Artificial intelligence in business: productivity and governance in 2026

AI adoption in business: where do things stand in 2026?

In just a few years, artificial intelligence has moved out of the lab and into everyday business use. Beyond the hype, it’s increasingly seen as a real driver of innovation and competitiveness. It helps companies streamline operations, improve performance, and develop new services.

If adoption is accelerating, it’s also because AI is becoming easier to access. Costs are coming down, ready-to-use tools are widely available, and competitive pressure is pushing companies to move faster. Many are now shifting from testing use cases to deploying them at scale.

This marks a new phase in digital transformation. Companies are no longer just experimenting with AI. They’re starting to integrate it into core business processes.

According to the Panorama 2026 de l’IA en entreprise by Bpifrance, 2026 represents a clear turning point. Pilot projects are gradually moving into large-scale deployment, confirming that AI adoption is accelerating across industries.

AI adoption in business in 2026

A global, cross-sector shift

AI adoption is now global, but it’s not moving at the same pace everywhere. Some regions are further ahead, even if common drivers (lower costs, easier tools, and growing demand) are pushing adoption forward across the board. In Europe, several countries are leading the way. France (44%), Spain (41.8%), and Ireland (44.6%) were all ahead of the United States (28.3%) and China (16.3%) in generative AI adoption at the end of 2025, according to the Microsoft AI Diffusion Report.

In France, around 10% of companies with more than ten employees are already using AI (INSEE). Adoption is particularly strong in certain functions. Marketing teams, for example, are widely using AI for summarization and translation (59%), while risk and compliance teams rely on it for anomaly detection (44%), according to KPMG’s Trends of AI 2026.

This growth is closely linked to the rise of generative AI. These tools can produce content, analyze documents, and synthesize information in seconds, making them easy to integrate into daily workflows.

Momentum is also building in Africa and the Middle East, where AI is becoming a driver of both digital transformation and economic growth.

This creates a strong ripple effect. As soon as companies start seeing quick wins, whether in automation, analytics, or content creation, they tend to accelerate their own AI initiatives. At this point, adopting AI is no longer a technical experiment. It’s a strategic decision.

As KPMG puts it, the question for companies is no longer whether to adopt AI, but how to integrate it sustainably, with the right governance in place. 

AI: what productivity gains are companies seeing?

Artificial intelligence is not just about automating tasks. It’s changing how companies work, how they process information, and how they make decisions.

According to Deloitte, 66% of organizations are already seeing productivity and efficiency gains from AI. These gains vary depending on the function, whether it’s marketing, HR, supply chain, or IT and cybersecurity.

In some roles, productivity can even double when AI tools are fully integrated into workflows. 45% of companies using generative AI say it has at least doubled employee productivity (Google Cloud/NRG 2024).

Positive ROI from generative AI

Since generative AI began scaling in 2022, productivity growth has accelerated significantly in the sectors most exposed to AI, such as financial services and software. According to PwC’s 2025 Global AI Jobs Barometer, growth in these sectors has almost quadrupled, from 7% (2018–2022) to 27% (2018–2024), roughly three times faster than in less exposed sectors.

The financial impact is also clear. In France, an analysis of 200 AI projects conducted between 2022 and 2025 shows a median ROI of 159% for SMEs, with payback reached in just 6.7 months on average. For mid-sized companies, it’s closer to 10 months (Stema Partners).

Companies are tracking these results closely. According to Deloitte, 86% of French organizations regularly measure the impact of their AI projects. Many see AI as a growth driver: 74% expect it to generate revenue growth, even if only 20% say they are achieving that today.

That said, results are not guaranteed. Gartner estimates that nearly 40% of advanced AI projects could be abandoned by 2027 due to unclear objectives or poorly defined processes. In other words, the benefits are real, but they only last when AI is rolled out with a clear strategy and strong governance. It also requires bringing teams along.

Concerns around trust, skills gaps, and resistance to change remain real barriers. Employee adoption is still one of the key success factors. 

It also requires bringing teams along. Concerns around trust, skills gaps, and resistance to change remain real barriers. Employee adoption is still one of the key success factors.

• In healthcare: AI is becoming a real support tool for medical teams. Some systems can quickly detect anomalies in lab results or medical images, helping identify certain conditions earlier and improve care organization. This helps improve diagnosis, secure treatments, and accelerate drug development.
   
