Artificial intelligence is no longer an innovation project. In 2026, AI has arrived in most organisations - at least on paper. Pilot projects, proofs of concept, and initial productive applications are widespread. Yet a clear divide is emerging in the market: some organisations - so-called Frontier companies - are achieving measurable, scalable business impact. Others continue to struggle with isolated use cases, unclear governance, and disappointing ROI. The key question is therefore no longer whether companies are using AI, but why so many AI projects never make the leap into productive operations.
The time for experimentation is over and organisations are shifting from asking what AI can do, to asking what value AI can deliver. CEOs now expect AI to deliver results that directly impact business KPIs such as growth, risk reduction, and time to market. IDC predicts that by 2026, 70% of EMEA1000 businesses will require clear proof of value before approving new AI investments, prioritising use cases that deliver meaningful impact beyond efficiency, driving growth and strengthening business resilience. Further, 51% of CXOs expect to achieve revenue growth through the application of AI in 2026, and 77% of CIOs surveyed stated that scaling AI is a priority for 2026. As a result, the pressure on decision makers to explain the ROI of AI is increasing.
Reasons Why AI Does Not Deliver ROI
There can be several reasons why integrating AI into organisations fails to generate ROI.
- AI is treated as a technology project rather than integrated into strategy: In many organisations, AI is still viewed as an isolated IT project. Pilot projects are launched within individual departments without a clear connection to overarching business strategies or measurable business objectives. As a result, technically functional solutions may emerge, but they often lack a direct link to concrete business value.
- Insufficient data maturity: Data is considered the foundation of every AI initiative - yet this foundation is often fragile. According to IDC, fewer than 4 out of 10 organisations are confident about their data-readiness for current AI priorities. Typical issues include data silos between departments, inconsistent data quality, and a lack of governance for data access.
- Lack of skills, organisational overload, and underestimated change management: Another bottleneck is people. IDC reports that most workers are acknowledging AI's impact on their roles, with 75% believing their roles will change. However, AI does not just change tools - it changes processes, roles, and decision structures. Without change management, typical symptoms emerge such as shadow AI outside official policies - meaning employees use AI tools outside the official framework, increasing security risks - and low adoption despite available solutions, or resistance due to uncertainty and fear of losing control.
- Difficult integration into existing systems and processes: Many AI PoCs work - but often only in isolation. Real business value only emerges when AI is integrated into existing business processes, enterprise data, and ways of working. This is where many organisations encounter structural barriers. Legacy systems, complex IT landscapes, and missing end-to-end processes make integration difficult and prevent AI solutions from moving from pilot projects into productive operations.
- AI remains stuck at the copilot stage: Many organisations primarily use AI as an assistance system - for example, for text generation, analytics, or simple automation of individual tasks. These so-called copilots can increase productivity but often remain limited to supportive functions. Decisions are still made entirely by humans, processes remain fragmented, and AI is not deeply embedded in operational workflows. This creates localised benefits but no structural transformation. The next step - agentic AI that independently analyses data, prepares recommendations, and initiates processes - is often not implemented. Companies therefore remain in a phase of assistance rather than true automation and decision support.
In summary, organisations that continue to launch isolated pilot projects, ignore data challenges, underestimate change management, and treat AI merely as an efficiency tool risk stagnating ROI results.
What Do Successful Companies Do Differently?
As stated in the IDC InfoBrief, sponsored by Microsoft , Frontier firms are seeing a return of 2.84 times on AI investments versus a return of 0.84 times for laggards. Overall, Frontier companies achieve up to four times better outcomes in growth, efficiency, and customer experience than other organisations. Further, 76% of Frontier firms describe their organisations’ overall adoption of GenAI as scaling (delivering both incremental and new value across the organisation) or realising (achieving consistent GenAI value across the organisation and in multiple business units) compared to 21% of laggards. So what do Frontier companies do differently to be so successful?
- AI is a business strategy - not an IT project: Frontier companies treat AI not as a technology initiative but as a strategic core capability. AI goals are directly linked to revenue growth, risk mitigation, time to market, and operational excellence. Management involvement is critical. AI is discussed at the executive level - not developed solely within IT teams.
- Data foundations instead of isolated dashboards: Many organisations invest in new platforms or dashboards without harmonising their underlying data architecture. Yet even the best visualisations create no value if the underlying data is fragmented, inconsistent, or difficult to access. Frontier companies take the opposite approach. They first invest in consistent data models, system integration, and clear data ownership. They also promote data literacy and decision literacy. Data literacy refers to the ability to understand, interpret, and use data effectively. Decision literacy goes one step further - it describes the ability to make sound decisions based on data. Dashboards still play an important role, but not as isolated reporting tools. Instead, they are part of an integrated decision architecture that makes data transparent, connects information from different systems, and enables leaders to understand developments in real time and make data-driven decisions. IDC predicts that by 2026, 30% of large organisations will evolve their hybrid clouds into integrated digital business stacks with federated data fabrics to realise business value, doubling AI production success through smooth data access and unified governance.
- Change management as a success factor and governance as a scaling accelerator: Technology changes processes. But AI transformation is not a technology project - it is an organisational project that must address people, governance, and culture equally. Frontier companies invest not only in systems but also in people. They provide structured training, emphasise transparent communication, and actively involve business units. According to IDC, over 75% of surveyed organisations rate transparency as very important. This figure jumps to 88% for Frontier firms. Governance therefore becomes a prerequisite for scaling - not an obstacle.
- AI is deeply integrated into systems and processes: Frontier companies understand that AI only creates real business value when it is seamlessly embedded in existing systems, data flows, and business processes. Instead of developing isolated applications, they integrate AI directly into operational workflows - for example in customer service processes, supply chains, IT operations, or management decision processes. AI accesses enterprise data, connects with core systems such as ERP, CRM, or content platforms, and supports employees exactly where decisions are made. This close integration makes it possible to translate insights from data directly into action and to continuously optimise processes. AI thus becomes not only an analytical tool but a core component of operational value creation.
- From copilot to agentic AI: Another difference lies in the use of agentic AI. While many organisations use AI as an assistance system, Frontier companies take a step further. They deploy AI agents that analyse data, evaluate scenarios, prepare recommendations, and trigger processes. This shifts the focus from pure analysis to active decision support. The future of AI lies not in reporting, but in action.
