AI at work: Why companies need to assess trust
4 min read
Updated 3 November 2026 · Originally published 17 June 2026
More than half of companies in Germany used AI in their business processes in May 2026, according to the ifo Institute. The adoption rate alone does not reveal what employees and customers think about these applications. This presents management teams with a practical task: they need to explain how AI is used, gather feedback and assess which transparency obligations apply to each application.
Key points at a glance
- AI use has risen. In the ifo survey of May 2026, 54.5 per cent of companies used AI in their business processes. The figure was 40.9 per cent the previous year.
- An adoption rate does not measure trust. Companies can ask employees and customers directly which applications they understand and where they need clarification.
- The AI Act governs specific transparency cases. The applicable obligation depends on factors including the system, the content and whether the company acts as a provider or deployer.
What the adoption rate tells us
In May 2026, the ifo Institute surveyed companies on whether they used AI in their business processes. The proportion stood at 54.5 per cent, up from 40.9 per cent the previous year. The survey therefore shows that use has become more widespread. On its own, it does not show how much employees trust the applications or whether that trust has risen or fallen over the same period.
Individual companies can therefore benefit from taking stock: Which tasks does AI already perform? Who uses the systems? Which people are affected by their results? Only on this basis can companies ask targeted questions about experiences and concerns.
Assessing trust within the company
When introducing a system, management and the project team can review its intended use together with the employees affected. This includes the system’s purpose, the data used, the application’s limitations and who checks its results. If an application drafts replies to customers, for example, it should be clear who reviews and corrects them before they are sent.
A short survey or a conversation with the teams involved can reveal which questions remain unanswered. Useful questions include: Is it clear what the system is used for? Do employees know which data they may enter? Can they report errors and request a review of a decision? Feedback should be recorded and collected again after changes to how the system is used.
For applications involving customer contact, the external perspective also matters. Companies can check whether the process is understandable, how customers can reach a person and whom they can contact if they receive incorrect information. These checks are management recommendations; their effect on project success cannot be quantified from the ifo survey.
When the AI Act requires transparency
Since 2 August 2026, the transparency obligations in Article 50 of the AI Act have applied to certain AI systems and content. They do not impose a general obligation to disclose every use of AI. The specific use case and the company’s role determine which rules apply.
Providers of systems intended to interact directly with people must design them so that the people concerned are informed that they are interacting with AI. An exception applies, for example, where this is obvious from the circumstances. Deployers must disclose, among other things, that certain deepfake content has been artificially generated or manipulated. The regulation also requires disclosure for AI-generated text on matters of public interest, but provides an exception where the text has undergone human review or editorial control and editorial responsibility has been assumed.
A separate obligation requires providers to mark certain generated content in a machine-readable format. For systems placed on the market before 2 August 2026, a limited transition period applies until 2 December 2026. Companies should record whether they are a provider or deployer for each application and which specific rule applies.
Questions before the next use
Before approving an application, the responsible team can document its intended benefits and limitations. This includes the data permitted for use, the person responsible for checking results and a way to report errors. For external services, companies should also establish which data is sent to the provider and which contractual terms govern its processing.
For communication, a simple set of questions can help: What must affected people be told under the applicable rules? What additional information would help them understand the process? And who answers their questions? Such an overview makes clear where legal obligations end and where the company wishes to provide further information voluntarily.
Frequently asked questions
Does the ifo survey demonstrate a trust gap?
The survey measures AI use in business processes. On its own, it does not establish a comparable trend in trust.
Must every use of AI be disclosed under the AI Act?
No. Article 50 sets out obligations for certain systems and content. The applicable requirement depends on the individual case and the company’s role as a provider or deployer.
How can companies gather feedback on AI use?
They can ask affected employees and, where appropriate, customers about their understanding of how AI is used, their open questions and their experiences. Errors and suggestions for improvement should reach a responsible team or individual.
Read more on MyBusinessFuture
MyBusinessFutureEU AI Act from August 2026: What SMEs must label nowMyBusinessFutureRegulator halts an AI model: The lesson for SMEsMyBusinessFutureThe real cost of generic AI-generated textMore from the MBF Media Network
cloudmagazinFrontier model taken offline by regulatory order: The architecture lessonDigital ChiefsWhen an AI model disappears overnight: Why CIOs need a plan BSecurityTodayDeepfake CEO fraud: How executives can defend against voice cloningImage source: AI-generated (June 2026), C2PA certificate embedded in the image
Translated from the German original using artificial intelligence. The German version is authoritative.
