It’s January 2026. As you read these lines, over 57,000 Telus employees are saving an average of 40 minutes per interaction with AI agents. Suzano, the world’s largest pulp manufacturer, has reduced query times by 95% for 50,000 employees thanks to an AI agent that translates natural language questions into SQL code.
These aren’t futuristic experiments. They’re current operational realities.
2025 was the transition year: from hype to practical implementation. Now, entering 2026, we face a much more concrete question: what skills are needed to thrive in a job market where AI is no longer an option, but the infrastructure itself on which careers and companies are built?
Year Zero of Agentic AI: What’s Happening Right Now
Gartner predicts that by 2028—just two years from now—15% of daily work decisions will be made autonomously by agentic AI, compared to 0% in 2024. 33% of enterprise software applications will include AI agents by 2028, compared to less than 1% today.
But what is the agentic AI everyone’s discussing at the start of 2026?
Traditionally, AI tools were “reactive”: you’d ask ChatGPT something, it would respond. Done. Agentic AI represents a qualitative leap: autonomous systems that reason, plan, and execute multi-step workflows toward defined goals, making independent decisions along the way.
Imagine an assistant that doesn’t just respond “Here’s the list of VIP clients to contact,” but autonomously analyzes CRM data, identifies high-value clients, drafts personalized emails based on previous conversations, sends them at the statistically optimal time, monitors responses, and follows up with those who haven’t opened within 48 hours—all without you lifting a finger after the initial input “Prepare VIP client outreach”.
This is agentic AI. And in 2026 it’s moving from pilot to production.
The Numbers Telling the Transformation (January 2026)
Deloitte’s “Emerging Technology Trends 2025” report (just published) reveals a curious situation:
- 30% of organizations are exploring agentic options
- 38% are piloting solutions
- 14% have deployment-ready solutions
- 11% actively use systems in production
- 42% are still developing strategic roadmap
- 35% have no formal strategy
In other words: we’re at the beginning, but acceleration is exponential. While 77% of organizations haven’t yet implemented agentic AI in production, predictions for end of 2026 indicate a tripling of deployments.
IDC predicts that by 2029, global IT spending dedicated to AI will exceed 25% of the total, driven precisely by agentic AI. AI investments will register an annual increase of 31.9% between 2025 and 2029, exceeding $1.3 trillion by the end of the four-year period.
The real question is no longer “If” but “How quickly”.
New AI Professions: Who’s Hiring Now
The European Commission estimates 1.7 million new jobs related to AI by 2030 in Europe alone. In Italy, according to ManpowerGroup, 78% of companies struggle to find talent with AI skills—a paradox in a country with 4.7 million workers at risk of automation.
Here are the professional figures emerging right now, at the start of 2026.
Prompt Engineer: The AI Language Architect
What they do: Design, test, and optimize text inputs to maximize AI response effectiveness. It’s not “just” writing good questions—it’s applied linguistic engineering.
A prompt engineer works with complex chains of reasoning (chain-of-thought), few-shot learning, temperature control, and advanced techniques like tree-of-thoughts reasoning. They must understand how different models (GPT-4, Claude, Gemini) respond to linguistic nuances, build reusable frameworks, and measure performance quantitatively.
Required skills:
- Deep understanding of NLP and LLM architectures
- Computational logic and algorithmic thinking
- Excellent natural language writing ability
- Knowledge of cognitive psychology and linguistic biases
- Iterative testing and A/B testing
Average salary 2026: $60K-$95K entry level, $100K-$150K+ senior (US). In Europe €45K-€120K+ depending on market.
Who’s hiring: Anthropic, OpenAI, Google DeepMind, AI-first startups, digital consultancies
Reality: On LinkedIn, terms like “Prompt Engineering”, “Prompt Crafting”, and “GPT” have been added to profiles by 75% of global users in the past year. Demand far exceeds supply.
A practical example: Klarna replaced 700 human customer service agents with an AI assistant handling the work of 700 people, providing responses in 23 languages. Who designed the prompts behind this system? A team of prompt engineers who iterated for months on thousands of scenarios.
AI Trainer: The Machine Professor
What they do: Train AI models through Reinforcement Learning from Human Feedback (RLHF), the technique that made ChatGPT so effective. Provide qualitative feedback on AI outputs to “teach” the model what’s desirable and what’s not.
AI trainers aren’t data scientists, they’re domain experts (doctors, lawyers, writers, educators) bringing human expertise into AI training. A medical AI trainer evaluates whether diagnoses suggested by AI are appropriate. A legal AI trainer verifies if generated contracts contain correct clauses.
