AI Agents in the workplace: from technology to people, what it means for your organisation and your job

AI Agents are performing work autonomously for the first time, changing roles, processes and even e-commerce. From the difference between probabilistic and deterministic processes to the collaboration between IT and HR, and the choice between scarcity thinking and growth thinking: a complete overview of what AI Agents mean for the future of work.

Working life is about to change drastically in the short term. Not in ten years, not "someday", but faster than most organisations and professionals are accustomed to. The reason is simple and historical: for the first time, we have technology that can perform tasks independently. We call this technology AI Agents. And yes, you can see that as bad news. Or as very good news.

What are AI Agents?

AI Agents are systems that not only provide answers but actively carry out tasks. Think of planning tasks, gathering information, processing data, creating reports, drafting emails, performing analyses, monitoring processes, and independently following up on steps, often with minimal human input. Where "classic AI" primarily offers support, AI Agents go a step further: they take action.

AI agents, powered by generative AI, thus transition from language to action. These models understand text and independently perform tasks within business processes. This could include administrative actions, scheduling, or even recruitment tasks such as screening candidates. The potential of this technology is enormous: faster processes, more efficient workflows, and a different distribution of responsibilities between humans and machines.

First, the right question: is your process probabilistic or deterministic?

Many organisations want to deploy AI Agents to accelerate processes, reduce costs, and make employees more productive. However, before adding AI to your operations, you must first answer one fundamental question: is this process probabilistic or deterministic? Although this distinction sounds technical, in practice it is a strategic choice that determines whether AI truly creates value or introduces new risks.

AI Agents that use language models work probabilistically. This means they generate answers based on probability. Under the hood, they run on mathematics, statistics, and pattern recognition, predicting what the most logical or most probable answer is given the context. This is not a system error but the core of the technology. Like humans, language models weigh context, interpret tone and intent, and add nuance to their responses.

This is clearly evident in customer contact. When a customer sends an angry email, an AI Agent can recognise the tone and adjust its response accordingly, improving the customer experience. The same applies to marketing, where AI can generate content based on target audience, channel, and tone of voice, thus supporting campaigns faster and more consistently. In these types of processes, interpretation is desired, and context truly makes a difference.

Yet, not every process is context-driven. Many core processes within organisations are designed deterministically, meaning they operate based on fixed rules where the same input must always lead to the same output. Think of financial reports, tax audits, compliance checks, or legal validations. In these processes, you want certainty, not variation. The answer should not be "probably correct" but demonstrably correct according to predefined rules. If you deploy a probabilistic system in such an environment without additional safeguards, you introduce variation where consistency is required, potentially leading to errors, discrepancies in controls, or compliance risks.

The right question, therefore, is not whether you can deploy an AI Agent, but whether the process requires interpretation or strict rule application. If a process revolves around nuance, customer experience, and context, probabilistic AI can add enormous value. If a process is about fixed rules, controllability, and repeatability, then AI must be tightly embedded with validations, business rules, and controls, or a completely deterministic solution must be chosen.

What does this mean for your job?

If you see AI Agents as bad news, that's understandable. You might think:

• "I have to change enormously."
• "I have to reinvent myself."
• "I'm attached to my current job and role."
• "I find that change difficult."

These are normal reactions. Change affects certainty, identity, and routine. But you can also see AI Agents as good news. For the first time in history, there is realistic scope to dedicate your time to what you are truly trained for and where you add the most value.

Marketers: from Excel to strategy

Instead of losing time filling out Excel sheets and manual reports, a marketer can focus more on:

• marketing strategy and positioning
• target audience insights and creative concept
• channel choices and campaign optimisation
• brand building and growth

Analysts: from re-keying to enriching

Instead of re-keying, cleaning, and repeatedly generating the same data exports, analysts can spend more time on:

• enriching data insights
• testing hypotheses
• building better KPI structures
• advising teams and stakeholders

In short: less "work for the sake of work", more work with impact. This shift is not gradual; it's an acceleration. Teams that cleverly deploy AI Agents will deliver faster, analyse more keenly, and organise more efficiently. This changes expectations within organisations: what is considered "normal" in terms of output, speed, and quality is shifting.

