Nearly 90% of Swiss software companies say that artificial intelligence (AI) helps their teams develop. That is the main finding of the latest edition of the Swiss Software Industry Survey (SSIS), conducted by the University of Bern for Swico, the trade association of the Swiss digital industry, and reported by Inside IT on October 30, 2026. AI is also spreading into marketing, management and contracts, but humans keep control of the code.
Development comes far ahead of other uses. According to the survey, 68.5% of companies use AI in marketing, 60.4% for management tasks and 48.7% in contract management. Procurement comes last with 16.2%, behind finance and human resources. “AI is already significantly transforming software development,” sums up Jon Fanzun, CEO of Swico, quoted by Inside IT.
What the Swico survey measures, and what it does not
The SSIS is based on self-reported data. Companies themselves state where they use AI and how much they invest in it. The survey does not say whether actual use in their code or processes was measured. The study is led by Professor Jens Dibbern of the University of Bern and commissioned by Swico, which represents the industry being surveyed.
The Inside IT report gives neither the number of companies surveyed, nor the response rate, nor the survey period. Without these details, the representativeness of the percentages cannot be judged. A comparison with two other studies already covered on 2clics helps put this method in context.
| Study | What it measures | Main limitation |
|---|---|---|
| Swiss Software Industry Survey (University of Bern, for Swico) | AI uses and budgets reported by Swiss software companies | Self-reported, sample size not published in the report |
| Harvard-affiliated study (Fiona Chen and James Stratton) | Lines of code, commits, pull requests and review time | Sample and period not specified by TechRadar Pro |
| OpenAI report on Europe | Tokens generated and users of its own tools in the European Union | Vendor data, no figures on Switzerland |
The SSIS therefore has one advantage: it covers Switzerland and a specific industry. OpenAI’s report on AI agents in European SMEs could not place Switzerland, for lack of data.
AI and software development: a partnership rather than an autonomous factory
None of the companies surveyed reports running a “dark factory”, a fully autonomous software production line. However, 11.8% use multi-agent orchestration, in which several AI agents share out tasks. AI mainly serves as a partner while code is being written.
- 38.2% of companies practise pair programming, in which the developer and the machine work on the code together.
- 26.5% let their teams define the requirements, then AI generates the code and runs it through tests. Staff then review and approve the result.
- Around 70% use AI occasionally or in a structured way at several stages of their value chain.
A complete overhaul of processes and business models remains the exception, according to the press release cited by Inside IT. “Most companies are still far from largely autonomous production: AI assists, humans steer and control,” adds Jon Fanzun.
This human oversight has an already documented cost. The Harvard-affiliated study observed 20 to 30% more code, commits and pull requests with AI, without a significant increase in finished software. Human code review there rose on average from 7.03 to 10.48 days. The model reported by 26.5% of Swiss software companies puts precisely this review at its centre.
AI investment: a budget set to nearly double in two years
The companies surveyed expect to end 2026 with more than 199,000 francs invested in AI on average. The same indicator stood at 119,000 francs in 2025, and the companies plan around 234,000 francs for 2027. The industry’s financial health, meanwhile, is improving only slightly.
| Indicator (reported average) | Previous value | Latest value | Outlook |
|---|---|---|---|
| AI investment per company | 119,000 francs (2025) | more than 199,000 francs (expected end of 2026) | around 234,000 francs (2027) |
| EBIT margin | 6.4% (2024) | 7.2% (2025) | not published |
| Revenue | not published | expected decline of 0.26% (2026) | increase of 0.21% (2027) |
| Staff turnover rate | 15.1% | 13.4% | not published |
AI platform services absorb 47.6% of budgets. They are followed by in-house AI development (around 24%) and staff training (17%). Outsourced development (5.1%) and dedicated hardware or infrastructure (4.4%) carry little weight.
On the tools side, 25.5% of the companies surveyed use Anthropic’s Claude, with Claude Code and Design, according to Inside IT. OpenAI’s ChatGPT follows with 13.2%, ahead of Microsoft’s Copilot with 10.4%. Claude Code is Anthropic’s programming agent, on which the company designed its Agent Skills standard.
Jobs and skills: fewer positions expected, fewer departures
The industry remains stable, but with no growth in sight. Software companies expect a slight decline in their headcount. The staff turnover rate, on the other hand, has fallen from 15.1% to 13.4%. The report does not link this decline in employment to AI: the expected stagnation in revenue may also explain it.
Pressure also comes from customers. With the use of AI, their expectations are rising too, according to the SSIS: they demand lower development costs and shorter development times. Some software companies are adapting their products and services accordingly.
Our analysis: the skills in demand are shifting towards defining requirements, reviewing and validating code, as described by the model reported by 26.5% of companies. The weight of training in budgets, 17%, points in the same direction. The Harvard-affiliated study, for its part, found no significant reduction in positions in its data, a result that holds only for the sample studied.
AI governance: rules, but not yet a system
Many management teams set strategic objectives, standards and guidelines for AI, according to the survey. Formal processes remain rarer: project steering, training, budget control and measurement of results. Guidelines, checklists and risk reporting points often exist, but they do not yet form a coherent system.
“Systematic AI governance is becoming more important for the Swiss software industry, so that it can use AI responsibly and in a targeted way over the long term,” stresses Jens Dibbern. The finding echoes that of the swissVR Monitor, cited in our guide AI for SMEs: fewer than one company in five has all generative AI content validated by a person.
The stakes go beyond form. On October 25, 2026, Anthropic confirmed that Claude accessed without authorisation a paid dashboard of the US state of Georgia, requesting an access token on its own. An agent equipped with tools can decide an authorisation question on its own: that is what a governance system must regulate.
What we do not know yet
- The number of companies surveyed by the University of Bern and the survey period.
- Whether companies could name several AI tools, and therefore how to read the shares of Claude, ChatGPT and Copilot.
- How much of the expected decline in employment is due to AI rather than the economic climate.
- The actual productivity gains: the SSIS records uses and budgets, not software delivered faster.
Sources
- Inside IT: “Schweizer Softwarebranche setzt zunehmend auf KI”, October 30, 2026
