When artificial intelligence disrupts Swiss tax

Artificial intelligence is no longer just a topic of digital strategy. It is now entering the tax field. In March 2026, the National Council adopted a postulate requesting the Federal Council to examine the impact of AI on the Swiss tax system, particularly the risk of tax base erosion linked to increasing automation.

For a Swiss SME, this debate may seem distant. It is not. AI is already changing the way of producing, selling, employing and accounting for costs. Yet tax precisely follows these flows: profits, salaries, investments, VAT, intellectual property. If value shifts from human labour to algorithms, licences or intangible assets, traditional tax bases are bound to be re-evaluated.

A political signal in a very decentralised tax system

Switzerland does not have a single tax controlled from Bern. Its system is based on three levels: Confederation, cantons and municipalities. This architecture gives the cantons significant tax autonomy and explains why two comparable companies can be taxed differently depending on their location.

The Confederation, for its part, can only levy taxes within the limits set by the Federal Constitution. The two main federal taxes, the direct federal tax and the value-added tax, must be regularly approved by the people and the cantons to be maintained. In other words, a profound adaptation of the Swiss tax system is never a simple technical adjustment: it involves institutional, political and cantonal balances.

This is what makes the postulate adopted by the National Council important. It does not create a new tax on AI. It requests an analysis. But this analysis can pave the way for broader reflections: what happens if an increasing part of the added value is produced by automated systems? How to preserve public revenues if forms of work, profit and investment evolve? And how far should tax be adapted without hindering innovation?

Geneva shows that AI has already entered SMEs

The tax debate is not theoretical. In Geneva, a survey conducted in the first quarter of 2024 among more than 200 SMEs showed that 54% of them had already integrated AI systems into their operational processes. Adoption is particularly high in financial services, at 76%, in IT, at 71%, as well as in communication, marketing and real estate, each at 67%.

These figures signal a shift. AI is not reserved for large platforms or research laboratories. It is found in tools for customer relationship management, data analysis, content generation, invoicing assistance, administrative automation or commercial forecasting. For a fiduciary, this means that discussions with clients will no longer only focus on purchasing software, but on a transformation of work organisation.

The Geneva study also indicates that 68% of the companies surveyed consider AI important or very important for their activity. And 72% expect it to contribute to increasing their revenues within one to three years. From a tax perspective, this expectation is central: if AI increases revenues, it can increase taxable profit; if it replaces certain tasks or reduces certain costs, it can change the very structure of the income statement.

The observed results go in this direction. Among the companies that have adopted AI, 73% have seen an increase in their productivity and 43% an improvement in the quality of their products or services. For the manager, the subject is therefore not only fiscal: it touches on margins, prices, delivery capacity and competitive positioning. But these gains must be documented, measured and correctly translated into the accounts.

When value leaves the payroll

The heart of the tax debate lies in the tax base. This notion refers to the base on which a tax is calculated: income, profit, taxable turnover, wealth, payroll or other economic indicator. If AI transforms the way value is created, it can also transform the bases on which the State levies tax.

The risk raised by the postulate adopted in March 2026 concerns in particular the possible erosion of the tax base due to automation. If certain tasks performed by people are replaced by automated systems, labour income may evolve. This can have indirect effects on personal income tax and, more broadly, on mechanisms related to salaries and social insurance.

For an SME, this point must be addressed without caricature. AI does not automatically mean job cuts. It can also shift skills: less manual entry, more control, analysis, customer relationship or tool supervision. But as soon as an employer reorganises functions, outsources certain tasks to platforms or transforms the remuneration of its employees, the fiduciary must look at the consequences on salaries, social charges, salary certificates and the coherence between contracts, payslips and accounting.

The subject is also managerial. The Geneva survey indicates that 40% of companies report not having the necessary internal skills and resources to implement and manage AI systems. The lack of knowledge about AI and its potential applications is cited as the main obstacle by 44% of companies. An SME that invests in AI without internal skills risks underestimating the costs of training, maintenance, data governance and human control. These costs are not neutral fiscally or accounting-wise.

Subscription, intangible asset or research project?

In the accounts of an SME, AI can take several forms. It can be a subscription to an online service, a tool integrated into existing software, a bespoke development, an internal project or an element related to intellectual property. The accounting and tax treatment then depends on the real nature of the expense, its usefulness over time and the way the company controls or not the tool.

This distinction is very concrete. A monthly invoice for an AI service used by the sales team is not read like an internal development intended to automate a strategic process. A foreign licence, a consultancy service, a configuration or a specific development can also raise different VAT questions depending on the supplier, the place of service, the company's VAT status and the actual use of the service.

Prudence consists in documenting decisions from the start: why the tool is purchased, who uses it, which process it replaces or improves, which costs are recurring, which costs relate to implementation, and which benefits are expected. This documentation facilitates the work of the fiduciary, but also the justification in case of control or request for clarification.

The taxation of innovation already exists in Switzerland in certain forms. The Tax Reform and AHV Financing, implemented in 2020, notably introduced the patent box, which allows favourable tax treatment of certain income from intellectual property to encourage research and development. AI can, in some cases, fit into this universe. But the concrete application depends on legal conditions, the canton and the exact nature of the rights or income concerned. A case-by-case analysis remains essential.

The robot tax remains a hypothesis, not a rule

The public debate sometimes evokes the idea of taxing robots or autonomous systems. Experts and specialised firms, including PBM, have raised the possibility of a specific tax status for autonomous robotic entities that would generate economic resources. This type of proposal illustrates the difficulty: the digital economy produces value with assets that do not resemble a traditional factory, a salaried employee, or a simple office software.

For now, an SME should not reason as if a Swiss robot tax were already in force. However, it would be wrong to ignore the signal. The legislator is questioning the ability of the current system to correctly capture the value created by automation. Depending on future choices, technology-intensive companies could see the rules affecting the deduction of investments, the taxation of intangible income, VAT on digital services or the way of assessing the economic substance of an activity evolve.

In the immediate term, the best reflex is less spectacular: keep readable accounts. An SME that mixes software subscriptions, consultancy services, internal developments and salary costs in too general accounts loses visibility. It also deprives itself of useful information to manage its profitability. AI must be monitored as a real business project, with a budget, responsible parties, expected effects and cost traceability.

The role of the fiduciary then becomes more strategic. It can help to ask the right questions before the investment: what impact on margins? What effects on employment and salaries? What VAT treatment for foreign or Swiss suppliers? What separation between current expenses and investments? What elements to document if the tool creates or exploits intellectual property? These answers are not standardised; they depend on the activity, the canton, the business model and the level of risk accepted by the company.

Artificial intelligence does not yet overturn Swiss taxation. But it already forces SMEs to look differently at their value flows. In a country where tax is constructed between Confederation, cantons and municipalities, the evolution will probably be gradual. Companies that have documented their uses, understood their costs and integrated their fiduciary early in the projects will be best placed to adapt if tax law follows, tomorrow, the speed of algorithms.