news & insights
back to all news
This guide is aimed at executives, HR Directors, CFOs and CIOs seeking to understand the actual state of artificial intelligence in business in 2026, beyond the media hype: which use cases are mature, what signals indicate that it is time to act, how to choose a solution, and which regulatory framework applies.
It provides a clear definition of AI terminology, an overview of use cases across the main business functions, a framework for assessing your own level of maturity, the criteria that really matter when selecting a solution, an up-to-date regulatory overview of the GDPR and the AI Act, as well as the most common mistakes observed in the field.
The main factor behind AI project failure is not the technology chosen, but data quality and change management (Deloitte, Global Human Capital Trends, 2025).
Artificial intelligence in business today encompasses a wide range of technologies, often grouped together under the same generic term. Establishing a shared vocabulary is essential before going any further.
It is important to distinguish consumer AI, such as conversational assistants used individually by employees, from integrated business AI built directly into enterprise management tools: HRIS, financial ERP, CRM, and expense management solutions. The latter is connected to the organisation's actual data, governed by defined policies, and generally subject to structured human oversight.
This distinction largely explains the gap between individual employee use of AI and its official adoption by organisations, a phenomenon often referred to in research as shadow AI: employees use consumer AI tools in their day-to-day work without the company having officially approved their use or established safeguards for the associated risks, particularly regarding data confidentiality.
AI adoption has reached a new milestone. According to McKinsey (State of AI, 2025), 88% of organisations now use AI in at least one business function, and 79% have adopted generative AI. However, only around 7% have deployed it at enterprise scale. This gap between isolated adoption and enterprise-wide deployment defines the strategic challenge for the next two years.
Three dynamics explain this acceleration. First, competitive pressure: organisations that industrialise AI are gaining a measurable advantage over their competitors in terms of productivity and processing costs. Second, pressure on skills: in a context of talent shortages in certain roles, AI makes it possible to redeploy teams' time towards higher-value tasks. Finally, regulatory pressure is introducing a strict governance framework, particularly for higher-risk use cases.
The rise of agentic AI is also changing the nature of the subject. These are no longer simply tools providing occasional assistance, but systems capable of executing tasks end to end. According to McKinsey, 62% of organisations are already experimenting with AI agents, and around 31% have at least one running in production, with banking and insurance among the most advanced sectors.
Key figures to remember
|
AI use cases vary significantly from one function to another, both in terms of maturity and the nature of the benefits delivered. Here is a concise overview by major business function.
Predictive recruitment, employee chatbots, personalised training, talent management, and people analytics are among the most mature AI use cases in HR. For a detailed overview with concrete examples: 6 concrete AI use cases in HR.
59% of CFOs were already using AI in their department in 2025, compared with 37% two years earlier (Daf-Mag, 2026). The most advanced use cases include automated bank reconciliation (85% of entries processed without human intervention, with the error rate reduced to below 0.3%), automated invoice data extraction (92–98% accuracy), accounting anomaly detection, and accounts payable automation. Expense management and T&E compliance automation, already widely adopted by companies using tools such as SAP Concur, are a natural extension.
Before choosing a tool, it is essential to understand where your organisation really stands. Market frameworks generally converge around five stages of maturity.
| Stage | What characterises it |
|---|---|
| Starting point | Individual, unstructured use (shadow AI), with no formal strategy. |
| Experimentation | A few pilots launched, limited scope, with results not yet scaled. |
| Formalisation | Use cases prioritised, with data governance being structured. |
| Industrialisation | Deployment across one or more functions, with ROI measured. |
| Transformation | AI is natively integrated into processes and decision-making culture. |
→ Assess your AI maturity in 5 minutes with our questionnaire
The market for AI solutions is crowded and evolving rapidly. Rather than comparing features that change every quarter, it is more useful to rely on cross-functional criteria that remain stable over time.
| Criterion | What to check |
|---|---|
| Interoperability | Does the solution integrate natively with your existing IT environment (HRIS, ERP, CRM)? |
| Data security | Where is the data hosted? Is it used to train third-party models? |
| Explainability | Can the provider document how a recommendation or decision is generated? |
| Regulatory compliance | How does the solution address the GDPR and the AI Act? |
| Change management support | Does the provider or integrator offer a training and adoption plan? |
| Measurable ROI | Are success metrics defined before deployment? |
The GDPR governs all processing of personal data, including when it is carried out by an AI system. The standard requirements apply in full: a lawful basis for processing, informing data subjects, defined retention periods, and conducting a Data Protection Impact Assessment (DPIA) for high-risk processing.
The European AI Act entered into force on 1 August 2024, with its obligations applying progressively. The ban on certain uses and obligations relating to general-purpose AI models have already applied since 2025.
The timeline for high-risk systems, which notably covers AI used in recruitment and employee assessment, was amended by the Digital Omnibus Regulation (EU) 2026/1744, which entered into force on 27 July 2026. The deadline originally set for 2 August 2026 has been postponed to 2 December 2027. This extension gives organisations additional time to achieve compliance, but does not remove the need to anticipate requirements: transparency obligations (Article 50), particularly informing users when they interact with a conversational AI system, remain subject to progressive application over the period.
Good to know The AI Act does not prohibit the use of AI in business. It requires high-risk uses to be documented, explainable, and subject to human oversight. The postponement of the deadline to 2027 should not be interpreted as a relaxation of vigilance: compliance takes time, and organisations that anticipate it now will avoid rushed overhauls at the end of the deadline. |
Arago supports its clients in unlocking the value of their data and integrating AI at the heart of their HR and Finance systems, through a complete value chain: native data extraction, workflow orchestration, value creation through Business Intelligence, adoption of new use cases, and more.
This approach always starts with strategic scoping before technical implementation. Before any technical project, our experts assess your organisation's data and AI maturity: evaluating the quality and governance of your existing data, mapping your ecosystem, identifying the use cases with the highest ROI, and more. This assessment is essential to any Data and AI project, providing the foundation for building an architecture or deploying agents on a solid, thoroughly assessed foundation. Beyond standard integration, our teams also develop bespoke solutions to help you increase your Data & AI maturity.
We combine strategy and execution, supported by dedicated experts and more than 200 certified consultants.
Predictive AI analyses historical data to anticipate an outcome. Generative AI produces content such as text, images, or code. Agentic AI goes further: it executes tasks autonomously by chaining together multiple actions, without requiring human validation at every step. These three categories are often combined within a single solution.
Start with an honest maturity assessment: mapping existing use cases (including unofficial ones), evaluating the quality of available data, and identifying an initial high-volume, low-risk use case. Starting directly with a complex project without this step is one of the main causes of failure.
The AI Act classifies AI use cases according to their level of risk and imposes transparency, documentation, and human oversight requirements for high-risk uses, particularly in recruitment and HR assessment. The compliance deadline for these systems has been postponed to 2 December 2027 by the Digital Omnibus Regulation, which entered into force in July 2026.
ROI should be defined before deployment, not afterwards: time saved on specific tasks, reduced error rates, or improved processing times. Organisations that fail to define these metrics upfront are also the ones that struggle most to justify scaling their AI initiatives.
AI is more likely to transform tasks than entire jobs. It automates repetitive, low-value-added activities within a role, freeing up time for work that relies on judgement, relationships, and decision-making. This transformation requires skills development and support, not just technical deployment.
Would you like to structure your AI strategy in HR or Finance?
