• According to Afef Belhadj, Director of ITN & Data and in charge of leading business communities at Orange, the key skills for 2026 include a scientific mindset (mathematics), proficiency in advanced tools (RAG, agents, MCP), and in-depth knowledge of regulations (AI Act, NIS2, GDPR).
• In recruitment, junior candidates must now demonstrate their ability to produce tangible results (mockups, prototypes) rather than relying on AI certifications.
More than three years after the generative AI boom, have tech professionals at Orange truly changed their work methods?
Afef Belhadj. Yes, and I now say without hesitation that it’s a revolution, even though I was initially rather skeptical about certain announcements of major transformation. What has changed everything isn’t ChatGPT itself, but the ecosystem that has built up around it—namely the frameworks, the diversity of LLMs, abstraction layers like MCP, agents, and so on. Today, the developers, architects, and integrators who build Orange’s products and networks will gradually change their methods, so that we are no longer just talking about a tool that improves monitoring or writing, but about a profound transformation in the way we produce.
With generative AI, it’s easy to put on an impressive show without real depth, but in reality, what truly impresses is what the candidate has actually produced.
What kind of profiles are you looking for now?
Trying to find a unicorn is a trap. At Orange, we build complex networks, and a network engineer who isn’t an expert in their field has no place here. But we can’t be closed off to adjacent technologies. Today, a network expert is increasingly proficient in software concepts and code generation, and is required to use AI tools validated by the group. This is what we call the hybridization of roles: remaining an expert while staying open to new areas. This shift began before generative AI, with the software-driven evolution of networks. AI has accelerated it.
What should be on your resume in 2026 to be credible in this completely transformed market?
First, a solid scientific mindset is necessary because agents, abstraction layers, and LLM architectures require scientific understanding and sound reasoning, in the sense that mathematics and algorithms are returning to center stage. AI tools will be used to enhance skills, not weaken them, with a view to increased productivity. These are the types of profiles we’ll be looking for: I know how to generate code, but more importantly, I know how to debug it and assess its quality.
Next, you need to stay up to date on the tools—not just consumer-facing generative tools, but frameworks like RAG, MCP, agents, and their orchestration.
Thirdly comes mastery of the issues underlying regulation. The AI Act takes effect this year with stricter requirements for high-risk systems, and someone who comes in with a grasp of algorithmic transparency, GDPR compliance as it applies to AI, and LLM security has a real advantage. Finally, even if it may seem trivial, having documentation skills has become strategic, particularly for accurately documenting one’s expertise so that agents can take over. This is a significant cultural shift, especially in companies like Orange where much expertise is passed down from employee to employee.
From what you’re saying, prompt engineering is disappearing from resumes…
In reality, it’s already on its way out as a differentiating skill. Three years ago, we were considering training programs on this topic, but today it’s as basic as writing an email or using Excel. In five years, natural language will be sufficient for everyday use, and agents will know their users so well that it won’t even be necessary to provide context anymore.
How do you assess these new skills in an interview, when everyone can claim to be “comfortable with AI”?
That’s the real challenge for recruiters. With generative AI, it’s easy to put on an impressive show without any real depth, but in reality, what truly impresses is what the candidate has actually produced: a mockup, a prototype, an application developed independently, and a creative mindset. The development lifecycle has shortened to the point where anyone in the chain can almost single-handedly produce an entire product. If you come in with that, you no longer need major certifications beyond a recognized science degree.
Has Orange implemented measures to support this upskilling?
Yes, and our partnership with Coursera is both an example and a key driver of this. For several years now, many Orange employees have had access to the platform at a negotiated rate well below the retail price. The offering covers data, the cloud, IT, and now the entire generative AI ecosystem. Completed training courses are tracked and recognized in career development, and experts can go well beyond generalist tracks. It’s a real accelerator.
Beyond training, Orange’s business communities play an essential role in skills development. Learning often progresses faster when we can draw on the experience of colleagues who share their expertise, best practices, and concrete use cases tailored to our environment. This takes the form of regular events, open to the entire Group and across all our regions. Here again, it is a powerful catalyst for skill development.
This text has been translated by an artificial intelligence.
Afef Belhadj
