Artificial Intelligence A reality too good to be true?

Content

03

Introduction Methodology Key takeaways

Sustainability & sovereignty

21

Strategy & budget

06

AI technologies

25

Excepted benefits VS actual impact

Employee enablement

30

Risks & barriers 17

Get in touch

34

A reality too good to be true?

This year’s Global AI survey surfaces an image of AI adoption that seems almost too perfect. Ninety - nine percent of organizations report deploying AI; 70% place it at the heart of their business strategy; and IT budgets dedicated to AI have reached an average of 13%. Autonomous agents are already proliferating and reported benefits span efficiency, customer value, and employee wellbeing. On paper, AI appears less like an emerging tool than a permanent fixture of corporate strategy. Yet many of these achievements may be overstated.

Behind the numbers: a fragile reality

Beneath the headlines lies a more nuanced reality. Many initiatives still lack rigorous ROI measurement, projects often overrun time and budget, and only about 30% of users have meaningfully changed how they work. Use cases are multiplying but most organizations are still only beginning to grasp how AI can truly transform business models and core processes. In many cases, these deployments echo what MIT calls the “GenAI Divide” — high adoption but low transformation, where visible pilots fail to scale.

This year’s Wavestone AI Leader Survey, based on insights from 500 executives across Europe, North America, and Asia — CIOs, CDOs, CTOs, and CISOs — confirms both the centrality of AI and the paradox of a technology celebrated in boardrooms yet fragile in execution.

A market in transition

Where the most advanced organizations stand out by building resilient, industrial - grade foundations. They are investing in data architecture, governance layers, and adaptive platforms that transcend one -off experiments. Others remain focused on quick wins, chasing new pilots and agentic tools without rooting them in enterprise - wide processes. Sovereignty has become a buzzword — cited by 84% — but in too many cases it ends up as sovereignty washing, invoked rhetorically while performance, cost, or speed still dictate choices.

The real challenge ahead

At the same time, the rise of agentic organizations - humans working alongside intelligent agents - is redefining expectations from AI. This shift brings both opportunity and peril: orchestrating

multiple agents requires new operating models, accountability frameworks, and ethical safeguards. Many are rushing to adopt, but few are prepared to govern.

AI is now embedded in strategy, but maturity remains uneven. Ambition is abundant; execution is inconsistent. In 2025, the question will no longer be who experiments with AI, but who can prove value, scale responsibly, and anchor it in trust and purpose. Only then will AI become a true backbone of competitiveness, not just another wave of corporate FOMO.

Imène Kabouya Partner, Wavestone

About the survey

Wavestone surveyed 500 mainly technology (but also cyber, data & AI) leaders based in the USA, the UK, France, Germany, Singapore, and Hong Kong in mid 2025.

Respondent organizations’ size were:

23% large business (5000+ employees),
41% mid business (3000-4999 employees),
36% small business (1000-2499 employees).

62% of respondents were Board/C-level positions and 38% were Senior Management and Director positions.

IT, technology & telecoms Retail, distribution & transport Financial services Manufacturing & production Construction & property Government & international institutions

Consumer services

Business & professional services Energy, oil/gas & utilities

Life sciences

Leisure & entertainment

Key takeaways

No strategy without AI

70% of organizations already place AI at the heart of their business strategy

99% of respondents report that their organization has already deployed one or more AI solutions, and 70% consider it a significant part of their business strategy. With an average 13% of IT budgets dedicated to AI, most companies are betting on shortterm performance and long-term transformation. The challenge ahead is to scale responsibly, with stronger governance, compliance and value tracking.

2

AI maturity is not only about quickwins, but about futureproof foundations

From everyday tools to infrastructure, leading organizations are laying the foundations for scalable AI

AI adoption often starts with accessible, > embedded tools (adopted by 89%), but the most committed organizations go further, laying the invisible foundations needed to scale. Investments in infrastructure and governance platforms enable them to move beyond experimentation, building a resilient and industrialized architecture.

sv

Everyone wins with AI, on paper

While most organizations see tangible benefits, 46% do not yet have a structured ROI measurement framework.

99% of respondents report significant time > avings with AI, translating into business alue (32%), productivity (24%), and employee wellbeing (24%). Its impact is set to be enterprise-wide, from core business to support functions, with the biggest gains expected in IT & cybersecurity (92%) and customer-facing functions (90%). ROI is not formally measured, but only 22% of companies see it as a barrier to adopting AI.

As AI maturity grows, hurdles shift

Among the most AI-mature companies, the top hurdles are market complexity (27%) and regulation (25%). While others struggle mainly with talent shortages (38%), interoperability (32%) and budget (32%)

Barriers to AI adoption reflect organizational maturity. For the most mature organizations, the challenge is adopting responsibly through governance and compliance; for less mature players, it is overcoming silos and budget shortages. Success depends on breaking barriers, building momentum, and proving value quickly

Is sovereignty the new green?