Required skills:
- Deep expertise in specific domain
- Ability to articulate qualitative judgments in structured way
- Basic understanding of supervised ML functioning
- Patience (continuous iterations)
- Critical thinking to identify biases
Average salary 2026: $45K-$70K (highly variable by sector)
Who’s hiring: Scale AI, OpenAI, Anthropic, companies developing vertical AI (healthcare, legal, finance)
Reality: This role became critical after discovering that better models come not just from more data, but from better quality data supervised by expert humans. Claude by Anthropic owes much of its “personality” to thousands of hours of qualitative feedback from AI trainers.
AI Ethicist: The Responsibility Guardian
What they do: Ensure AI is developed and implemented according to ethical, legal, and socially responsible principles. Analyze biases, assess impacts on minorities, design guardrails to prevent harmful uses.
With the EU AI Act entering force in 2025 and similar regulations coming globally, AI ethicists have moved from “nice to have” to legal requirement. Forrester predicts 60% of Fortune 100 will appoint a Head of AI Governance in 2026.
Required skills:
- Background in philosophy, ethics, law, or social sciences
- Technical understanding of how ML algorithms work
- Knowledge of regulations (GDPR, EU AI Act, etc.)
- Ability to translate ethical principles into implementable technical requirements
- Effective communication with boards and non-technical stakeholders
Average salary 2026: $70K-$120K (US), significantly higher in EU for AI Act compliance
Who’s hiring: Big tech, banks, healthcare, public sector, consulting firms
Reality: When Anthropic launched Claude, the entire process was supervised by a team of AI ethicists who designed the “Constitutional AI principles”. This isn’t marketing—it’s operational infrastructure. Companies ignoring AI governance risk hefty fines under the new European regulatory framework.
Machine Learning Engineer: The Intelligence Plumber
What they do: Translate ML models from research prototypes to scalable production-ready systems. While data scientists experiment, ML engineers build pipelines that handle millions of requests daily.
Required skills:
- Advanced programming (Python, Java, C++)
- MLOps, Docker, Kubernetes
- Cloud architectures (AWS, GCP, Azure)
- Performance optimization and debugging
- Understanding deep learning frameworks (PyTorch, TensorFlow)
Average salary 2026: $130K-$220K (US), €60K-€95K (Italy)
Who’s hiring: Everywhere. Literally every tech company.
Reality: Demand for ML engineers continues to exceed supply. In Q4 2025, 63% of tech companies reported difficulty finding these profiles—an improvement from 72% in 2023, but still critical.
Data Curator & Knowledge Engineer: The 2.0 Librarians
What they do: Organize, label, and maintain the quality of data used to train AI. With agentic AI needing “grounding” on specific company knowledge bases, this role has exploded.
The “data labeling” market will reach $3.5 billion by the end of 2026. But it’s not just labeling cat and dog images. Data curators manage complex taxonomies, ensure semantic consistency, and design ontologies that allow AI to “reason” about data.
Required skills:
- Information science, librarian background (surprisingly useful!)
- Understanding of relational databases and graph databases
- Maniacal attention to detail
- Domain-specific business understanding
- Tools: SQL, knowledge graph platforms, annotation tools
Average salary 2026: $45K-$75K
Who’s hiring: Enterprises implementing RAG (Retrieval-Augmented Generation), companies with large legacy knowledge bases
Conversation Designer: The Bot Screenwriter
What they do: Design the conversational experience of chatbots and voice assistants. Combines UX design, copywriting, and understanding of how people actually communicate.
It’s not “writing responses”. It’s architecting conversational flows that anticipate intent, handle ambiguity, recover from errors, and maintain personality consistency across hundreds of touchpoints.
Required skills:
- UX/UI design background
- Excellent copywriting skills
- Pragmatic understanding of language (how people really talk)
- Familiarity with conversational AI tools (Dialogflow, Rasa, etc.)
- Empathy and user-centric thinking
Average salary 2026: $50K-$90K
Who’s hiring: Banking, insurance, e-commerce, customer service operations
Reality: Salesforce and Google Cloud are building cross-platform agents using the Agent2Agent (A2A) protocol. Who designs how these agents converse with each other? Conversation designers.
Traditional Jobs Evolving (Not Disappearing)
The “AI will replace everyone” narrative is simplistic. Reality: 75% of workers expect AI to influence their role by end of 2026, but this means transformation, not extinction.
Doctors → AI-Augmented Physicians
AI-assisted radiological diagnoses. Personalized treatment plans generated by algorithms analyzing millions of case studies. Drug discovery accelerated by ML.
