If 74% of companies see AI returns, why have only 13% scaled as planned?
BearingPoint surveyed 1,050 senior leaders across Europe, the United States and China. The results suggest that AI can create value, but they measure executives' reports—not audited profit caused by AI.
The short version
- What happened
- BearingPoint surveyed 1,050 C-suite executives and senior leaders online in August 2026 across public and private organisations in Europe, the United States and China.
- Who it affects
- Workers may experience redesigned roles, retraining or reduced hiring even when a company reports value without full enterprise-scale deployment.
- What remains unproven
- The study is authored by a consultancy that advises organisations on technology transformation, and the public results are not an independent audit.
Keep in mind
The study is authored by a consultancy that advises organisations on technology transformation, and the public results are not an independent audit.
Respondents are senior leaders reporting on their own organisations; the public summary does not disclose response rates, full country bases or the questionnaire.
Reported revenue, cost and workforce effects are associations and perceptions rather than verified causal estimates.
Multi-source analysis · 2 linked sources. Primary analysed source date: 30 September 2026. Latest linked supporting source: 1 October 2026.
Read the full report · 6 min The evidence, methods, implications and limitations
The headline combines two very different tests
BearingPoint's new study asks whether organisations obtain measurable business value from AI and whether they can reproduce that value at enterprise scale. Seventy-four per cent of respondents whose organisations had implemented AI reported a measurable top-line or bottom-line effect. Around four in ten said they experienced both revenue growth and cost reduction. Those answers suggest that AI is no longer confined to demonstrations, but they do not show the size, duration or audited cause of every reported gain.
Scaling is the more demanding test. Only 13% said their AI initiatives had scaled completely in line with the original business case. Almost three quarters had adjusted the scope or achieved less scale than anticipated. A pilot can improve one workflow without surviving integration, security, compliance, data quality, procurement or workforce constraints across a whole organisation. The apparent contradiction—many reporting value but few scaling exactly as planned—is therefore plausible rather than proof that one figure is false.[1][2]
Who was surveyed, and what the denominator means
The methodology note says the study used online interviews with 1,050 C-suite executives and senior leaders from public and private organisations in Europe, the United States and China during August 2026. Every percentage describes responses from that leadership sample or a stated subgroup. It is not a census of companies, a survey of employees or a measurement of every AI project. The public summary does not provide a response rate, sampling frame, organisation-size distribution or country-by-country base sizes.
The 74% value result applies to organisations described as having implemented AI, not automatically to the full population of businesses. The 13% scaling figure likewise reflects respondents' interpretation of the original business case. One executive may define measurable value as a documented saving in a process; another may require a change in audited profit. Without the questionnaire and subgroup denominators, close comparisons between countries or sectors should be treated as directional.[1]
Value appears concentrated and difficult to convert into accounts
Nearly half of the surveyed organisations reported that AI's effect was below 4% of costs and below 2% of revenue. Reuters reports that 24% claimed AI-related cost savings of at least 10%, while only 4% reported revenue growth of the same magnitude. That asymmetry matters. Removing repetitive work can produce an operational saving relatively quickly; generating durable new revenue requires products, customer demand, distribution and a defensible advantage, not only a capable model.
BearingPoint classifies a group of 'AI Leaders' using its maturity framework. Forty-seven per cent of that group said they scaled completely as planned, compared with 6% of 'Implementers'. Seventy per cent of Leaders linked more than half of their AI projects to financial indicators, versus 34% of Implementers. These comparisons describe associations inside the consultancy's categories. They do not establish that adopting the recommended management practices caused higher returns; already capable and better-resourced organisations may be more likely to qualify as Leaders.[1][2]
The workforce result needs particularly careful reading
Sixty-two per cent of respondents reported AI-induced workforce overcapacity of at least 10% today, and 95% expected that level by 2030. Overcapacity is not the same as a recorded redundancy. It can mean that a team has more available hours after automation, that hiring plans change or that capacity is moved to other work. The report says only 48% currently embed strategic workforce planning in their AI roadmaps, making redeployment and training a central implementation issue rather than a side effect.
For workers, the practical question is what organisations do with released time. Management can redesign roles, invest in skills, move people into understaffed services, reduce recruitment or remove posts. Those choices affect whether productivity becomes better service, lower prices, higher margins or job loss. Employers should publish observed changes in hours, headcount, pay, quality and workload by function instead of treating an executive expectation as an employment forecast.[1][2]
What blocks scale—and what would change the assessment
BearingPoint says complex regulation and integration with legacy systems were the most frequently reported barriers. Reuters gives the attributed figures as about 40% naming legal regulation and 34% legacy integration. Fifty-four per cent of executives identified high-quality, trusted data as critical to scale. These constraints are connected: a system cannot be governed, tested or connected reliably when data ownership, definitions and access remain unclear.
Confidence in the value claim would rise with the full questionnaire, response and weighting information, country and sector bases, and longitudinal results tied to independently verified financial accounts. It would fall if reported gains disappear after integration, supervision, security and restructuring costs are included. The study supports a useful conclusion: many senior leaders perceive real AI value, while repeatable scaling remains uncommon. It does not prove that 74% of all companies have earned a positive return caused by AI.[1][2]