Triathlon Group’s AI Study 2026
Triathlon Group’s Generative AI Study 2026 assesses how Sweden’s largest manufacturing companies are moving beyond general purpose GenAI tools such as ChatGPT & Copilot, with emphasis on what enables or prevents collaboration with GenAI startups. The study is based on in-depth interviews with decision makers from manufacturing companies and leaders of prominent GenAI startups, supported by a mapping of the startup ecosystem across five business functions.
You can access the full report here: Triathlon - Generative AI Study 2026
GenAI adoption is widespread, but tailored, partnership-driven solutions remain limited
Mainstream GenAI solutions are used by 100% of Sweden’s largest manufacturing companies, however only 45% of the surveyed companies have implemented customized, domain-specific GenAI services. For customized GenAI, manufacturing companies prefer to develop solutions internally, with 77% developing in-house and 23% via partnerships, reflecting priorities around control, customization, and data security.
Three main roadblocks stand in the way of successful GenAI partnerships
According to the study, three main roadblocks hinder GenAI partnerships: procurement processes, data security/compliance, and uncertain ROI. Lengthy, multi-stakeholder procurement processes in large manufacturing firms may render GenAI solutions obsolete even before the procurement process is complete. Security concerns push companies toward in-house builds, and ROI uncertainty, often as a result of limited solution awareness, stops many collaborations early.
Bridging the gap, actions for manufacturers and startups
To close the gap and enable partnerships, manufacturers should invest in structured exploration, streamline procurement, security, and compliance processes, and ensure alignment between startup maturity and internal AI readiness. Startups, in turn, must demonstrate clear differentiation, proactively address security and compliance requirements, and make integration and implementation as simple and transparent as possible.