The interesting thing is: in a world where AI and AI Agents play a more prominent role, knowledge and experience become more important than ever. AI can perform tasks, but you determine the goal. AI can provide options, but you choose what is appropriate in context. AI can generate output, but you monitor quality, nuance, and strategy. AI can accelerate, but you provide direction and responsibility. Value shifts from "doing a lot of manual work" to "being able to guide, assess, and decide well."

IT as the new HR?

Jensen Huang, CEO of NVIDIA, recently stated that IT is set to become the new HR department. A striking statement that gives pause for thought. In light of the rise of AI agents, the boundary between IT and HR indeed seems to be blurring.

HR departments have been tasked for years with optimally utilising talent. With the rise of AI, this responsibility does not change, but rather becomes more complex. Where HR previously only looked at human skills, they must now also consider what AI systems are capable of. For example, an AI agent can take repetitive administrative tasks off an employee's hands. This means that, in addition to the question of which employee is suitable for a task, HR must also determine how AI can support or even replace them. It's about finding the right symbiosis between humans and machines.

Traditionally, IT and HR have been separate worlds. But with the integration of AI into business processes, these departments are becoming increasingly interdependent:

• IT understands technology: they know what AI systems can do, where their limitations lie, and how they can be deployed.
• HR understands people: they know what employees can do, what their talents are, and how best to utilise them.

Together, these departments can develop a strategy in which humans and AI reinforce each other instead of competing. This requires joint strategic planning, education and training so employees are prepared for collaborating with these systems, and continuous evaluation, because what works today may be outdated tomorrow. Therefore, start the conversation between IT and HR: determine for each process which tasks remain with employees and where an AI agent can provide support. That is the basis from which everything else follows.

AI Agents in e-commerce: optimising for the new digital shopper

The way consumers find and buy products is fundamentally changing. Where e-commerce for years revolved around visibility, branding, and conversion optimisation, the focus is now shifting to something else: decisions made by AI Agents. Increasingly, consumers are asking AI assistants for the best product within their budget or for their specific situation. This AI compares, analyses, and selects. Not based on emotion or appearance, but on data, reliability, and consistency. This means e-commerce is shifting from marketing optimisation to agent optimisation.

1. Make your product data fully machine-readable

AI Agents do not read mood images or branding. They analyse structured data. Therefore, ensure that titles are clear and logically constructed, attributes are correctly and fully entered, specifications are standardised and consistent, and stock status and delivery times are current. This is the new SEO: the cleaner and more consistent your data, the greater the chance an AI agent will select your product.

2. See reviews as your reputation with the AI Agent

An AI agent primarily looks at reliability signals. Verified reviews carry significant weight: the number of reviews, the average score, the content and relevance of the feedback, and how you, as a seller, respond. Reviews are no longer just customer service, but an important ranking factor now that AI provides purchasing advice.

3. Ensure omni-consistency in your product information

An AI Agent checks information across multiple sources. If specifications on your own website differ from those on other channels, it undermines trust. Ensure that specifications are identical everywhere, product titles are consistently formulated, and important features do not vary per channel. Consistency is trust, and trust is decisive for AI.

4. Optimise for Answer Engine Optimization (AEO)

Consumers are increasingly using AI tools instead of traditional search engines. They ask a question in natural language and expect a concrete answer. Therefore, formulate content around real customer questions, ensure product information answers specific use cases, and work with structured data that AI can easily interpret. No longer about "being found in the search bar", but "being the best answer to the question".

5. Use AI Agents yourself in your operation

The shift to AI means you need to optimise for agents, and you can deploy them yourself, for example, for dynamic pricing, inventory management, review analysis, listing optimisation, and advertising optimisation. Start small: choose one process and implement one purpose-built agent that performs one task extremely well. E-commerce is thus shifting from being visible to humans to being reliable for machines. Tomorrow's winners are not necessarily the brands with the most beautiful visuals, but the companies with the cleanest data, the strongest reputation, and the most consistent information.

Why AI Agents must consider human behaviour

AI agents are taking an increasingly active role within organisations. While AI has long been primarily used to perform analyses and make predictions, we now see systems that autonomously formulate recommendations, make decisions, and directly interact with people. Precisely in this shift from analysing to acting, one fundamental problem becomes apparent: many AI agents are designed as if they operate in a world where people react more or less the same way to the same stimuli. That world does not exist.

