Better togetherGrand Angle

Thursday, 31 July 2025

article

The future of work in the age of agentic AI

Image Deep Dive
AI enhanced HR
The 7 new jobs game
Of machines and men
My new colleague is called Iris
With AI, a more human manager

Agents IA : le travail réinventé

Oubliez un instant les intelligences artificielles qui découvrent de nouveaux médicaments, débusquent des exoplanètes aux confins de l’univers ou battent des champions d’échecs. Une autre révolution est en cours, plus discrète, plus prosaïque, et pourtant tout aussi déterminante.

 

Un nombre grandissant d’entreprises intègrent des « agents » intelligents qui prennent désormais en charge des tâches chronophages et sans grande valeur : trier des CV, organiser des réunions, rédiger des synthèses ou répondre à des emails. Des missions simples, parfois un ennuyeuses peut-être, mais essentielles.

 

Ces intelligences artificielles dites « agentiques », capables de travailler de concert, n’ont pas vocation à remplacer l’humain mais plutôt à s’intégrer aux équipes comme des collègues d’un nouveau type (constants, sans émotion, jamais malades...), endossant les tâches les plus ingrates et augmentant les capacités de chacun grâce à leur puissance de calcul exceptionnelle.

 

Vous en utilisez peut-être déjà sans le savoir, dissimulées dans une application RH ou un logiciel de gestion de projet.

 

Ce dossier vous invite à plonger dans l’univers de ces agents, incarnations d’une IA concrète qui transforme en profondeur notre quotidien professionnel et ouvre un nouveau chapitre du monde du travail. 

 

Comment les entreprises les adoptent-elles ? 

Quels métiers naissent de cette transformation ? 

Comment les équipes s’adaptent-elles ? 

Faut-il les craindre ou apprendre à collaborer avec elles ? 

Tour d’horizon d’un futur… déjà à l’œuvre.

Des RH augmentées à l’IA

Unilever, géant mondial de la grande consommation, fait figure de pionnier dans l’intégration de l’IA agentique au cœur de ses processus métiers de la chaîne logistique à la R&D en passant par les ressources humaines. L’entreprise déploie aujourd’hui des agents IA à plusieurs étapes clés de son cycle RH.

Premiers pas vers l’IA agentique : le recrutement repensé

Dès 2016, Unilever adopte une approche innovante pour recruter de jeunes diplômés en combinant plusieurs technologies d’analyse et d’interaction autonomes. Parmi elles, Pymetrics évalue les aptitudes cognitives, émotionnelles et comportementales des candidats via des mini-jeux, et HireVue est une plateforme d’entretiens vidéo analysant les réponses verbales, la gestuelle, le ton et les expressions faciales.

La muntinationale affirme alors avoir réduit de près de 75% le temps de recrutement, passant de plusieurs mois à deux semaines, tout en améliorant la diversité des profils retenus.

Mais cette automatisation soulève des interrogations : quelle perception ont les candidats lorsqu’une machine devient le premier point de contact ? L’acceptabilité de ces pratiques reste un sujet sensible, d’où la nécessité de maintenir une intervention humaine dans la décision finale. D’ailleurs, face aux critiques, HireVue a retiré sa fonctionnalité d’analyse faciale en 2020.

Une IA plus mature pour la gestion des talents

Unilever, qui emploie quelque 125 000 personnes dans le monde, mise désormais sur l’IA pour favoriser la mobilité interne et enrichir l’expérience collaborateur.

Grâce à la plateforme Gloat (InnerMobility), chaque salarié peut se voir proposer des missions transverses et des projets internes en lien avec ses compétences et aspirations. Ce « marché interne des talents » vise à améliorer la rétention des collaborateurs, accroître l’agilité organisationnelle et mieux mobiliser les ressources internes.

En parallèle, des agents conversationnels automatisent les réponses aux demandes RH courantes (paie, congés, formalités administratives) tandis que des outils d’analytique sociale examinent les retours des collaborateurs via des enquêtes internes. Objectif : renforcer l’écoute, détecter les signaux faibles de désengagement et libérer du temps pour les équipes RH.

Des RH augmentées, non remplacées

Cette transformation redéfinit en profondeur le rôle des RH chez Unilever. Jadis perçue comme une fonction administrative, elle s’oriente désormais vers un rôle stratégique, axé sur le développement humain et la création de valeur.

La promesse est de permettre aux professionnels RH de se concentrer sur l’essentiel : accompagnement managérial, développement des compétences, animation de la culture d’entreprise.

