EY’s Post

Trust is certainly a prerequisite for AI in the public sector, but I wonder if the defining issue is something even more fundamental. Every significant technology shift changes the relationship between institutions and the people they serve. AI raises the stakes because it doesn't just automate processes—it redistributes judgment. Once that happens, trust is no longer built by the technology itself, but by the quality of the decisions surrounding its use. The organizations that will lead this transition may not be those that deploy AI the fastest, but those that can redesign governance, accountability, and decision-making without weakening public confidence. That is a far more complex transformation challenge than a technology challenge. Perhaps the real question is not whether citizens will trust AI, but whether institutions can evolve quickly enough to remain trusted in an AI-enabled world.

Great perspective. As AI continues to evolve, the real differentiator will be how organisations translate adoption into measurable outcomes. Trust matters, but so does the ability to embed AI into decision-making and create value at scale.

Having worked in the NHS/public sector during my career before EY, the challenge resonates strongly with me. Rigid policy, structure and hierarchy doesn't lend itself to AI transformation - both with internal organizational structures and public services/products. The one thing I do remember is the quality of the people however, and their loyalty to the public that they serve - they can help drive the transformation if leadership gave them the means, information and opportunity.

Trust is often the factor that determines whether new technology is embraced or resisted. As AI adoption accelerates, helping people build confidence in how it's used will be increasingly important.

I agree, and I’d go one step further. Readiness is not a test government can complete before it starts; it is built through disciplined, practical delivery. Buying licences creates access. It does not create organisational capability. Public value appears when AI is attached to a real problem, the process is redesigned around it, reliable knowledge and proportionate controls are built in, and the people doing the work have the authority to test, challenge and improve it. My strongest lesson has been that adoption follows evidence and ownership. Give frontline colleagues a bounded use case, a safe environment and genuine agency; measure quality, time, risk and user outcomes; then scale and reuse the patterns that work. Trust grows from seeing the system perform, and from knowing where human judgement remains accountable. The public sector already understands accountability. The missing piece is often an operating model that connects experimentation, governance, skills and delivery. The objective cannot be “more AI”. It must be demonstrably better public services, supported by an internal capability that keeps learning.

Trust is the right frame. In public sector work, AI fails the same way large ERP programs fail: weak inventory of what is actually running, unclear ownership when the model is wrong, and procurement language that buys a pilot instead of an operating capability. Skills and infrastructure matter, but accountability is what turns AI from a press release into a service citizens can rely on.

Worth adding that trust in public services behaves differently from trust in a market, and it changes what ready has to mean. In a market, low trust is self-correcting: customers leave, revenue signals the failure, the vendor responds. A citizen cannot switch benefits agencies. No exit means no churn signal, so a public-sector AI failure does not show up as lost revenue. It shows up years later as an ombudsman finding or a public inquiry, by which point the decision logs either exist or they do not. That has a practical consequence procurement usually ignores: in government, the evidence that a decision was lawful and explainable has to be manufactured at the moment of the decision, not reconstructed after the complaint. Legacy systems are the visible problem. Undocumented discretion is the expensive one. So when a public body buys AI today, is anyone asking the supplier the question that will matter in five years - can you reproduce on demand why this specific citizen received this specific outcome?

A timely read. We see this daily at Syrosoft-organizations get caught up in the hype of what AI can do, and forget to define what it should do. Whether you are leading a government agency or a mid-market enterprise, deploying AI without clear governance just creates operational risk. Trust is earned when leaders act as the synthesis layer—ensuring these tools enhance human judgment rather than replacing it. You have to build the guardrails before you hit the gas.

Non basta implementare l’AI, bisogna meritare la fiducia che la rende davvero utile. La sicurezza, la protezione dei dati e la responsabilità non sono dettagli tecnici, ma condizioni di delega. Senza infrastrutture moderne e senza un approccio trasparente, l’AI rischia di crescere in un terreno che non è pronto a sostenerla.

Well said. The challenge isn't buying AI, it's building the trust, skills, and modern infrastructure needed to make it deliver real public value. Technology is only one piece of the transformation.