AI is intended to make employees more productive and free up time for more demanding tasks. However, when using AI itself becomes a performance benchmark, it can create the wrong incentives. The study “The Pulse of Work in 2026” shows that 42 percent of employees in Germany, Austria, and Switzerland have already used AI-generated results even though they did not trust them.
In theory, the use of Artificial Intelligence (AI) in companies is intended to help employees become more productive – and thus free them up to tackle more challenging tasks. However, it seems that the use of this technology is sometimes misinterpreted as a ‘performance target’ in itself, creating a wrong incentive: Employees then use AI simply because it is expected of them, and adopt its results even when they do not trust them. The fact that this risk is indeed significant is demonstrated by the recent study “The Pulse of Work in 2026” by GoTo and Workplace Intelligence: 42 per cent of employees in Germany, Austria and Switzerland have already used AI-generated content, even though they considered it to be of poor quality or suspected it contained errors or fabricated information.
If the use of AI is taken into account in performance appraisals, the proportion rises to as much as 59 per cent. What appears to be a technological problem thus becomes a question of leadership and organisation. For this study, carried out between November 2025 and January 2026, a total of 2,500 permanent knowledge workers and IT decision-makers were surveyed – including respondents from Germany, Austria, Switzerland, the USA, Canada, the UK, Ireland, India, Mexico and Brazil. Companies should assess whether their AI strategy takes into account not only efficiency but also quality, accountability and human judgement. In its statement, GoTo sets out five steps to help with this.
AI-generated ‘workslop’ poses a problem in terms of quality and trust
Generally speaking, AI brings measurable benefits in the workplace: 88 per cent of those surveyed said that the technology had already helped them. On average, they estimated that they saved 1.5 hours a day. However, these time savings lose their value if others then have to spend a great deal of time checking and correcting the results.
This very problem is already becoming apparent today: 71 per cent reported that checking AI-generated work produced by colleagues creates extra work for them. More than half feel that this impairs their own productivity. An equal number stated that flawed AI content raises doubts about others’ competence and fosters resentment within the team.
‘Workslop’ – that is, AI-generated content that looks professional but lacks substance – thus becomes not only a quality issue but also a trust issue. GoTo sets out five steps for the responsible use of AI below:
- Jointly assess usage and benefits
Before companies draw up new AI guidelines, they should gain a realistic picture of the current situation: “Which AI applications are already in use? For which tasks? What data do employees input? And is AI perhaps being used in unauthorised ways?”
This assessment should not be carried out solely for the sake of monitoring. The key is to define the intended added value and to determine for which decisions human judgement remains indispensable.
Without this strategic integration, AI remains an isolated tool in the hands of individual employees. Whilst this might result in isolated time savings, it would not lead to a company-wide productivity gain. - Clear guidelines rather than long lists of prohibitions
An AI policy will only be effective if staff are aware of it, understand it and can apply it in their day-to-day work. A lengthy set of rules that mainly lists prohibitions is unlikely to do justice to the dynamic nature of the technology.
Modern governance must be flexible and regularly adapted to new tools, risks and application scenarios.
The guiding principle here is ‘empowerment rather than punishment!’. Rules should enable staff to use AI safely and confidently, rather than forcing them into a grey area out of fear of consequences. This requires ongoing dialogue between the IT department, the legal department, management and the relevant specialist departments. - Secure tools rather than ‘shadow AI’
Guidelines alone are not enough to prevent ‘shadow AI’. If authorised applications were complicated, slow or unsuitable for the task at hand, employees continue to turn to private and often free alternatives.
This increases the risks to data protection, compliance and intellectual property.
Companies should therefore provide secure ‘tools’ that are just as user-friendly as freely available options and can be seamlessly integrated into existing IT systems. - Firmly establish accountability and quality control
AI can generate content, but cannot take responsibility for its consequences. Companies must therefore clearly define who checks and approves the results and who takes corrective action in the event of errors. Human oversight must not be optional, particularly when it comes to external content, customer communications and decisions relating to legal, financial, personnel or security matters.