• In the public sector: AI helps make services more efficient and more accessible. Administrations can anticipate needs, guide users more quickly to the right information, and better manage flows in critical services such as emergency care or support platforms. It is not just about automation. It is about improving the experience and freeing up time for teams.
   
• In companies: the benefits often appear within the first few weeks. Teams work more efficiently when analyzing data, processing requests, and coordinating actions. AI also supports operational innovation by making it easier to test ideas, simulate scenarios, and improve processes.
  Beyond immediate productivity gains, AI ROI now also includes service quality, operational security, and innovation in business practices. It is not just about financial return. It also means better organization, safer services, and teams that can focus on higher-value work.

The most effective AI use cases in business

In 2026, AI is embedded in core business processes. Use cases are expanding across all functions, from cybersecurity and training to supply chain and customer service.

These real-world applications show that AI is not just a trend. It is a practical tool that helps teams work better and more efficiently.

Cybersecurity and IT: AI on the front line

Marketing: still leading the way

Across businesses, marketing remains the most advanced function when it comes to AI adoption. According to KPMG’s 2026 Trends of AI, it continues to lead in terms of use cases, even if these are still largely focused on content creation. Content generation is at the top of the list. 55% of French companies already use AI to produce text, messages, or marketing materials. More broadly, nearly 75% of marketers rely on generative tools to analyze customer data and make recommendations, whether that’s predicting behavior, personalizing campaigns, or refining targeting.

Beyond marketing, these tools are also making their way into day-to-day tasks across other functions. In procurement, for example, 29% of companies use AI to draft professional messages, while 17% use it to translate documents or when working with multi-lingual teams.

HR: more personalized learning

In human resources, AI is making training more tailored and more effective. Around 38% of French companies use intelligent tools to personalize learning paths and track skills, helping improve both program efficiency and employee engagement, according to KPMG.

Supply chain: anticipating disruptions

In supply chain operations, AI helps simulate disruptions and optimize inventory planning, making it easier to respond quickly when things change. According to KPMG’s Trends of AI 2026, around 30% of French companies have already deployed or tested these solutions to anticipate issues, reduce delays and losses, and improve customer satisfaction.

Healthcare: a valuable ally

In healthcare, AI is helping teams deliver more personalized and effective care. Some solutions can quickly analyze medical records and test results to identify patterns or anomalies, supporting clinical decisions and reducing errors. As highlighted in our article Augmented health: when AI helps us care better, AI is becoming a practical, everyday ally for improving outcomes and responsiveness.

Customer service: faster, smarter responses

In customer service, AI helps teams respond faster and handle complex requests more accurately. Conversational assistants and semantic analysis tools can now manage a large share of routine interactions, guide users to the right information or contact, and provide support around the clock. By analyzing past interactions and customer data, these systems also help personalize responses and anticipate needs, improving both service quality and team efficiency.

These use cases show that AI is already delivering real, tangible benefits. It improves efficiency, enhances service quality, and supports better decision-making, while making everyday work easier for teams.

But for these benefits to last and remain responsible, simply adopting AI is not enough. Companies need clear rules and frameworks to ensure it is used in a way that meets legal and ethical requirements. That is where AI governance and the AI Act come into play.

AI and regulation: the governance challenge

For a long time, AI in business was mainly seen as an innovation topic. From 2026 onward, it also becomes a matter of regulatory compliance. The full implementation of the European AI Act on 02 August 2026 marks a major shift. AI is now governed by a harmonized legal framework across all 27 EU member states. Companies need to put in place real AI governance structures that can identify risks, document usage, and ensure transparency across systems.

A European framework based on risk levels

At the core of the AI Act is a classification of AI systems based on four levels of risk:

• Unacceptable risk: certain uses are banned, such as mass biometric surveillance.
   
• High risk: systems subject to strict requirements, for example in healthcare, recruitment, or critical infrastructure.
   
• Limited risk: transparency obligations for users.
   
• Minimal risk: little to no constraints.

This approach makes it possible to regulate sensitive uses without slowing innovation in less critical areas. For organizations, this means mapping AI usage in detail.

The question quickly becomes very practical: Which tools rely on AI, in which business processes, and with what level of regulatory risk? 

