Is your company drowning in data in spite of AI? Do you feel like, despite setting up expensive AI-pilots and giving all your employees access to Copilot (and turning a blind eye when they use Claude or OpenAI) nothing seems to be changing? Do you feel like pulling investment in the whole farrago, declaring it a pointless bubble, and just going to back to how things were before you got so distracted? Have you maybe already done exactly that?

If so, you’re not alone. A recent study by Bain [https://www.bain.com/insights/your-ai-budget-is-growing-your-returns-arent-heres-why/] found that despite sizeable investments of time and capital, most AI implementations are failing to generate either cost savings or productivity improvements, and the number one reason that they underperform is that companies cannot get reliable access to their own data – and this despite a decade of data modernisation projects that run to the hundreds of billions of dollars globally.

The study goes on to cite evidence, however, that this reason is somewhat spurious – and that companies that cite it are often revealing that they are going about AI implementation in the wrong way. The businesses that are deploying AI effectively are likely to have very similar structural data access problems to those that aren’t, but instead of letting this stop them they are breaking the pattern by focussing their AI efforts in areas where their data is already bounded and accessible, and are using AI itself to then improve how data flows through the organisation.

“Amazon's Finance Technology team did exactly this with World Wide Watch,” the Bain repot says, “a generative AI solution that tracks valued-added tax (VAT) regulatory updates across global markets. What previously took tax teams 26 minutes per regulatory update now takes 2 minutes—a 92% reduction—with 80% of the AI generated summaries accepted without modification by human experts. That is not a moonshot. It is a bounded, specific workflow where the data was already there, and AI replaced the manual assembly.”

Instead of waiting for large-scale data modernisation efforts to mature, the approach here is to automate one repeatable, high-value workflow where humans are currently pulling data manually, consolidating spreadsheets, and producing reports, and replace that entire sequence with AI.

The key step is, then to follow this through by reinventing the jobs of the employees involved. In an agent-led operating model, employees are no longer moving work along a process; they are orchestrating, supervising, and making the high judgment calls that agents cannot. That is a fundamentally different role to the one they were previously doing, when they were handling the steps of the process themselves, and it requires deliberate investment in role redesign, new ways of working, and change management that most AI programs currently treat as an afterthought. If done properly, it is these employees who then become the vanguard for further change within your organisation – because far from being made irrelevant, their jobs will have in fact become more interesting, dynamic and strategic than before.

Deloitte’s “State of AI in the Enterprise” report for 2026 [https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf] had similar findings. It quotes the former VP of observability at a major telecommunications company: “We thought we were going to automate jobs,” she said. “The truth is, you’re not. You’re going to give existing workers force multipliers where they can be more effective. Maybe someday these things will start to become headless where they just feed off a dashboard metric and you can pull back staff to wait on an alert that wakes somebody up or flashes red on the screen if something really bad happens. But initially it is going to be more work for those people. They’re not going to be cooling their heels; they’re going to be watching these agents, making sure the volume metrics are right, making sure the qualitative metrics are right, and being there to interact with them if they hit a human-in-the loop gate and need to interact with a human for accountability purposes.”

This kind of change cannot come from the bottom up. It has to be set as explicit policy by the CEO, both because of the impact on roles that will necessarily take place, and because of the need to completely clarify responsibilities. CEOs must explicitly answer the question of who is personally accountable when an AI agent makes a consequential wrong decision in production? In most organisations, governance is split almost evenly between IT, business functions and central teams, with no clear owner in most organisations. As the Bain report states: “When an agent makes a consequential error in a production system, accountability cannot be improvised in the moment. It must be established in advance.”

This is an organisational risk, not a technical one, and it changes the nature of governance. “Governance is no longer a compliance exercise,” the Deloitte report says, “it’s the mechanism that enables rapid, confident scaling. Enterprises where senior leadership actively shapes AI governance achieve significantly greater business value than those delegating the work to technical teams alone. True governance makes oversight everyone’s role, embedding it into performance rubrics so that as AI handles more tasks, humans take on active oversight. This shared responsibility empowers employees to help identify challenges and guide safe, trusted AI use.”

At Securys we echo these insights and have built our data and AI governance and AI-by-design services in accordance with them. We help our clients define where humans should remain in control, how automated decisions and data use are audited, and which records of system behaviour should be retained, and we establish relevant governance frameworks early so that scale does not outpace control. At the same time, we make sure that this work balances risk management with innovation, so that oversight enables experimentation rather than constraining it. The objective is not to add bureaucracy but to create clear, adaptive guardrails that allow responsible progress at speed – turning your data from a risk into an asset that can start to properly power AI.

Keen to turn your data from a risk to a strategic asset?

The smallest data errors can undermine client trust and impact regulatory compliance and business performance. Get in touch to discuss how we can help you improve data quality, strengthen data governance and create the trusted data foundation needed for growth, innovation and AI adoption.