84% of organizations integrate sovereignty into their strategy, while focus on environmental impact has fallen to 29% (vs. 46% last year)

While concern for AI’s environmental footprint was high in 2024, today sovereignty clearly dominates, driven by geopolitical tensions, regulatory debates, and dependency risks. This “sovereignty turn” reflects the new context — where cost, control, and strategic autonomy increasingly outweigh environmental considerations.

From bots to agents: the next challenge for AI leaders

76% of companies report deploying AI agents able to plan and act

The age of scripted bots is over: with 76% of organizations already deploying advanced agents and even 10% experimenting with full autonomy. These figures may appear surprisingly high, reflecting both ambition and early pilots. The critical hurdle, however, lies ahead: moving from experimentation to enterprise-scale deployment, with the governance, reliability, and integration required to make AI agents a true operational backbone.

#

Strategy & budget

Ghislain de Pierrefeu AI Expert - France

Our clients are rapidly shifting from an opportunistic approach — limited to a handful of focused use cases (assistants, RAG, etc.) — to a more strategic, business-driven approach, reimagining the processes most likely to deliver tangible ROI.

AI is now a board-level priority

The survey shows that AI has moved past the exploratory phase to become a board-level priority.

A clear majority (70%) already treat AI as a significant part of their business strategy, while another 20% include it without making it central. The laggards are both very few and under pressure, as even the 9% planning to adopt within a year will soon converge with the mainstream. This points to an environment where ignoring AI is no longer viable, and the real differentiator lies not in whether AI is adopted, but in how effectively it is scaled and embedded.

Regional patterns reveal contrasts in maturity. Companies in Hong Kong (78%) and Singapore (80%) show the strongest commitment to embedding AI strategically, while European firms — France (55%), UK (58%), and Germany (70%) — take a more measured approach. This relative caution is shaped by stricter regulatory frameworks such as the AI Act, but may also allow for more deliberate, targeted integration of AI into core processes.

Overall, the data makes one point clear: AI has become a non-negotiable element of corporate strategy, and leadership will depend on turning ambition into scaled, responsible adoption.

Which of the below best describes your organization’s strategic approach to AI?

AI has entered the strategic mainstream: 70% of companies now embed AI significantly in their business strategy.

C-suites put real money behind AI

Organizations are making substantial financial commitments to AI, underscoring its strategic importance. Over 65% of businesses now allocate at least 10% of their IT budget to AI, with 14% dedicating more than 20%. This is no longer experimentation; it’s a core investment.

This financial commitment scales directly with strategic intent: organizations that consider AI a "significant part of their business strategy" are far more likely to invest heavily, with 75% allocating over 10% of their IT budget to AI. Even those just planning for AI in the next 12 months show strong intent, with 40% ready to allocate 10-15%. These investments are directed toward cloud infrastructure, proprietary models, AI integration, and data governance: the invisible yet essential foundations of a long-term adoption strategy.

Larger enterprises (over 3,000 employees) are generally outspending their smaller peers, reinforcing that scale often drives deeper AI investment. Regionally, countries like the US and Singapore lead in high-budget allocations.

Sector-wise, IT & Telecoms, Financial Services, and Energy/Utilities are at the forefront of AI spending, consistently allocating higher percentages of their IT budgets. This reflects the intense data demands and transformational potential AI holds for these industries. Ultimately, these budget increases underscore AI's transition from an emerging technology to a foundational strategic asset for a wide array of organizations.

These commitments point to a potential upward trend in AI spending a trajectory we expect to accelerate, driven by strong C-level appetite.

What percentage of your organization’s IT budget is allocated to AI? (Average)

On average, 13% of the IT Budget is allocated to AI

AI confidence is rising — but is it overplayed?

The responses show a striking consensus: AI is seen as both transformative and already financially impactful, even though organizations vary in how advanced they are compared to peers. This marks a sharp shift from last year, when 86% reported a “wait and see” attitude, and reflects growing confidence not only in AI’s capabilities but also in organizations’ relative positioning.

Nearly all respondents (97%) agree that AI is a game-changer that will deeply disrupt work and business within the next two years, with almost half (47%) strongly agreeing.

A similarly high 96% agree that AI already has a positive impact on profits and financial planning, which strongly validates the case for sustained investment. This perception of value is widely shared across geographies and sectors.

Yet this optimism contrasts with what is really happening on the ground, where scaling remains uneven and maturity levels vary widely. These results may reveal more about the fear of being seen as laggards than about true enterprise-wide transformation.

At the same time, 66% acknowledge that concerns about AI-related risks are slowing down progress or scaling, showing that risk management is not peripheral but a significant brake on adoption speed. This reflects on the need for governance and compliance frameworks.