Doctors aren’t disappearing—they’re becoming supervisors of AI diagnostic systems. Google DeepMind developed an algorithm detecting breast cancer with 89% accuracy, surpassing human pathologists at 74%. But doctors are still needed to contextualize, communicate with patients, and make final decisions.
New skills required: AI literacy, ability to interpret algorithmic outputs, critical validation of AI suggestions
Lawyers → Legal Engineers
Automated jurisprudence research. Contract analysis via NLP. Litigation outcome prediction via ML.
40% of routine legal activities are already automatable. But advocacy, negotiation, litigation strategy? Remain human. Law firms now hire “legal engineers” who build hybrid workflows where AI handles discovery and research, allowing lawyers to focus on high-value elements.
Marketing Managers → AI-Augmented Marketers
Customer segmentation via unsupervised learning. GPT-4 assisted copywriting. Sentiment analysis on millions of social interactions.
Marketing has become applied data science. 2026 marketers don’t just write copy—they define parameters for AI systems that generate and test thousands of variants. They don’t choose an image—they train brand recognition models.
Developers → AI-Fueled Coders
We already discussed vibe coding in another article. But it’s worth reiterating: GitHub Copilot is just the beginning. In 2026, developer tools like Cursor, Replit Agent, and Windsurf are transforming programmers from “code writers” to “system architects”.
Developers are becoming managers of agentic workforce. They no longer write every function—they supervise AI agents that generate, test, and debug code. Their value shifts to system architecture, design decisions, and critical code review.
Cross-Cutting Skills Everyone Will Need (2026-2030)
Regardless of profession, some skills will become as universal as Excel was in the 2000s.
AI Literacy: You’re No Longer “Non-Technical”
According to the World Economic Forum’s Future of Jobs Report 2025, 39% of workers’ current skills will need to be updated or transformed by 2030.
What does “AI literacy” mean?
- Understanding what AI can/cannot do (avoid both hype and skepticism)
- Writing effective prompts for tools like ChatGPT
- Interpreting AI output critically (recognize hallucinations)
- Understanding basics of supervised/unsupervised/reinforcement learning
- Familiarity with terms: LLM, neural networks, training data, bias
How to acquire it: Coursera/edX courses on AI for Everyone, hands-on experimentation with ChatGPT/Claude/Gemini, follow AI newsletters (The Batch, Import AI)
Critical Thinking and Complex Problem Solving
Paradoxically, the more AI handles routine tasks, the more human value shifts to judgment, creativity, and solving ambiguous problems.
The World Economic Forum identifies “analytical thinking and problem solving” as the single most sought-after skill. Why? Because AI excels at well-defined problems but fails on ill-structured challenges requiring contextual judgment.
Human-AI Collaboration: The New Literacy
Not “using” AI. Not “being replaced by” AI. Collaborating with AI.
Citizens Financial Group is training every employee to “guide, supervise, and optimize digital colleagues”. It’s no longer an option—it’s the basic operating mode.
What it involves:
- Decomposing tasks into parts AI can handle vs. those requiring human judgment
- Designing hybrid human-AI workflows
- Monitoring AI performance and providing feedback for improvement
- Identifying when to delegate and when to directly supervise
Change Management and Adaptability
In a world where new AI tools launch weekly, the ability to learn quickly, unlearn obsolete approaches, and adapt to new workflows is critical.
ServiceNow predicts that “AI agents built on proven, deterministic workflows” will require new “agent management” skills that don’t exist in any university curriculum today.
Real Challenges We’ll Face (2026-2028)
Not everything is enthusiasm. There are concrete problems that will emerge in the next 2-3 years.
Shadow Agents: The New Shadow IT
Snyk predicts that in 2026 “a surge in shadow agentic AI will create one of the largest blind spots in enterprise security.”
What are shadow agents? Employees creating autonomous AI agents without IT approval, using accessible tools like Make, Zapier with GPT integration, or direct API calls to OpenAI. These agents access company data, make decisions, and operate without governance.
It’s the Shadow IT 2.0 problem, but with higher stakes because agents can act autonomously, not just read data.
The solution: Governance frameworks balancing innovation and control. Forrester predicts 60% of Fortune 100 will appoint a Head of AI Governance in 2026.
The AI Skills Gap
ManpowerGroup reports that 74% of global companies and 78% of Italian ones struggle to find talent with adequate AI skills.
The speed of technological evolution far exceeds the speed of the education system. Universities are still teaching curricula designed in 2020, while the market demands skills on technologies launched in 2024.
The response: Massive corporate upskilling. Companies can’t wait for universities to update—they must build internal programs. Amazon invested $1.2 billion in AI upskilling for employees. Google offers free AI certifications.
Ethics and Bias: Who’s Responsible When an Agent Errs?