Concrete requirements from 2026

For many organizations, AI compliance in 2026 brings new requirements:

• Documenting AI systems in use
   
• Tracking data and automated decisions
   
• Assessing ethical and regulatory risks
   
• Putting human oversight in place for certain use cases

The stakes are real. Non-compliance can lead to fines of up to €35 million or 7% of global annual revenue.

As a result, companies are taking this seriously. According to a KPMG study, 60% of French companies have already set up cross-functional AI governance, and 86% have adopted responsible AI guidelines.

From compliance to trust

Beyond regulatory pressure, governance can also become a source of trust. Companies that can demonstrate transparency in how their AI systems operate help reassure customers, partners, and employees alike.

This is why many European tech players are prioritizing infrastructures that meet European standards, especially for data hosting and AI model processing.

To support this shift, some solutions now integrate governance and traceability features directly into AI tools.

Because ultimately, the question is no longer just how to use AI, but how to use it responsibly, securely, and in compliance with regulations. In Europe in 2026, that capability is becoming a real competitive advantage. 

Key AI trends in 2026, explained simply

Generative AI marked a first turning point for businesses. In 2026, a new phase is emerging. AI is not just responding anymore, it is starting to act.

Two trends stand out in particular: agentic AI and multimodal AI.

These terms may sound technical, but the ideas behind them are actually quite straightforward.

Agentic AI: from assistant to teammate

For years, AI worked like a fast assistant. You asked a question, it gave you an answer.

Agentic AI goes further. It acts more like a digital teammate that can handle a sequence of tasks on its own.

Think of an assistant that not only suggests how to plan a business trip, but also compares flights, books the hotel, and updates your calendar automatically. That is what autonomous AI agents can do. They analyze a situation, make simple decisions, and carry out a series of actions. 

This shift is already visible in companies. According to Deloitte, 58% of organizations are using or experimenting with agentic AI.

Adoption is global, although uneven. Around 40% of companies in Europe are exploring these technologies, compared with 30% in France (Adobe).

Scaling remains gradual. Only 13% of French companies have deployed agentic AI at scale, showing that many are still in the testing phase.

Multimodal AI: combining text, image, and voice

The second major trend is multimodal AI. Until recently, most AI systems were specialized. Some handled text, others images, others voice. Multimodal AI brings these capabilities together. It works more like a human, able to process different types of information at once.

A single AI system can now:

• analyze an image
• read a document
• process a voice message
• and generate a coherent response by combining all of these inputs.

This ability relies on the latest generation of generative AI models, which are becoming increasingly versatile.

Gartner highlights how 80% of companies worldwide are expected to use generative AI technologies by 2026, accelerating the adoption of systems that can combine multiple types of data.

In practice, these solutions often rely on several models working together. Some platforms, for example, use multiple large language models to select the most relevant one for each task.

For companies, this opens the door to deeper transformation. While 60% of large French companies have already implemented cross-functional AI governance, many smaller organizations are still exploring without a formal strategy. But those preparing now already see AI as a long-term transformation strategy, well beyond short-term productivity gains. (KPMG, Trends of AI 2026).

To explore these developments further and understand their impact on organizations, you can take a look at this analysis from the Orange Business: Présent et futur de l’IA : quelle trajectoire pour les entreprises ?

Rolling out AI across your organization

Between excitement and widespread experimentation, many companies are now asking the same question: how do you move from idea to a real AI project that actually delivers?

The good news is that implementing AI often follows a fairly simple path. Start small, test quickly, then scale progressively. What really makes the difference is building trust from the outset with both teams and users. That is what drives adoption and unlocks real value.

1. Identify the right use cases

The first step is to focus on two or three priority use cases that can deliver visible results. Customer service automation, document analysis, logistics optimization, marketing support… the goal is not to move fast at any cost, but to focus on high-impact projects with measurable outcomes.

According to Stema Partners, quoted above, these initial use cases can generate ROI in around 6.7 months for SMEs, and about 10 months for mid-sized companies. That makes them a strong starting point for launching an AI initiative.

2. Launch a pilot

Once use cases are defined, it is time to experiment. The objective is not to transform everything overnight, but to test an initial version of the solution in real conditions. This phase is similar to a prototype in industry. It helps confirm that the technology works, that the data can be used effectively, and that teams can adopt the tool.

