Why Good Data Comes Before Automation and AI

A practical perspective on data, governance, and finance transformation from the ground up.

AI, dashboards, and automation dominate conversations in finance transformation. Yet many businesses still struggle with data quality, inconsistent reporting, manual workarounds, and a lack of trust in their numbers.

Recently, I sat down with Becky Snarey, an FCA-qualified senior finance professional with deep experience in data-led transformation, to discuss where businesses should really start.

Her perspective was refreshingly pragmatic.

Rather than rushing to outputs, she argues that strong data foundations, ownership, and governance are what ultimately enable effective automation, analytics and AI.

Why data foundations matter more than ever

Data has always existed in business. Historically, it was used to support operations and historic reporting and was closely controlled in terms of access. Now, data is strategic—it moves faster, is shared in real-time, is used across multiple systems, and is used to predict future outcomes and enable decision-making, so it’s vital that we get it right. And that’s before we even start to consider the increases in types of data, volumes and regulations!

Modern tools allow organisations to combine, analyse and visualise data at scale. The risk, however, is that poor-quality or poorly understood data is now presented through tools that look highly credible, creating false confidence in decision-making.

Before investing in dashboards, AI or automation, Becky stresses the importance of understanding:

• Where data comes from
• How it flows between systems
• Who owns it
• What controls exist around it
• The quality of the data

Without that understanding, businesses risk scaling problems rather than solving them.

Data governance is not just about quality

Good data governance is often misunderstood as simply improving accuracy. In reality, it goes much further.

Key elements include:

• Clear data ownership, so issues can be resolved quickly
• Consistent definitions, avoiding multiple versions of the truth
• Understanding data lineage, particularly as systems become more complex

A common example is something as simple as headcount. There may be valid reasons why a headcount definition for Finance purposes might differ from an HR or Corporate Property Services team definition, but if the senior leadership team are presented with different datasets without adequate explanation, trust will very quickly be eroded in the numbers.

Importantly, governance does not always need to be complex and require bespoke applications to manage. A high-level log (taxonomy) of data types and owners and simple process maps showing systems, integrations and data flows can be a good starting point and be scaled over time.

Understanding the data landscape before spending money

One of the most practical points from the discussion was this: you cannot prioritise investment in automation efficiently if you do not understand your data landscape.

Having a high-level view of systems, data sources, and processes allows organisations to:

• Identify duplication of datasets
• Spot high-risk areas
• Focus spend where it delivers the most value

This understanding is also essential for managing change, helping teams assess the impact of systems, data or process changes before they happen and communicate effectively with impacted stakeholders.

Data architecture and the role of data lakes

Modern data architecture, often using cloud-based data lakes, allows organisations to consolidate data from multiple systems into a central structure.

Compared to traditional data warehouses, data lakes:

• Reduce duplication
• Enable reuse of data models
• Simplify connections to reporting and analytics tools
• Enable AI

However, Becky highlighted that success depends on careful decisions around what data to bring in and at what level of detail, as well as close collaboration across finance, business and technology teams. A clear view of who will consume the data and for what purposes informs the right data model and access design.

Process automation should be pragmatic

Not all transformation needs to be large-scale or rely on change management resources.

Automation can range from:

• A single dashboard improving visibility and access to data
• Through to multi-year finance system transformation programmes

The key is prioritisation. Becky advocates a risk versus reward approach, focusing first on areas with regulatory risk exposure or significant manual effort.

In reality, many opportunities start with a review of EUCs (End User Computing) – that’s your spreadsheets, low-code tools, and applications that allow end users (non-IT staff) to create, manage and interact with data themselves.

The goal is not to eliminate EUCs but to enable agility while managing risk, quality, and compliance, and this is where good data and records management can help.

Data owners and processors within the business should have basic data literacy and be able to highlight data issues, risks, and, importantly, opportunities for improvement. These subject matter experts are key, partnered with the right technical expertise, to evaluate ideas, consider benefits, and drive positive change and continuous improvement.

Where AI really fits

AI should not be the starting point.

Instead, it becomes valuable once:

• Data quality is strong
• Controls are understood
• Leadership is aligned

That’s not to say that we shouldn’t all be experimenting with AI. Used well, AI acts as an enabler and personal assistant and is embedded in tools many businesses already use.

Guardrails are essential, as is a “human in the loop” to validate outputs and to ensure that data is not shared outside the organisation without the appropriate permissions.

What this means for SMEs

For SMEs in particular, the message is reassuring.

This is not about big budgets or large teams. It is about:

• Starting small
• Understanding data hand-offs between teams
• Addressing known issues first

Often, the biggest step is having someone in the business who can bridge finance, technology, and operations, helping translate data into meaningful insight.

Final thoughts

Whilst AI presents many opportunities for efficiency and automation, it’s a well-quoted saying that if you put poor-quality data in, you’re going to get poor-quality outputs.

For businesses considering automation, analytics, or AI, the most impactful first step is often the least glamorous: getting the data foundations right.

If this resonates and you are thinking about data, automation or AI within your finance function, I am always happy to have a conversation about what “good” looks like in practice and introduce people who can help.