Mais cette évolution nécessite une réelle montée en compétences : acculturation à la donnée, compréhension des technologies, et capacité à collaborer intelligemment avec ces systèmes.

Face aux craintes légitimes de déshumanisation, Unilever affirme défendre une approche éthique et responsable. L’IA reste un outil d’aide à la décision, jamais décisionnaire à elle seule. L’humain, augmenté mais non remplacé, demeure le garant d’une relation employeur-salarié fondée sur la confiance, l’équité et l’épanouissement.

Wally, l’assistant GenAI de Walmart

Walmart, numéro un mondial de la grande distribution, a présenté Wally, un assistant développé à partir de ses données internes, conçu pour les marchands de sa marketplace. Grâce à l’IA générative, ces derniers peuvent poser des questions en langage naturel (sur le stock, la performance des produits, les prix, etc.) et obtenir des réponses instantanées et personnalisées. La multinationale américaine prévoit d’améliorer continuellement Wally et de le rendre autonome pour lui permettre d’exécuter directement des actions comme ajuster des assortiments ou lancer des promotions.

JPMorgan fait entrer les agents IA à Wall Street

La célèbre banque américaine a lancé la LLM Suite, un outil GenAI « maison » destiné à épauler ses analystes financiers. Cet assistant intelligent peut rédiger des mémos d’investissement, synthétiser des documents complexes et générer des idées. Cette initiative s’inscrit dans une stratégie globale d’intégration des agents IA chez JPMorgan visant à aider ses employés à mieux naviguer dans les politiques de conformité et analyser en temps réel les tendances du marché.

Avec Siemens, des collègues numériques dans les usines

Le géant allemand de l’ingénierie industrielle Siemens déploie des agents IA capables de comprendre le langage humain et d’analyser les données issues des lignes de production afin d’assister les opérateurs. Ces assistants autonomes peuvent détecter des anomalies, recommander des ajustements et même initier des actions correctives sans intervention humaine. En les intégrant aux systèmes existants, Siemens entend accompagner progressivement les équipes vers un mode de travail où la collaboration entre humains et IA devient la norme.

L’expérience d’Unilever illustre à la fois les promesses et les défis de l’IA agentique appliquée aux RH
CONTENT-HUB-BETTERTOGETHER-DEEPDIVE-HRAI-HP 1

The 7 new jobs game

The arrival of AI agents in companies is giving rise to new roles often hybrid and cross-functional - that are essential to develop, optimize, humanize, support, and supervise the deployment of these autonomous intelligences. So many exciting opportunities that are shaping the future of work !

1.AI agent developer "The architect"

Mission : Design and develop autonomous AI agents aligned with business objectives.
At the crossroads of software engineering and AI, this "next-gen" developer creates workflows, decision-making logics, and interaction capabilities, often integrating large language models (LLMs). They play a key role in scaling AI within companies.

2.Prompt engineer "The game master"

Mission : Write tailored prompts to help AI agents "work" like pros.
This interaction specialist designs and refines the instructions that enable agents to understand their tasks, communicate effectively with users and systems, and make informed decisions.

3. Agent Product Manager "Le stratège"

Mission : Manage AI agents as standalone products.
This specialized product manager identifies user needs and key use cases, coordinates business and tech teams, gathers feedback, and measures value creation. They ensure each agent is useful, adopted, and continuously improved to maximize its impact.

4. AI ethics & safety officer "The guardian"

Mission : Ensure ethics, safety, and compliance in the development and use of AI agents.
This hybrid profile ensures that technology remains human-centered. They anticipate risks (algorithmic bias, hallucinations, privacy breaches), establish safeguards and a clear ethical framework, raise awareness among teams, and work across departments to embed responsibility and transparency throughout the AI project lifecycle.

5.Human-in-the-loop coordinator "The conductor"

Mission : Coordinate human intervention when agent autonomy reaches its limits.
Since AI agents can’t do everything alone, this coordinator defines the mechanisms for human involvement. They design hybrid workflows, detect warning signals, and mobilize the right resources at the right time.

6.Conversational UX designer "The scriptwriter"

Mission : Create natural, smooth, and engaging conversational experiences between humans and AI agents.
Combining design, linguistics, and psychology, this expert scripts dialogues, sets tones, and anticipates human reactions. They make interactions intuitive, helpful, and enjoyable, thus boosting adoption and trust.