According to the AI Barometer led by Denis Atlan (2022–2025), early AI projects in French SMEs show a median ROI of 159%, with returns typically seen within 6 to 7 months. Results will always vary, but the potential is clear.

3. Prepare your data

AI is powerful, but only as good as the data behind it. If data is scattered or poorly structured, results will be limited.

Data is often described as the fuel for AI. Before scaling, companies need to organize their data sources, secure access, and ensure quality.

This step is often the least visible, but it is one of the most critical for a successful deployment.

4. Train your teams

5. Put governance in place

With the rise of generative AI, a new phenomenon is emerging: “shadow AI.” According to Microsoft/YouGov 2026, around 61% of employees in France use AI tools outside official frameworks, often through personal accounts.

This highlights why strong governance is essential. Without clear rules, companies expose themselves to real risks, including data leaks, non-compliance, or flawed automated decisions. Setting clear guidelines helps unlock the full value of AI while keeping it secure.

That includes defining which tools can be used, how sensitive data is protected, when human oversight is required, and how to detect potential bias in outputs.

The goal is simple: ensure AI is used in a responsible, ethical, and secure way. This becomes even more critical as European AI regulations come into force.

6. Scale up

Once early results are validated, the next step is scaling.

AI models are integrated into business tools, use cases expand, and benefits become structural. This is when AI moves from experimentation to a real performance driver.

Orange Business and AI: scaling with confidence

Implementing AI is one thing. Scaling it across an organization is another. Data, security, compliance, integration with business tools… successful transformation requires a strong technology ecosystem and the right expertise. That is exactly the approach developed by Orange Business, supporting companies in their digital transformation and AI adoption.

With Live Intelligence, the company has already reached more than 100,000 users in under three years, leading to the creation of 20,000 personal AI assistants and 300 industrialized AI agents. Adoption continues to grow among enterprise customers.

This success is built on a trust-based approach, ensuring security, transparency, and ethical use of AI solutions. 

Scalable technological expertise

To support AI projects, Orange Business relies on a broad ecosystem of experts and researchers.

Today, nearly 3,900 specialists in data, AI, and digital transformation are working on it, covering every stage of a project, from strategy and architecture to deployment and governance.

This operational expertise is reinforced by the Group’s research capabilities. Orange Innovation brings together around 700 researchers, including 110 PhD and postdoctoral researchers, working on technologies such as AI, networks, and cybersecurity. Research plays a central role in Orange’s innovation strategy, with an open approach that connects technological trends with real-world uses.

Accelerating AI adoption

To make AI easier to adopt, Orange Business develops ready-to-use solutions. One example is the Live Intelligence platform, which gives companies access to generative AI capabilities in a secure and trusted environment. It is built on an architecture that orchestrates multiple AI models such as GPT, Gemini, Claude, or Mistral, while maintaining full control over how data is used.

This approach brings clear benefits: selecting the most relevant model for each task, optimizing performance, and ensuring data sovereignty.

Today, Live Intelligence has more than 50,000 regular users, reflecting growing demand for secure AI environments. The platform has recently expanded with Live Intelligence Studio, enabling the development of advanced agent-based solutions to automate key business processes, while maintaining the same standards of security and trust. Pour faciliter l’adoption de l’IA, on développe chez Orange Business des solutions clés en main.

Discover Live Intelligence

Responsible, more efficient AI

Beyond performance, Orange Business also places responsibility at the core of its approach through a “Frugal AI” strategy.

In practice, this means encouraging both clients and teams to choose the most relevant AI model for each use case, rather than defaulting to the most powerful one.

This approach promotes more efficient solutions that optimize resource use and reduce energy consumption, while maintaining strong performance.

At the same time, Orange is working to increase transparency by making the carbon impact of AI models visible and measurable. This gives companies a clearer view of their footprint and helps them manage and reduce it.

From experimentation to transformation

The goal is not just to test new tools. It is to embed AI sustainably into business processes.

To support this, Orange Business offers end-to-end support:

• Identifying use cases
• Structuring data
• Ensuring regulatory compliance and risk management
• Developing and testing AI solutions
• Scaling deployment
• Supporting change and training teams

This comprehensive approach helps companies move from experimentation to real, data-driven transformation.

Ultimately, AI is becoming a strategic infrastructure, just like cloud or networks. The challenge now is to integrate it in a way that is reliable, secure, and scalable.