About organizations’ perceptions of AI and its impact

46% of respondents say they are behind most competitors, down from 75% in 2024

TAKEWAY

Strategy & budget

AI is no longer a peripheral technology—it has become a central pillar of corporate strategy for most organizations. With 70% already embedding AI significantly, the focus is shifting from adoption to effective scaling and integration. Regional differences persist, but the overall trend confirms that ignoring AI is no longer an option.

01

02

Mature organizations are no longer experimenting with AI—they are building the foundations for its sustained integration. Investments are increasingly directed toward invisible but critical enablers signaling a shift from short-term pilots to long-term transformation. This reflects a growing maturity, where AI is seen not just as a tool, but as a structural component of future business models.

AI is becoming increasingly central to our clients’ strategies. For most, the approach is still opportunistic and fragmented: identifying use cases that can quickly improve part of a process. This pragmatism is positive — it delivers fast, tangible results while progressively building AI skills and foundations “on the go.”

The most advanced players, however, are shifting to an AI-driven strategy. Rather than optimizing pieces of a process, they are rethinking entire workflows through AI — from KYC in banking to CMC in life sciences, supply chain in retail, and customer or employee service desks.

For these leaders, investments are rising sharply reflecting not only the magnitude of potential benefits but also the need to manage the heavy computational costs of largescale AI.

#

Expected benefits VS actual impact

Johann Chazelle AI Expert - France

Financial services are already ahead in adopting AI, with tangible impact in areas such as fraud detection, scoring and targeting, pricing, and risk management. Today, a new era is unfolding with generative and agentic AI — transforming KYC, customer experience, and claims management — as our clients reorganize to fundamentally reimagine these processes with AI at their core.

AI is primarily expected to drive efficiency, customer value and, innovation

Organizations expect AI to deliver on both fronts: 39% cite efficiency gains, while 37% look to enhance customer experience and 37% to improve products and services.

Workforce productivity (33%) and satisfaction (27%) are also in focus, while competitive advantage (30%) outweighs simple catch-up (24%), showing that firms aim to lead rather than follow. Direct financial outcomes rank slightly lower (29%), suggesting that revenue is seen as a consequence of efficiency and differentiation rather than an immediate target.

AI benefits materialize at different speeds. Quick wins often come from copilots or task automation, while transformationa l gains s uch as new business models take longer to emerge. Moreover, ROI hinges on data quality, process redesign, and adoption: where data is a barrier, revenue expectations drop to 24%; where change management is weak, efficiency gains fall to 31%.

The lesson is clear: organizations must balance short-term wins with long-term transformation, invest in strong enablers such as data and governance, and actively manage adoption if they want to turn pilots into enterprise-wide impact.

Since nearly all organizations plan to adopt AI within two years, competitive advantage will come not from adoption itself, but from the ability to scale it responsibly, embedding it deeply, and turning ambition into measurable impact

What benefits would, or is, your organization hoping to gain from adopting AI within the next 2 years?

Respondents were asked to select up to 3 options

80% expect at least one operational benefit — such as reduced inefficiencies, enhanced product development, or higher employee productivity. 73% expect at least one business-related benefit — such as improved customer satisfaction, competitive advantage, or revenue growth.

AI’s positive impact is expected everywhere in the enterprise

The data reveals a clear and consistent expectation: AI will have a positive impact on organizations within the next two years. From strategic decision-making to operational execution, AI is increasingly seen as a catalyst for efficiency, innovation, and performance moving beyond pilots to become embedded in core business processes.

IT and cybersecurity stand out as the most immediately affected domains. Automated threat detection, anomaly detection, and AI-driven code analysis are already being deployed, establishing new standards for risk management and resilience.

Productivity and knowledge management follow closely, with the rise of AI agents expected to fundamentally reshape how organizations capture, share, and leverage knowledge.

In customer-facing areas, such as marketing and sales, adoption is accelerating through use cases like hyper-personalization, automated content generation, and predictive analytics. This underlines that AI’s value extends beyond operational efficiency — it’s also about enhancing customer experience, deepening engagement, and unlocking new paths to growth

Sectoral differences add important nuance to this picture. Manufacturing companies, for instance, show particularly high expectations of AI’s impact on operations and supply chains — 92% anticipate a positive effect, compared with 84% in non-manufacturing sectors. This gap suggests that perceptions are closely linked to each industry’s level of exposure to AI use cases. It also raises the possibility that, in sectors where applications are less visible, AI’s potential may still be underestimated by non-specialists.

Looking at functional areas within your organization, where and to what extent do you expect AI to make an impact within the next 2 years?

IT & Cybersecurity lead (92%), while Legal, Risk & Compliance trail at 78% — yet all functions will feel the impact.