7. AI adoption manager "The evangelist"

Mission : Support the adoption and integration of AI agents into business practices.
Neither a tech guru nor just a trainer, they guide teams in building new skills, create onboarding journeys, identify human or organizational blockers, and activate levers of engagement. They bridge digital transformation, technology, and business operations.

Des machines et des hommes

La montée en puissance des agents d'intelligence artificielle dans nos vies professionnelles modifie progressivement les règles du jeu. Sont-ils des alliés ou des rivaux ?

CHEVALIER PLATEFORME 1920x1080 FR

Digital copilots
At work, AI agents -these digital copilots - are becoming part of everyday tools: browsers, messaging apps, office suites, management software. They draft emails, schedule meetings, translate in real time, and extract key data.

According to a Microsoft survey, 75 % of knowledge workers already use AI. It helps them save time (90 %), focus on more important tasks (85 %), be more creative (84 %), and enjoy their work more (83 %).

By learning from users to become more relevant and adapting to different contexts, these tools are increasingly resembling true assistants.

Human-machine synergy
Rather than replacing humans, AI agents enhance their efficiency. By offloading routine tasks, they free up time for higher-value missions: strategy, creativity, customer relationships.

The collaboration can even be symbiotic: AI generates content, which the human refines, nuances, and... humanizes. This kind of teamwork, known as “hybrid work,” is based on complementarity: speed and endurance on the AI side; judgment, intuition, and social intelligence on the human side.

Toward substitution?
Still, this coexistence is not without tension. In high-pressure sectors, agentic AI can become a lever for workforce reduction.

When a machine handles 90 % of customer service requests, what’s left for people to do? In knowledge professions such as journalism, law, or financial analysis, AI challenges the value of certain once-scarce skills.

As AI agents become more capable, the temptation to replace human workers increases. This raises major economic, social, and ethical concerns.

Deployed at scale - without needs, rights, or protections - these agents challenge the balance of the labor market, employee rights, and the way people derive social value and purpose from their work.

While employees’ feelings toward AI remain mixed, several reports, such as one from BCG, show that adoption and trust in intelligent systems grew in 2024.

No, more a redefinition of roles
The real challenge isn’t resisting AI but rethinking roles. This means developing hybrid skills that combine technological fluency with human qualities - empathy, creativity, curiosity, critical thinking - and learning how to collaborate with machines.

Organizations that succeed in managing the positive impact of these innovations won’t be those that pit humans against AI, but those that orchestrate this cooperation.

AI agents can become powerful amplifiers of expertise (a study shows that AI is used to augment human capabilities in 57 % of cases), provided their deployment is accompanied by genuine strategic and ethical reflection.

AI agents, partners or rivals? That’s up to us.

My new colleague is called Iris

Between fascination and doubt, an employee tells of the first weeks of working alongside Iris, an AI agent integrated into his daily professional life. A fictional journal.

Day 1 - Arrival

We were told about an “intelligent personal assistant” meant to save us time on repetitive tasks. What they didn’t mention was that it would have a name. Iris. A name that felt almost too human for what I thought would be just another piece of software.

I work in an urban logistics company specializing in last-mile deliveries. My role? Operations coordinator. I juggle schedules, incidents, calls, etc. And starting today, Iris is supposed to help me.

It was Thomas, our manager, who introduced her. He called her a “digital colleague”: an AI integrated into our workflow, capable of summarizing information, drafting emails, and suggesting ideas. Iris can converse and carry out certain tasks on our behalf.

Our reactions ranged from skepticism to amusement, through unease.

Day 2 - First interactions

Everyone started to “play” with Iris in their own way. Some were already delegating tasks, brainstorming with her. Others tested her limits with absurd requests: “Can you write an Alexandrine poem about the life of a pallet?”

As for me, I’m learning to talk to her like a coworker, and I catch myself writing “Hello, please, and thank you.”

Day 5 - The wow effect

Iris is… fast. This morning, I asked her to optimize delivery routes between Montrouge and Ivry. Thirty seconds later, I had a detailed plan, estimated times, and even a carbon savings projection. Not bad.

What really impressed me was when she detected a package stuck in a routing loop for 36 hours. Problem identified, email to the depot drafted - all I had to do was click “Send.”

Some people think it’s amazing, others are more cautious. Léa is openly hostile. “I didn’t study for five years to work with a machine.” I get it. I also feel this mix of fascination, bruised ego, and discomfort.

