Swiss Software Survey Finds Nearly Nine in Ten Firms Build With AI, but Only 11.8% Use Agent Orchestration

Almost nine in ten Swiss software companies now use artificial intelligence to build software, but very few have handed development to teams of autonomous agents, according to the new Swiss Software Industry Survey reported by Netzwoche on October 27. The study, conducted by the University of Bern on behalf of the ICT industry association Swico, drew up to 115 responses per question from software firms in Switzerland. More than 88 per cent of respondents said they use AI in software development, most often for writing code, followed by design and testing, and least often for integration and planning.

Humans remain in charge. Thirty-eight per cent of companies describe their approach as pair programming, with people and AI developing solutions together; 27 per cent use spec-driven development, where staff define requirements and then review AI-generated code; and 23 per cent call it assisted coding, with AI making suggestions. Only 11.8 per cent use multi-agent orchestration, and none reported a fully AI-run “dark factory.” Outside coding, firms use AI most in marketing and sales (68.5 per cent), management tasks (60.4 per cent) and contract management (48.7 per cent).

Spending is climbing. Respondents invested an average of CHF 119,000 in AI in 2025 and expect CHF 199,000 in 2026 and more than CHF 234,000 in 2027. Over 47 per cent goes on AI platform services and licences, consistent with proprietary tools such as Claude, ChatGPT and Copilot being the most used, while 23.7 per cent flows into in-house AI development.

What it means in practice

The survey suggests the industry is still early in its transformation: over 69 per cent use AI in isolated or structured process steps, 17.8 per cent say it has fundamentally changed existing steps, and fewer than 9 per cent put it at the centre of their business model. Governance is the weak spot, with guidelines and checklists often not yet joined up into a closed system. Study lead Jens Dibbern said systematic AI governance is gaining importance. The findings fit a wider picture of steady adoption, from SMEs driving a 32 per cent surge in Swiss AI job ads to Zurich startup Klarent raising money for repeatable AI software testing.

Customers, meanwhile, expect faster development and cheaper software thanks to AI, while vendors mainly hope for efficiency gains. With the average EBIT margin up from 6.4 to 7.2 per cent but revenue forecast to slip 0.26 per cent in 2026, that gap in expectations may shape pricing talks next year.