Scenario: An autonomous AI agent in recruiting unconsciously filters against candidates from certain zip codes correlated with ethnic minorities. Who’s responsible? The software vendor? The company that implemented it? The HR team that configured it?
The EU AI Act of 2025 begins to answer, but practical jurisprudence will develop through real cases in coming years. Companies are entering unexplored legal territory.
ROI and Hidden AI Costs
While global spending on AI systems will reach $300 billion in 2026, over 40% of agentic AI projects will be canceled by end of 2027 due to rising costs, unclear business value, or inadequate risk controls.
Implementing AI has hidden costs:
- Compute infrastructure (GPUs aren’t cheap)
- Data curation and preparation (80% of ML work)
- Continuous testing and validation
- Change management and employee training
- Maintenance and monitoring
Companies that win will be those measuring realistic ROI, not those chasing hype.
The Distant Future: 2030-2035
Looking beyond the immediate horizon, what can we reasonably predict?
Narrow AI Everywhere, AGI Never (Probably)
Artificial General Intelligence—AI with general human intelligence—remains speculative. Most prudent researchers place it 20-50+ years in the future, if ever.
But specialized Narrow AI? Will be ubiquitous. Every business process will have dedicated AI agents. Every smartphone will have increasingly capable AI assistants. Every car will be at least partially autonomous.
The revolution won’t be AGI replacing humans. It will be billions of Narrow AIs amplifying human capabilities in specific domains.
Human-Machine Fusion: Brain-Computer Interfaces
Elon Musk’s Neuralink demonstrated in 2024 the first brain implant allowing a paralyzed person to control computers with thought. Synchron and other competitors are progressing rapidly.
By 2035, it’s plausible that brain-computer interfaces will become optional for professionals wanting to operate at thought-speed with AI tools. Is it science fiction? Maybe. But so was the idea you’d have a supercomputer in your pocket 15 years ago.
AI Abundance Economy?
If AI handles most cognitive work and robots handle physical work, what happens to the economy based on human labor?
Universal Basic Income experiments are happening in Finland, Netherlands, Kenya. As AI generates unprecedented productivity, how to distribute benefits will become the central political question of the 2030s.
Your Personal Strategy for the AI Future (Concrete Actions 2026)
Theory is useful. Action is necessary. Here’s what you can do concretely in 2026.
In the Next 30 Days:
- Experiment with AI tools daily: ChatGPT, Claude, Gemini—use a different one each week, learn limits and strengths
- Identify 3 repetitive tasks in your work: which could be automated or AI-assisted?
- Complete a free “AI Literacy” course: Coursera “AI for Everyone”, edX “Introduction to AI”
In the Next 6 Months:
- Build a personal AI project: use Teachable Machine or similar to solve a real problem
- Get a certification: Google AI Essentials, IBM AI Fundamentals, or similar
- Network with AI professionals: LinkedIn groups, local meetups, virtual conferences
In the Next 12 Months:
- Redefine your role: How can I become “AI-augmented” in my current profession?
- Consider specialization: Which new AI profession aligns with my existing skills?
- Teach others: Best way to consolidate knowledge is explaining it
For Companies:
- Start small with targeted pilot projects
- Measure ROI rigorously
- Invest in upskilling existing employees before hiring externals
- Build AI governance before it becomes a problem
- Treat AI as transformation, not just technology
Conclusion: Writing the Future, Not Suffering It
As we enter 2026, the question isn’t “Will AI take my job?” but “How can I evolve alongside AI to create value only I can offer?”
The history of technology has always been the same: new tools eliminate routine tasks, freeing humans for high-value activities. The industrial revolution eliminated 90% of agricultural jobs but created entire manufacturing industries. The internet destroyed traditional publishing but generated the digital economy.
AI will follow this pattern. Yes, some jobs will disappear. But many others will emerge—roles that don’t even exist in the boldest projections today.
The privilege and responsibility of living through this transition is that we can choose how to participate. We can be passive spectators or active architects of this future.
In 2026, every professional—from doctors to lawyers, from marketers to developers—is redefining what “expertise” means in a world where AI excels at routine cognitive tasks. Expertise shifts from “knowing everything” to “navigating ambiguity, exercising contextual judgment, and collaborating effectively with artificial intelligences”.
The professions listed in this article—Prompt Engineer, AI Trainer, AI Ethicist—are just the first. In the next 5 years, dozens more will emerge that we can’t even imagine today.
The final question isn’t “What will AI do?” but “What will I do with AI?”
Start today. Experiment tomorrow. Build the day after. The future of work isn’t something that happens to us—it’s something we build together, one acquired skill at a time.
