Sopra Steria Next estimates that fewer than one in five business decisions shaped or carried out by artificial intelligence has all three governance safeguards, pointing to an accountability gap as companies delegate more actions to software.
The consulting arm of Sopra Steria published the finding Oct. 20 in its CIO Compass report. Sopra Steria Next says fewer than 25% of AI-driven decisions already in production can be followed across their full path, audited and replayed at a standard comparable to critical information technology transactions.
The report extends AI sovereignty beyond where data is stored or which infrastructure runs a model. It treats responsibility for automated outcomes as another part of the question: organizations need to establish how decisions are made and who can answer for them. A proposed nine-dimension framework includes a five-layer stack: data, models, platforms, application programming interfaces and decisions. It asks companies to assess their operational control, how easily a layer could be replaced and where accountability sits.
Sopra Steria Next says generative and agentic AI increasingly prioritize tasks and initiate or carry out decisions, making ownership more consequential when systems act beyond recommendations. The report also identifies “Shadow AI” and weak model governance as risks to traceability and cost control.
The report treats the ability to intervene as part of governance. “A system that cannot be stopped in real time is not governed: it is merely observed,” said Valérie Thiblet, Partner Data & AI at Sopra Steria Next.
It recommends a Chief Data & AI Officer to clarify responsibility for AI decisions. The proposed role is intended to balance oversight with innovation, sovereignty, compliance and performance.