However, this responsibility should not rest solely with individual staff members. When managers call for the use of AI, they must at the same time provide suitable tools, realistic quality standards and effective review processes. AI governance is therefore not a peripheral technical or legal task, but a management responsibility.
Quality assurance also requires an open ‘culture of error’ – employees must be able to question dubious results and report problems without being penalised for doing so. An AI output that sounds plausible should never carry more weight than a well-founded expert assessment. - Continuously review skills and performance
A guideline on its own cannot prevent so-called AI hallucinations or incorrect decisions. Employees need to learn to check AI outputs for facts, sources, biases and missing context. Practical training should therefore be based on real-world workflows and provide role-specific guidance on where the technology helps and where its limitations lie.
This is not primarily about in-depth technical knowledge. What is crucial is the ability to control AI precisely, to evaluate its results critically and to consciously apply human judgement. Creativity, communication and professional expertise are not rendered redundant by AI; rather, they become even more important.
Measuring success must also reflect this approach. The number of tools used or the amount of content generated says little about the actual benefits. According to GoTo, what is more relevant is whether tasks are completed more quickly without increasing the error rate or the amount of follow-up work required. Equally important is the question of whether the outcomes are better for customers and whether employees’ workloads are actually reduced. The policy itself should be regularly reviewed and refined in the light of these findings.
Clear responsibilities, reliable tools and binding quality standards are what make AI valuable
The key question is no longer whether employees use AI, but under what conditions they do so. Anyone who promotes the use of the technology – or even makes it a criterion for assessment – must at the same time establish clear responsibilities, secure tools and binding quality standards.
An effective AI policy is therefore not a bureaucratic obstacle to innovation – it lays the foundations for translating time savings into genuine economic added value.
“Only when employees know what is permitted, how results are verified and when their own judgement takes precedence can AI play to its strengths – without jeopardising the quality of work, data or trust.”
Key findings from the DS editorial team
- Just as a traditional workshop has ‘orgware’ – i.e. guidelines for material tools such as work instructions and accident prevention regulations – there should be clear rules on the use of AI to ensure it can be applied in a way that is both low-risk and useful.
- The use of AI in businesses goes far beyond the technological dimension and touches on key aspects of leadership and organisation; it therefore falls within the overall remit of decision-makers.
- Just as accidents involving physical tools can never be entirely ruled out, however careful one may be in day-to-day operations, a clear framework for dealing with problems arising from the use of AI must be established in the context of emergency and recovery planning.
- In this context, the so-called ‘culture of error’ plays a key role – although it would be more accurate to refer to a ‘culture of learning’: Errors should be identified and reported at an early stage in order to prevent damage and learn from them for the future.
- Senior management is responsible for providing staff with all the necessary tools to use AI safely and to achieve high-quality results – the introduction of ‘shadow IT’ must be prevented in general, and therefore also in the context of AI.
Conclusion
Safe and effective AI use requires clear rules, appropriate technical and organisational frameworks, and an open learning culture. Responsibility lies with leadership, which must establish suitable tools, processes, and guidelines while preventing shadow systems. The key is not only to avoid risks, but also to identify errors early, learn from them, and continuously improve the quality of AI use.
Further Information
GoTo
Hallo. Wir sind GoTo.
GoTo
The Pulse of Work in 2026
DATENSICHERHEIT.DE, 11.08.2026
KI-Vorfälle nur Symptome – IT-Sicherheit muss auf Identitäten fokussieren
DATENSICHERHEIT.DE, 07.08.2026
KI als Produktivitätstreiber: Beschäftigte gewinnen bis zu acht Stunden pro Woche
DATENSICHERHEIT.DE, 29.07.2026
PINKTUM-Checkliste zur KI-Kompetenz für Unternehmen