AI time savings fuel both business value and workforce transformation

Organizations are leveraging the time savings from AI in ways that balance hard business outcomes with employee-centered benefits.

The top answer, chosen by 32%, is driving greater business value, such as revenue growth or client satisfaction. This reflects the executive-level focus on ensuring AI translates into tangible impact on performance and competitiveness.

At the same time, nearly half of organizations emphasize the human side of productivity gains: 24% prioritize employee wellbeing (better work-life balance, reduced administrative workload), and another 24% focus on increasing output without adding headcount, underscoring efficiency and cost management. Together, these responses highlight AI’s dual use: enhancing employee experience while optimizing operations.

A smaller but significant share (19%) reinvest time savings in employee learning and upskilling, suggesting that some organizations view AI as a catalyst for capability building rather than mere efficiency.

Regional and organizational differences add nuance: Singapore leads in wellbeing focus (42%), Hong Kong emphasizes efficiency (36%), and the U.S. is most business-value oriented (35%). Mid-sized firms appear more agile in capturing value (39%), whereas larger organizations adopt a more balanced, multi-dimensional approach.

Overall, these findings point to a dual orientation: companies are using AI to create measurable business impact while simultaneously investing in people — signaling a broader shift from pure ROI metrics toward workforce transformation

Which of the following is the most valuable benefit your organization gains from the time saved through use of AI, if any?

99% of respondents reported time saving with AI

Measuring AI’s ROI is the new maturity milestone

Most organizations are moving beyond experimentation and treating AI as a measurable driver of operational and business performance.

Over half of respondents (54%) now have a formal process in place to track AI’s financial impact, showing that AI is increasingly integrated into standard performance management and ROI frameworks. This marks a clear maturity milestone: AI is no longer viewed merely as innovation, but as a contributor to enterprise value that must be measured with rigor.

Another 32% rely on informal expert assessments, indicating that while impact evaluation is happening, it often lacks standardized KPIs and depends heavily on qualitative judgment. Meanwhile, 10% assess value only case by case, and 4% have no structured approach at all — meaning roughly 15% remain in an ad hoc or immature measurement stage..

Overall, the data reveals a decisive shift: the real divide is no longer about adopting AI, but about proving its value. Leaders embed AI into governance and financial systems, while laggards risk blind spending and unverified ROI.

Do you know how to measure the value generated by AI at the enterprise level?

46% of organizations do not yet have a structured ROI measurement framework

TAKEWAY

Expected benefits VS actual impact

AI is now seen as a pervasive driver of value, with organizations identifying pockets of impact ranging from operational efficiency and customer experience to employee wellbeing and innovation.

01

Certain domains stand out as most exposed to disruption, notably IT and cybersecurity, knowledge management, and customer-facing functions such as marketing and sales.

02

The value is visible but not fully captured: while use cases multiply, ROI remains uneven, creating a divide between leaders embedding governance and metrics and those at risk of blind spending.

There is often a gap between vendors’ optimistic claims about AI’s transformative impact and the reality employees and customers experience. For many, the “arduous tasks” remain, budgets have barely shifted, and customer experience evolves slowly. So far, the most visible changes are largely confined to the private sphere (personal assistants, consumer apps) or to the marketing efforts of big tech.

Yet we are seeing clear acceleration among enterprises in their demand for valuedriven AI. This is reflected in the rise of “AI Control Towers” and a stronger involvement of business lines and finance teams in forecasting and measuring ROI.

At Wavestone, our value-based tools and methodologies highlight a striking disparity in where gains are realized. Certified benefits still come mostly from traditional AI use cases (scoring, targeting, fraud detection, predictive maintenance), but the share of GenAI-driven value — through copilots, coding assistants, field chatbots, service desk automation, and multi-agent systems — is growing rapidly.

#

Risks & barriers

Cecilia Edwards AI Expert - USA

Industry has kept up with AI experiments but struggles to scale them. Risks around security and compliance, combined with the rigidity of core systems (ERP, PLM, WMS, TMS…), hold back seamless integration of AI into day-to-day operations. Generative AI opens new possibilities by unlocking value from rich unstructured data, but it will only be a game changer if built on open architectures, robust governance, and effective change management. For the most advanced players, these are now the real challenges.

AI risks are recognized but unevenly managed

AI comes with a complex and critical risk landscape.

Nearly all organizations (94%) acknowledge substantial risks, reflecting a growing maturity in recognizing AI’s potential downsides.

Security and data privacy (43%), regulatory and reputational exposure (38%), and ethical concerns (36%) dominate the agenda—placing compliance, cybersecurity, and trust firmly at the forefront of enterprise risk management. Highly regulated sectors like Financial Services and Energy feel these pressures most acutely, given the sensitivity of their data and operations.

