What Does It Take to Trust AI Before It Goes Live?
Mon 3 Aug 2026
This is the final week of this month’s AI in Practice theme, focused on building the skills and trust needed to move AI into production.
The answers differ by market. In the UK, the focus is on ownership and accountability. In Germany, readiness is treated as ongoing work across data, engineering, security, and the business. In France, the question is whether AI can be made understandable, accessible, and trustworthy enough for people to use with confidence.
UK: Ownership Has to Be Visible
The UK discussion starts with changes inside government rather than the enterprise itself.
A recent reorganisation has moved AI closer to the centre of government decision-making, with a new Prime Minister’s AI Taskforce focused on delivery and AI Minister Kanishka Narayan attending Cabinet. The AI Security Institute has also moved closer to the centre.
For enterprise leaders, the useful signal is the structure around those decisions.
Responsibility for AI delivery is becoming more explicit. Adoption and governance are being brought closer together, while accountability can be more frequently traced to named leaders.
That reflects questions many organisations are facing as AI moves beyond isolated experiments: who owns the outcome, who governs the system, and who has the authority to intervene when something goes wrong?
It also raises a skills question. Trust depends on people who understand what the system is meant to do, how to challenge it and when to step in.
Germany: Readiness is Continuous Work
The German perspective challenges the idea that organisations pass through a neat preparation phase before launching AI.
Artur Yatsenko, Director of Data Engineering at Urban Sports Club, described readiness as “continuous discovery” around the resources, expertise, and capabilities an organisation already has.
That means data quality, governance, and collaboration cannot simply be completed and signed off before launch.
Naveen Kanneganti, a data, analytics, and AI leader, argued that engineering, data, security, and business teams need to make decisions about value, risk and feasibility together from the beginning rather than reviewing them separately at the end.
The same applies to cost.
“AI cost is not only token consumption,” said Kanneganti.
Compute, data movement, orchestration, monitoring, integration, security, testing, and human review all contribute to the economics of operating AI at scale.
Readiness, from this perspective, is not a box to tick. It is the ability to keep operating AI reliably, securely, and economically as the system changes. That also depends on teams having the skills to work across those functions together.
France: trust depends on understanding and inclusion
The French perspective, drawn from Gloria Fataki’s interview with ONU Femmes France, brings the question down to the user level.
Fataki said AI can help translate medical jargon into plain language, support patients preparing for appointments, and help them understand where specialists disagree. But she was equally clear that AI should not replace the doctor. Its role is to reduce the gap between people who already know how to decode technical language and those who do not.
That same logic applies to design.
Fataki argued that accessibility is not only a question of technology. Tools need to be simple, available in several languages, and adapted to different levels of digital literacy. They also need to be designed with the women who will use them, not for them.
She linked that to data quality as well. If women are under-represented in the data, the outputs will reproduce the same blind spots the technology is supposed to address.
For Fataki, the safeguards are clear: transparency, protection of personal data, explainability, and human supervision. AI remains an aid to decision-making, with responsibility staying with health professionals.
That makes trust a practical question. It depends on whether people can understand the system, whether the data behind it reflects the people it is meant to serve, and whether it still leaves room for human judgement.
Trust is Built Around the System
Across the three markets, technical capability is only part of the picture.
The UK highlights ownership and accountability. In Germany, readiness is treated as ongoing work across data, engineering, security, and the business. In France, the test is whether AI can be made understandable, inclusive, and safe enough to support people in practice.
Together, they suggest trust is built through the people, skills, and operating decisions around AI, not just through the model itself.
That may be the more useful definition of AI readiness: whether an organisation has the people, processes and decision-making discipline to keep asking if an AI system remains useful, safe and worth operating.
