-jobs-barometer.html
IBM. IBM Study: One in Four Malicious Breaches are AI-Enabled, Costing Companies $6 Million on Average. July 29, 2026. https://newsroom.ibm.com/2026-07-29-ibm-study-one-in-four-malicious-breaches-are-ai-enabled,-costing-companies-6-million-on-average
IBM and Ponemon Institute. Cost of a Data Breach Report 2026. July 29, 2026. https://www.ibm.com/reports/data-breach
European Commission. AI Act, Shaping Europe’s digital future (page last updated 3 August 2026). 3 August 2026. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
European Commission. Guidelines on transparency obligations for providers and deployers of AI systems. 20 July 2026. https://digital-strategy.ec.europa.eu/en/library/guidelines-transparency-obligations-providers-and-deployers-ai-systems
NIST. AI Risk Management Framework (AI RMF 1.0 released January 26, 2023; page updated August 13, 2026). August 13, 2026. https://www.nist.gov/itl/ai-risk-management-framework
NIST. NIST AI 600-1, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. July 26, 2024. https://doi.org/10.6028/NIST.AI.600-1