Human behaviour can be modelled, but never fully captured in fixed rules. In practice, people assign different weights to the same factors. What is decisive for one person plays hardly any role for another. Two individuals with the same information and in the same context can therefore arrive at totally different choices, without either acting irrationally.

In data and models, we only see a part of what influences choices. Many individual motivations remain unseen: personal values, previous experiences, risk attitude, trust, or emotions that do not neatly fit into a dataset. In statistics, this is called unobserved heterogeneity: systematic individual differences that influence behaviour but are not directly observable. Modern models explicitly recognise this by not assuming a single decision-making process, but a distribution of preferences within a population.

Many AI agents do the exact opposite: they abstract away differences, optimise a single strategy, and learn from average patterns. As soon as agents work with real people, frictions arise. Recommendations that are valuable for one user evoke resistance in another. Nudges that help one person feel intrusive to another. Optimisations that seem efficient in the short term undermine trust and acceptance in the long term.

The central question for AI agents is therefore no longer what works best on average, but for whom something works, in what context, and why. This means that personalisation is not an extra layer added later, but a fundamental part of the decision-making process itself. AI agents designed this way make better decisions, hold up better in complex environments, and build more sustainable trust in interactions with people. The future of AI agents lies not in even more data or even more complex models alone, but in human-centred design.

The real choice: the spreadsheet or the soul?

The, sometimes justified, fear surrounding agentic AI is that it costs jobs and makes organisations more efficient but also colder. Technology takes over work, companies cut costs, profits rise. The classic automation story. But perhaps we're looking at it wrong.

Leaders who primarily see AI as a cost-saving machine reduce their role to spreadsheet management. They optimise margins, restructure FTEs, and present a tight quarterly result. That is not leadership with ambition, but management without a soul. A true leader does not start with the question: "Where can we cut?" but with the question: "Where do we want to grow? What more do we want to achieve? Which ambition dare we finally realise?" Agentic AI is not an glorified cost item; it is a capacity multiplier. It increases thinking and execution power. The question is therefore not how many people you can replace, but how much value you can add. Profit derived from contraction is finite. Profit derived from value creation is scalable.

Yes, you can reduce a support team from ten to two. But suppose those other eight people retrain and move closer to the customer, innovation, or the mission. Then the conversation shifts from "saving costs" to "increasing impact." Companies heavily focused on customer experience understand this. IKEA decides not to put fewer but more people in the store, precisely to enhance the experience. Sales organisations deliberately choose to have teams engage more often and more deeply with customers. Technology takes over repetitive work, allowing people to focus on relationships, creativity, and trust.

The same applies to the government. If agentic AI carefully takes over processes, with an emphasis on meticulous, ethical, safe, and sovereign operation, space is created. Space for customisation, for human contact, for a conversation that goes beyond a form. Agentic AI can enable less administrative pressure and more time for actual service: not fewer civil servants, but civil servants who are allowed to be civil servants again. Of course, all of this must happen ethically, safely, and responsibly. Transparency, human control, data security, and digital sovereignty are indispensable. But let's not narrow the debate to only risks and control measures. Perhaps the real contrast is not human versus machine, but scarcity thinking versus growth thinking.

Conclusion

AI Agents will drastically change working life in the short term. That is a fact. The question is not whether it will happen, but how you deal with it. Organisations that understand the distinction between probabilistic and deterministic processes deploy AI where it adds flexibility and intelligence, and protect their core processes where consistency is paramount. Organisations that enable IT and HR to collaborate find the right balance between technology and human talent. And organisations that use agentic AI to pursue greater ambitions instead of merely cutting costs will ultimately prevail. If you see AI Agents as a threat, it feels like a loss. If you see it as an opportunity, space is created: for focus, for craftsmanship, and for work that truly matters. And that is precisely why now is the time to adapt.

Want to start working with AI yourself? Check out our AI Agents e-learning or AI workshops for teams.

• Best AI tools
• AEO: discoverable in AI search engines
• Writing AI prompts