Around 29% worry about dependency on third parties or diminished employee purpose. These figures underscore that AI risks extend beyond technology into organizational culture and workforce dynamics.

By contrast, strategic autonomy (19%) and environmental impact (14%) receive less emphasis, suggesting that long-term risks remain underweighted compared to immediate concerns.

Risk perception also shifts with maturity: early adopters worry more about reputational damage and internal resistance, while advanced players focus on governance and compliance. Managing AI risks requires more than technical fixes—it demands robust governance, ethical safeguards, and human-centric planning.

What risks, if any, do you associate with adopting AI at your organization?

94% of respondents acknowledge the risks associated with AI adoption

AI obstacles evolve along the maturity curve

Most organizations (92%) face barriers in their AI journey, with talent shortages (34%) leading the list. Structural issues follow closely, with fragmented systems and siloed data (28%) and budget limitations or unclear priorities (27%) slowing momentum. At the same time, organizations struggle with a fast-evolving market (27%), regulatory uncertainty (24%), and unclear ROI measurement (22%), highlighting the difficulty of making confident investment decisions. Human and cultural barriers are equally significant: resistance to change (26%) and leadership misalignment (23%) underscore that adoption is as much organizational as it is technical.

Yet responses vary widely depending on maturity. For advanced organizations, the challenge is no longer whether to adopt AI but how to do so responsibly. Their priorities are navigating growing market complexity (28%) and regulatory alignment (25%). This requires strong governance, anticipation of the upcoming AI Act, and investment in specialized expertise—legal, ethical, and technical. Less mature organizations, by contrast, face more fundamental roadblocks: talent shortages (38%), technological silos (32%), lack of prioritization (30%) and resistance to change (28%). For them, the real challenge is cultural and organizational: breaking down silos, raising awareness, and delivering early, visible results to bring teams on board.

What barriers, if any, is your organization facing in its adoption of AI?

Last year, the leading challenge was “data reliability and quality” (31%).
This year, it has shifted to “skilled staff and internal AI expertise” (34%).

TAKEWAY

Risks & barriers

01

Nearly all organizations acknowledge substantial AI risks, with security, compliance, and trust dominating the agenda. Risk perception evolves with maturity: beginners worry about resistance and reputation, while advanced players focus on governance and regulatory alignment. Managing these risks demands more than technical fixes —it requires robust governance, ethical safeguards, and human -centric planning

02

Barriers to adoption are nearly universal, but their nature shifts with maturity. Less mature organizations struggle with talent shortages, silos, and cultural resistance, making change management and early wins critical. Advanced players, by contrast, face the challenge of scaling responsibly —navigating market complexity, regulatory requirements, and the need for specialized expertise.

The risks are now well recognized by most of our clients — from cyber threats to bias and hallucinations — but remain poorly addressed in practice.

MLOps and LLMOps frameworks, along with their control mechanisms (monitoring, drift detection, reproducibility, grounding, bias correction, red -teaming, etc.), are advancing among the most mature players. Yet they are still less prioritized — especially financially — than flashy GenAI POCs, which are too often launched without a solid foundation. At Wavestone, we have developed AI -Sec -Ops frameworks that significantly mitigate these risks while ensuring compliance with regulations.

Barriers, however, are less clearly understood. Today, the focus is still too often on technical obstacles (skills, data silos, technology choices), whereas in reality, scaling AI mainly runs into governance and change management challenges. How can we prevent uncontrolled proliferation? How can we reshape jobs and processes to fully capture AI’s potential? And how do we govern environments where multiple AIs — workplace tools, vendor solutions, open -source systems — coexist and interact?

#

Sustainability & sovereignty

Uta Niendorf AI Expert – Germany

…Digital sovereignty and sustainability are no longer optional. They are the new table stakes, and choosing the right AI is as much about resilience and responsibility as it is about performance.

Sustainability slips back: AI’s green promise still uneven

Last year, 46% of respondents said environmental impact was considered in all GenAI projects, and 40% said it applied to some. Even allowing for the fact that GenAI projects are among the most energy-intensive and therefore more closely scrutinized, this year’s results suggest a step back in environmental considerations (29%).

Still, 91% of organizations do measure AI’s environmental footprint in some way, showing growing awareness but uneven integration of sustainability strategies. Strategic adopters are far more likely to measure systematically pointing to a clear link between AI maturity and environmental responsibility. Yet practices remain fragmented: while 29% report enterprise-wide measurement, many adopt selective or inconsistent approaches.

This unevenness reflects both the complexity of defining standardized methodologies and varying organizational readiness. Regulatory and eputational pressures are pushing companies to act, but gaps persist, especially where governance and strategic alignment are weak. Encouragingly, organizations that prioritize digital sovereignty and responsible AI also tend to lead in environmental measurement, indicating a more holistic approach. As sustainability becomes a key marker of AI governance maturity, those investing early in structured measurement frameworks will be best positioned to meet rising expectations and regulatory demands.