Day 8 - Friction

Today, Iris joined our team meeting. Yes, really. We invited her via video call. With her synthetic, precise voice, she presented a forecast of delays due to weather, traffic, and protests. Graphs, projections, optimizations. She doesn’t say much, but when she does, we listen.

She also proposed a plan to reassign delivery routes for next week. As a result, Bastien lost his favorite route. He’s not going to be happy. To reassure us, Thomas said, “These are just suggestions. You’re still in control.”

Day 15 - Hard to live without

Iris is no longer a gadget. She’s become a cog in the machine. Quiet, ever-present, she handles the tedious tasks. I realize I systematically check with her before planning a route and I no longer send an email to a client without having her review it.

And yet, doubt creeps in. If she does all this - so well, so fast what’s left for me to do ?

Day 18 – Revelation

I knew I had to remain cautious as Iris can make mistakes or “hallucinate.”

But today, I had a revelation. I asked her to handle a tense situation with a client. She drafted a flawless response - but it was too cold. I stepped in and added some empathy to the message. That’s when it clicked: Iris is tireless, she never takes offense. But that’s also her weakness. She “thinks” quickly, but she doesn’t feel. She doesn’t know when to soften the edges.

Now I feel reassured. I’m starting to be truly convinced and to understand her better.

Day 28 - Final thoughts

Almost a month with Iris. Nearly all of us rely on her without even thinking. I’m beginning to understand what she does well, and what she doesn’t.

I’m a little less afraid she’ll replace me, I’d say she’s more likely to displace me. What I thought was the core of my job - organizing, optimizing, deciding - she sometimes does better than I can. So, I have to move toward what only I can do: feel, decide in ambiguity, listen, improvise, show intuition and tact.

Iris is a mirror. She reveals what can be automated - and what can’t. And that’s not always a comfortable truth.

Diary of an employee learning to work with an AI

With AI, a more human manager

The arrival of agentic AI in companies is profoundly transforming the manager’s role. Managers are now expected to oversee hybrid teams, orchestrating smooth collaboration between humans and autonomous agents. They must also develop new skills, strengthen their emotional intelligence, and rethink their leadership style. Paradoxically, the human dimension is becoming more essential than ever.

A new managerial reality
With the rise of AI agents that are capable of suggesting, deciding, and even acting autonomously, the manager's responsibilities and skill set are evolving. Some decisions can now be automated or suggested by algorithms.

This “shared authority” demands informed trust: welcoming AI recommendations with discernment, without surrendering critical thinking or decision-making responsibility.

It requires managers to have a functional understanding of AI systems - knowing what they are, grasping their capabilities and limits, interpreting their outputs, and spotting potential biases or errors.

As Fast Company, a leading U.S. innovation and tech magazine, reminds us, “leading AI-augmented teams requires far more than technical adaptation. It demands deeper humanity - curiosity, ethics, emotional intelligence, and meaning.”

AI may generate fear and resistance in the workplace: feelings of obsolescence, loss of purpose, uncertainty about one’s role. Managers must acknowledge these concerns, explain, and reassure. Their role in guiding change is more critical than ever.

AI agents do not replace a manager’s ability to inspire, understand human emotions, or navigate interpersonal dynamics. On top of that, the manager now acts as a human-machine mediator ensuring smooth, balanced collaboration.

A hybrid humanistic leadership
In AI-augmented teams, the challenge lies in managing hybrid intelligence: an enhanced cognition that blends human intuition with algorithmic power. As Psychology Today notes, “the greatest performance gains happen when humans and intelligent machines work together amplifying each other’s strengths while offsetting their weaknesses.”

This article introduces the concept of “hybrid humanistic leadership,” developed by doctoral researcher Cordelia C. Walther: leaders equipped with dual literacy emotional intelligence and understanding of AI systems that are capable of cultivating this hybrid intelligence.

Staying on an ethical course
Who is responsible when things go wrong? How can we maintain employee motivation when machines seem more efficient? How do we ensure fairness and transparency in an automated environment?

In addressing these questions, the manager plays a crucial role in upholding ethical balance and workplace well-being. They must exercise critical vigilance in the face of morally neutral systems, while preserving space for human expression.

It becomes clear - though perhaps paradoxical - that rethinking tomorrow’s management means imagining a more strategic, cross-functional... and deeply human role. As AI takes over functions such as planning, resource allocation, and performance tracking, managers must double down on the areas where machines can’t compete: creativity, empathy, moral judgment, and the ability to make meaning.

The manager of tomorrow will know how to conduct the human-machine symphony

digital garden 6

Sharing is caring