AI confidence starts with trusted data
Organisations are moving quickly from AI ambition to AI use cases. Employees are already using AI tools in their daily work, while leadership teams want practical ways to improve efficiency, customer experience, decision-making, and competitiveness.
“AI confidence starts with the data estate behind the model,” says Nathi Dube, Director: Innovation at PBT Group. “Before an organisation can rely on AI, it needs to understand what data it has, where that data comes from, who owns it, how sensitive it is, and whether it can be trusted for the use case in front of it.”
AI can accelerate work and open new possibilities, while also exposing weaknesses that have often been tolerated inside fragmented data environments. Poor data quality, unclear ownership, inconsistent definitions, weak lineage, and outdated documentation all become more visible when AI depends on that data to produce an output.
The data estate beneath the AI ambition
For many organisations, the first step is to map the data landscape. This includes understanding the systems that generate data, the format in which it is stored, whether it sits on-premises or in the cloud, who has access to it, how fresh it is, and what governance applies to it.
Dube says this foundational view is essential because doubt in the source data carries through to the AI output built on top of it.
“The model may look sophisticated, but the business still needs to know whether the data is complete, secure, accessible, traceable, and fit for purpose. A useful AI output must be explainable, traceable, and trusted enough to support the decision that follows.”
This is where many organisations encounter data chaos. One business unit may define a KPI differently from another. Finance and marketing may calculate revenue differently. Customer data may be fragmented across systems. Documentation may be incomplete or out of date. Lineage may be unclear, leaving teams unsure which source is correct or how a number has been produced.
Those issues may not stop a business from producing reports or running daily operations. However, they do limit the confidence with which AI can be applied, and trusted.
When definitions do not align
A fragmented data estate creates confusion long before AI enters the picture, and AI only raises the stakes. When a business cannot agree on how a customer, revenue measure, risk category, or operational KPI is defined, an AI-enabled process can reproduce that confusion faster and at greater scale.
“Standardisation becomes critical. When the same measure means different things in different parts of the business, AI inherits that fragmentation. You cannot build confidence in AI outputs when the business has not agreed on the meaning of the data being used,” says Dube.
Trusted data starts close to the source. Business owners understand how data is generated, how KPIs are defined, which exceptions exist, and which systems carry the most reliable information. Data teams need that business knowledge to build quality measures, document definitions, and create a common understanding across the organisation.
Without it, modernisation projects can become exercises in reconstruction, with teams forced to piece together meaning from inherited systems, incomplete documentation, and assumptions about how data has been used.
Embedding governance into the flow
Governance is often discussed as a framework, policy, or compliance requirement. In the AI era, it needs to be more practical than that. Governance has to be built into the way data is created, accessed, secured, monitored, and used.
This includes data ownership, stewardship, access controls, quality rules, lineage, security, and approved AI usage. In regulated industries such as financial services and insurance, data sensitivity and accountability cannot be treated casually.
“Governance cannot sit as a document that is applied after the fact. It needs to be embedded into processes and workflows so that trust is built into the way the organisation works with data,” says Dube.
People are central to this. AI is forcing organisations to treat data as an enterprise asset rather than an IT concern. Business users, data owners, technology teams, and leadership all have a role in improving trust. Improved AI adoption will depend on data and AI literacy across the workforce, supported by clear guidance on approved tools, acceptable use, and risk management.
From experimentation to confidence
Dube believes organisations should act, but with discipline. A sensible path starts with lower-risk processes where teams can learn, improve governance, strengthen the data foundation, and build confidence before applying AI to more critical business areas.
This also requires leadership commitment. AI policies, training, workshops, and clear communication help employees understand where AI should be used, where it should be avoided, and how organisational data must be protected.
“AI can become a trusted partner in the business, but only if the foundation is right. That means trusted data, clear ownership, strong governance, and people who understand both the opportunity and the risk,” concludes Dube.
AI confidence is built by creating the conditions that enable responsible AI use. Introducing AI will expose data chaos where it already exists, which makes the foundational work more urgent. When that work is done properly, experimentation can move into everyday use without losing trust, control, or business value.