To what extent does your organization measure the environmental impact of AI across use cases?

91% of respondents measure AI’s environmental impact in some way, but only 29% systematically assess it across all use cases

Digital sovereignty rises in AI strategy, yet trade-offs persist

The results highlight that digital sovereignty has moved firmly into the mainstream of AI strategy, though the intensity of commitment varies.

Half of respondents (50%) consider it important and factor it into decisions when feasible, while more than a third (34%) elevate it to a strategic priority, actively working to reduce dependence on foreign technologies and jurisdictions. Taken together, this means that 84% of organizations explicitly integrate sovereignty into their AI strategies. By contrast, 15% acknowledge sovereignty concerns but ultimately prioritize performance, cost, or innovation—illustrating a pragmatic trade-off approach.

This distribution reflects how geopolitical tensions and regulatory debates—ranging from the AI Act to data localization laws and supply chain risks—are pushing enterprises to weave sovereignty into their strategic frameworks. Yet for many, it remains conditional (“when feasible”) rather than absolute, underscoring that competitiveness and performance continue to dominate decision-making.

In short, sovereignty is broadly recognized and influential but still weighed against practical imperatives such as cost efficiency and access to cutting-edge capabilities. The divide between those treating it as a non-negotiable priority and those relegating it to a secondary concern mirrors global geopolitical rivalries that are increasingly shaping the world.

In light of rising geopolitical tensions and the shifting global economic order, to what extent is digital sovereignty a consideration in your organization’s AI strategy?

84% of respondents express strong concern about sovereignty

TAKEWAY

Sustainability & sovereignty

While 91% of organizations report measuring AI’s environmental impact, only 29% do so systematically across all use cases. This reflects uneven maturity in sustainability practices. Even if there is a step back from the previous year, strategic AI adopters lead the way, linking governance with environmental responsibility.

01

02

Digital sovereignty is now a mainstream concern, with 84% of organizations factoring it into their AI strategies. However, only 34% treat it as a top priority, while others balance it against performance and cost. Geopolitical and regulatory pressures are driving this shift. The divide reflects broader tensions in global AI ecosystems.

Sustainability remains a critical issue for large enterprises — yet often undermined tone’s visionby the opacity of AI vendors. Without transparent data from LLM providers, calculating a reliable climate footprint is nearly impossible.

At Wavestone, we advocate a frugal approach: choosing the right AI for each use case. In lots of enterprise use -cases, the biggest LLMs bring little to no performance advantage over lighter, more specialized models. And once all control functions (LLMOps) are added, the cost of an enterprise use case — both financial and carbon — can multiply by a factor of 2 to 5. Using compact SLMs for hyperspecialized functions (e.g., “LLM -as - a -judge”) often keeps this closer to 2x rather than 5x.

Sovereignty, meanwhile, remains poorly understood. Too often it is reduced to the nationality of the LLM vendor, while the rest of the value chain is ignored: rare earths and metals, semiconductors, hyperscalers for middleware and orchestration. On these layers, dependence on American or Chinese players remains far harder to overcome.

#

AI technologies

Stéphan Mir AI Expert - France

Low-hanging fruits — such as AI features embedded into productivity suites or enterprise software — are easy to deploy, but insufficient to deeply transform in-house, domainspecific processes. To address these, our clients leverage a wide variety of models, reflecting the commoditization of LLMs, and select those best suited to their agentic workflow.

One foot in, one foot out: the hybrid reality of AI partnerships

The survey reveals a fragmented landscape in how companies structure their relationships with AI providers. The largest share, 31%, adopt a hybrid model that blends internal development with external partnerships—a pragmatic way to move fast while maintaining some control.

Another notable pattern is the reliance on a single strategic partner, chosen by 21% of respondents, most often a hyperscaler such as Microsoft, Google, or AWS. A further 10% spread usage across several vendors but still concentrate on one dominant provider, meaning that over a third of organizations accept a significant level of dependency.

By contrast, 18% pursue a truly diversified multi-vendor strategy to mitigate lock-in risks.

Independent approaches remain in the minority. Only 10% primarily rely on in-house capabilities, while 8% give priority to open-source solutions, despite their rising prominence in public debate.

Overall, the findings highlight a delicate balance: pragmatism drives partnerships, major providers anchor much of the market, yet a parallel movement is emerging to preserve independence and sovereignty in AI.

Which of the following best describes your organization’s approach to working with AI technology providers?

Hybrid approach (includes both external partners and internal development)

Primarily on a single strategic partner (Microsoft, Google, AWS, etc.)

