Generative AI and AI agents in business: from assistant to key player
Artificial intelligence (AI) is increasingly taking on more than just individual tasks within organisations. Systems are now able to plan multiple work steps, integrate digital tools and data sources, and carry out multi-stage processes semi-autonomously. Employees are increasingly taking on the roles of coordinators and supervisors. A new white paper from the Plattform Lernende Systeme shows how tasks, roles and organisational structures are changing – and what conditions companies must put in place to ensure the safe and productive use of AI agents.
Download the white paper (German)
In many companies, there is a significant gap between the initial applications of generative and agent-based AI and their productive deployment. Despite potential efficiency gains, many projects remain at the pilot stage. A renowned study shows that 95 per cent of the companies surveyed have so far been unable to demonstrate a measurable return on investment (ROI). A lack of reliability and controllability, high costs, and shortcomings in integration, governance and accountability make sustained operation difficult.
Meanwhile, potential applications span the entire value chain: generative AI analyses information, supports simulations and idea generation, and taps into organisational knowledge. Employees are shifting their focus from creating and searching to verifying and categorising. AI agents can also control experiments, process business workflows or adapt manufacturing processes.
From prompt to process
Generative AI, for example, generates text, code or analyses. AI agents go one step further: within defined limits, they can manage entire processes aimed at achieving objectives. They plan sub-tasks, delegate them where necessary, evaluate results and make adjustments. Employees define objectives and scope for action, monitor agent networks and intervene in the event of uncertainties, conflicting objectives or safety-critical decisions.
“When AI takes on tasks autonomously within a company, the skills that people need also change,” says Rahild Neuburger of Ludwig Maximilian University of Munich and a member of the ‘Future of Work and Human-Machine-Interaction’ working group at the Plattform Lernende Systeme. “They must be able to make informed decisions about what tasks to delegate to AI, critically evaluate its results and communicate with the AI. Specialist knowledge remains indispensable – complemented by experience, the ability to manage and delegate, and a sound understanding of what AI is capable of and where its limits lie.”
Autonomy requires clear boundaries
The more steps a system carries out autonomously, the more important it is to have transparent rules and effective means of control. Errors in agent-based systems can be amplified across process chains, whilst ambiguously defined solution paths complicate testing, certification and explainability. Cross-system access and the long-term memory of AI agents also raise questions regarding trade secrets, personal data, access rights and data deletion policies.
The authors of the paper recommend introducing AI alongside target scenarios for processes, competencies and work structures. This requires binding guidelines, an established quality management system, testing standards tailored to specific target groups, and the systematic development of staff competencies. Exceeding authorised scope should be prevented through safeguards and restricted access rights. Pilot phases should be used to assess the areas of application, data basis and process maturity of AI. This is because success is not measured solely by time savings, but also by factors such as process stability, quality of results, degree of innovation, trust and the actual reduction in workload.
“Human-centred design of AI means that automation is integrated into work in such a way that it relieves the burden on people, whilst also equipping and empowering them – and, moreover, creating new jobs. Only in this way can AI contribute to the sustainable development of productivity and value creation in companies,” emphasises Norbert Huchler, Institute for Social Science Research and member of the ‘Future of Work and Human-Machine-Interaction’ working group of the Plattform Lernende Systeme.
The vision for companies should be a complementary working system between humans and technology: humans set the purpose, objectives and limits and remain the final decision-makers, whilst AI agents take on the tactical execution and accelerate the achievement of objectives.
About the whitepaper
The white paper “Generative AI and AI agents in business: from assistand to key player” was authored by members of the “Future of Work and Human-Machine-Interaction” working group of Plattform Lernende Systeme. It is available for download (German) free of charge.
Further information:
Petra Brücklmeier Press and Public Relations
Lernende Systeme – Germany's Platform for Artificial Intelligence Managing Office | c/o acatech Karolinenplatz 4 | D - 80333 Munich