Diversified multi-vendor strategy to reduce dependency on any single provider

Build and maintain in-house AI capabilities

Several providers but one plays a dominant role

Open-source models and platforms

No AI solution implemented yet

68% of respondents seek to avoid excessive dependency on a single provider

From plug-and-play to built-to-last: the maturity curve of enterprise AI

Enterprise AI adoption is driven more by integrated and accessible solutions than by bespoke developments.

The most common entry point: AI features embedded in productivity suites such as Microsoft 365 Copilot or Google Gemini, used by 63% of respondents.

Half also rely on AI capabilities within enterprise applications (CRM, ERP, HR, cybersecurity) and public tools like ChatGPT or Gemini (53%). This reveals a dual dynamic — top-down through embedded business software, and bottom-up through employee experimentation. Beyond convenience, structural investments are gaining traction: 43% have upgraded their data and cloud foundations to scale AI. Meanwhile, 39% develop proprietary models for internal use, 31% build external-facing chatbots, and 33% invest in dedicated AI platforms for governance and risk control.

For most organizations, AI adoption typically begins with plug-andplay tools, expands through infrastructure modernization, and matures with governance platforms and in-house model development — illustrating the shift from experimentation to enterprise-wide integration.

Which of the following AI solutions has your organization implemented or accepted for use so far?

Respondents selected an average of three options.

The rise of AI agents – and the urgent need for governance

Organizations are rapidly advancing toward more sophisticated AI agents. Only 3% still rely on basic chatbots (Level 0), while the vast majority have moved beyond scripted responses: 35% now operate Level 2 agents that dynamically select tools and actions, and 31% use Level-3 agents capable of generating and executing multi-step plans with limited human input. Remarkably, 10% are already experimenting with Level-4 agents — systems that autonomously handle complex, highvalue tasks.

This acceleration reflects both the promise and the hype of agentic AI. The lure of high-value automation — and a growing FOMO effect — has pushed companies to fast-track adoption. Last year, many felt late to GenAI; this year, they’re racing to test autonomy.

As agents proliferate and gain independence, the challenge shifts: from building them to governing them — ensuring accountability, strategic alignment, and responsible integration within enterprise frameworks.

To what extent has your organization adopted or planned for the use of AI agents?

Level 0: We have implemented conversational agents (e.g., chatbots or virtual assistants that respond to user input with predefined answers)

Level 1: We have implemented AI agents that can follow predefined deterministic workflows or sequences to complete tasks

Level 2: We have implemented AI agents that can dynamically select tools or actions based on real-time context or input

Level 3: We have implemented AI agents that are capable of generating and executing multi-step plans with minimal human input

Level 4: We have implemented AI agents that can operate independently on complex, high-value tasks, making decisions and taking actions without human oversight

We have not yet implemented AI agents, but they are part of our future strategy

We have not implemented AI agents and have no plans to do so

Most respondents use level 2 (35%) or level 3 (31%) agents.

TAKEWAY

AI technologies

01

The survey reveals a fragmented AI ecosystem. Most organizations pursue hybrid strategies that mix internal development with external partnerships, balancing speed and control. While over a third remain tied to a dominant hyperscaler, 68% actively seek to avoid excessive dependency, signaling growing awareness of vendor lock -in risks. Independent approaches are still rare, but the pursuit of greater autonomy is clearly gaining momentum.

02

AI adoption starts with plug -and -play tools embedded in everyday software, often boosted by employee experimentation. As organizations scale, focus shifts to solid foundations —data, cloud, governance, and custom models —for enterprise -wide integration. Meanwhile, AI agents are spreading fast, pushing autonomy further. The challenge now is not building, but governing —ensuring trust, compliance, and alignment in an increasingly complex landscape.

The fragmented survey responses mirror what we observe in the field: it would be misleading to claim that a single “right” AI or agentic target architecture exists today. The market is overflowing with options — hyperscalers, ERP/CRM/SD vendors, agentic platforms, open -source frameworks, low -code data platforms embedding AI, and specialized governance solutions. Still, a few trends are beginning to stabilize:

Multi-LLM strategies to avoid dependence on inherently unstable publishers, whether financially or in terms of performance.
MLOps and LLMOps practices to ensure better control and scalability.
Cloud/on-premise hybridization in sensitive industries where data sovereignty and security are paramount.

Beyond these foundations, many organizations remain in exploration mode, testing and building skills across multiple solutions, particularly in the fast -moving agentic space (open source, vendor -led, hybrid).

#

Employee enablement

Richard Graham AI Expert – UK

AI transformation is not won in the innovation lab. It is won in the workplace. Success depends on equipping every employee with the skills, confidence, and guardrails to thrive alongside intelligent systems.

From training to transformation — building workforce confidence in AI

AI workforce enablement is now a near-universal. Most organizations are adopting a multi-pronged approach, combining role-specific training (52%), transparent communication (51%), and formal change management (46%) to ensure adoption is both contextual and trusted. On average, respondents selected three initiatives, underscoring that no single lever is sufficient—true enablement requires a mix of technical upskilling, trust-building, and organizational transformation.

Leadership engagement is emerging as a key differentiator: AImature organizations are far more likely to embed AI-related objectives into leadership goals (51% vs. 43% on average), reinforcing the link between executive commitment and accelerated adoption. External partners also play a strong role (41%), helping organizations compensate for scarce internal expertise.

Companies are investing simultaneously in top-down leadership alignment, bottom-up experimentation, and external support to build both capability and confidence : full ROI will only be realized once AI is embraced across the workforce, not just piloted in isolated pockets.

How is your organization supporting employees to build confidence and capability with AI?

Respondents selected an average of three options

AI is reshaping work but the transformation remains partial

The data shows that AI is already reshaping work practices, but the transformation remains partial.

On average, respondents estimate that 30% of their target users have meaningfully changed the way they work due to their organization’s main AI initiatives. This confirms that AI adoption is no longer marginal, though it has yet to reach a majority of the workforce.

The distribution is telling: one-third of organizations place adoption in the 30–40% range, which appears to be the current ceiling. Only 7% report that more than half of their users have significantly changed their practices, highlighting that broad-based adoption is still rare. At the other end, just 5% are still in the very early stages, with fewer than 10% of users impacted.

Overall, these results reflect an intermediate stage of maturity: AI is delivering real change but mostly within specific populations or use cases. Breaking through to majority adoption will require stronger investment in change management, targeted training, and seamless process integration, to move from experimentation to enterprise-wide transformation.

In your view, what percentage of target users have meaningfully changed how they work due to your organization’s main AI initiatives?

TAKEWAY

Employee enablement

01

AI is already reshaping work, but only partially: on average, 30% of target users have meaningfully changed the way they work. Adoption remains concentrated, with most organizations capped at 30 –40% and only 7% reaching more than half their workforce. This points to an intermediate stage of maturity, where real change is visible but not yet at scale. Reaching majority adoption will require stronger investment in change management and process integration.

02

This is where workforce enablement comes in. It is now nearly universal, with organizations combining training, communication, and change management to build both trust and skills. Leadership engagement and external partners are emerging as key differentiators.

We’re still far from a world where every employee governs autonomous agents. To get there, organizations must first build AI maturity and literacy: while many already experiment with AI, these efforts remain fragmented and disconnected from real work.

The priority is to design concrete, differentiated learning experiences that quickly move from generic use cases to high -value, role -specific applications — with workflows purpose -built for AI rather than retrofitted. Each solution should serve a clear outcome, with embedded guardrails to ensure trust and adoption.

On this foundation, organizations can evolve step by step toward a truly AI - driven model — where humans and machines collaborate seamlessly — provided leaders lead by example and HR ensures continuous learning, coherence, and equitable access to new capabilities, building the culture glue that sustains transformation.

Conclusion & takeways

Key takeaways from 2025 Global AI survey

“Our survey highlights a major strategic turning point: while nearly all organizations acknowledge AI’s transformative potential, success now depends on their ability to translate ambitious strategy into disciplined execution, managing both risks and competitive pressure. Yet discipline alone will not be enough — competitive advantage will belong to those who can rethink their organizational models and business processes in a world shaped by intelligent systems”

FOR IT LEADERS & CTOS

Functional Leadership: IT/Cybersecurity shows highest confidence (50% expect significant impact) - position as internal AI champions Implementation Focus: Productivity & knowledge management (48%) and marketing/sales (46%) are prime areas for quick wins Budget Alignment: Ensure AI investments match strategic priorities - those with higher strategic focus allocate more budget Risk vs. Progress: Balance innovation speed with risk mitigation - don't let perfect be enemy of good

FOR C-LEVEL EXECUTIVES

Strategic Imperative: 70% of tech leaders already have AI as significant
business strategy - this is not optional
Investment Reality: 13% average IT budget allocation shows serious
commitment, but strategic priority directly correlates with investment levels
The AI Paradox: While 95% believe AI is a game-changer with positive profit
impact, 66% admit risks are slowing progress - need balanced risk
management
Competitive Positioning: 45% feel behind competitors - urgent need for
differentiated AI strategy

Authors/Get in touch

Key contacts by region

Report author

Report main contributors

Germany Uta Niendorf Partner

UK
Richard Graham
Partner

Imène Kabouya Partner, France

Ghislain de Pierrefeu Partner, France

Julien Floch Associate Partner, France

US
Cecilia Edwards
Partner

Asia

Romain Gagliardi Associate Partner

2025全球人工智能调查报告：AI应用悖论.pdf免费PDF下载