From pitch-dark data to decision advantage: The opportunities and risks of a unified data and AI landscape

Most organisations have no shortage of data. The challenge is that much of it can’t be used with certainty. It sits across systems, spreadsheets and platforms, but it is inconsistent, poorly understood, or difficult to rely on at the point decisions are made.

This is where ‘pitch-dark’ data shows up. It’s exists, but isn’t visible in any meaningful way because the context around it is missing, including ownership, definitions, quality, sensitivity and permissions. When data stays pitch-dark, teams fall back on workarounds, reconciliation and judgement calls. This slows decisions and increases risk for the organisation.

When those barriers are removed, something shifts. Data that was previously underutilised can start to inform strategic decisions, support planning and give AI the context it needs to produce more reliable outputs. This is the opportunity of a unified data and AI landscape: bringing pitch-dark data into the light in a controlled way so it can be used safely as decision context, not just stored away.

However, it also introduces material risk. Without deliberate design around security and governance, greater visibility can increase exposure faster than it creates value. 

Ultimately, this isn’t just a technology change. It is a leadership decision about what data and AI should enable, where the boundaries sit, and how outcomes will be measured.

The opportunity: making pitch-dark data usable 

Across most organisations, significant volumes of pitch-dark data sit unused. It might be buried in operational systems, legacy platforms, file shares, inboxes, project tools, or data stores that were never set up for analysis. It often contains value, but it isn’t decision ready.

The opportunity lies in making this data usable by bringing together fragmented structured data (e.g. SQL databases, ERPs and HRIS) and unstructured sources (e.g. emails, documents, logs, audio) into a governed, unified platform.

This allows the data to be discovered, integrated and enriched with metadata so it can be understood and used with confidence.

A simple way to think about the progression from pitch-dark to decision-ready is: 

• Bronze: Ingest raw structured and unstructured data into a unified platform, making it discoverable through metadata and cataloguing.
• Silver: Apply AI to extract meaning, standardise information and align definitions across domains. 
• Gold: Curate governed, reusable data products that support decisions and scale across AI use cases.

But structure alone does not deliver value. What matters is how the data is managed.

Clear ownership, agreed definitions and governance embedded into day-to-day use are what turn pitch-dark data into something that teams can rely on. Without that, it simply becomes more visible, not more useful.

When done well, the impact is practical. Leaders spend less time debating whose numbers are ‘right’. Teams reduce rework and reconciliation. Decisions move faster because people have confidence in what they are seeing. Analytics and AI become more dependable because they are working from shared, understood context.

The risk: more visibility, less control

In many organisations, identity and access management rules have become increasingly complex over time. They’re applied differently across systems, recreated when data is reused, and often not revisited until there is an issue. That approach doesn’t scale in a more connected environment. When more people can see and use data, without consistent control, a few things tend to happen:

• Sensitive information is shared more broadly than intended
• Controls are applied inconsistently depending on how data is accessed
• Confidence drops, particularly when it comes to using AI beyond experimentation.

The issue is not bringing data together. It’s doing so without thinking through how it will be governed. Organisations need to be clear on what data can be shared, what needs oversight, what should remain restricted and what is not suitable for use with generative AI tools at all.

This is especially important with pitch-dark data. Some of it is ‘dark’ because no one has looked closely at what it contains, who can access it, or whether it includes sensitive or regulated information. Before you increase visibility, you need guardrails that hold regardless of where the data sits or how it is accessed. Managing these risks early is far easier than trying to contain them once access has expanded. 

In practice, this means moving away from ad-hoc controls towards something more consistent and built in. Governance needs to work continuously: data is understood, sensitivity is identified, access is defined, and usage is traceable as it is used, not after the fact.

A practical lens: start broad, then go deep

One of the most effective ways to manage both value and risk is to keep things simple at the start. Start with a broad, shared view of performance. Then apply a specialist lens to deliver depth and action. 

A finance example makes this clear. When everyone is working off the same version of financial performance, leadership teams can align quickly. They’re not debating which number is right. 

From there, specialist analysis adds value. You can look at labour costs, for example, and break them down into drivers like overtime, demand patterns or rostering inefficiencies. Without the shared starting point, these insights tend to be challenged. With it, they are acted on. 

This is also a sensible way to introduce pitch-dark data. Start with the shared, high-confidence data that builds alignment. Then bring in ‘darker’ datasets where they add real decision value, with the right controls in place.

The same principle applies across other areas. A clear, consistent foundation builds confidence, but more targeted analysis is where the greatest value comes through.

How to get there: focus on what matters first

The organisations that get the most value from data and AI don’t start with infrastructure. They start with purpose.

Start with the business outcome 

Instead of asking where the data lives, start with why it needs to be connected:

• Which business decisions, products or AI use cases must this data enable? 
• Where are decisions slow, manual or inconsistent because data is fragmented? 
• Where is confidence or trust low due to multiple versions of the truth? 

From there, the value becomes clearer, delivering the outcomes leadership teams care about: improving profitability, reducing effort and cost, strengthening compliance and risk management, and improving customer and employee experience. 

A useful test is this: if you turned the lights on tomorrow, what decision would improve first? If there isn’t a clear answer, you’re likely connecting data for activity rather than value.

Manage risk before visibility increases 

Before increasing access, it is important for organisations to define where the information boundaries are: 

• What data is safe to share broadly? 
• What can be shared with controls and monitoring? 
• What must remain restricted? 
• What should not be used with generative AI at all? 

This is where pitch-dark data requires extra care. If data has not been classified, understood, or assigned an owner, it should not be treated as ‘ready’ simply because it is now discoverable. Define the boundaries first, then increase visibility.

Start small but make it count 

There is no need to solve everything at once.

A better approach is to focus on one dataset that matters to the business, define it properly, and use it to support a small number of real decisions. This creates a reference point for what good looks like and demonstrates value early.

Choose a dataset where the ‘darkness’ is currently creating friction or risk such as slow decisions, duplicated effort, inconsistent reporting, or uncertainty about sensitive information. Then turn it into a gold standard example others can follow.

Connect and prepare deliberately 

From there, it is about expanding carefully.

Each new dataset should be easy to find, clearly owned and governed from day one. Access shouldn’t be assumed; it should be defined. And wherever possible, data should be prepared once and reused, rather than copied and recreated. This keeps things consistent and avoids unnecessary complexity.

Use data in context, not in isolation 

When data is consistent, and well understood, it becomes far more useful for both analytics and AI. 

AI performs better when it has clear boundaries and context. It’s more reliable, easier to explain, and less likely to produce outputs that need to be second-guessed. 

Success is not measured by how much data is connected. It shows up in practical ways instead: decisions are made more quickly, manual effort is reduced, and confidence in outcomes improves.

Make trust part of how data is used 

A unified approach to data and AI only works if people trust it.

That trust comes from knowing that sensitive data is identified, access is managed consistently, and usage is visible.

Over time, this becomes less about policy and more about how the organisation operates day to day. Data is understood, responsibilities are clear, and people are confident in how it can be used.

When that happens, organisations can move faster without increasing exposure. 

How BDO can help 

BDO’s data, analytics and AI team works with organisations to take a practical, outcome-focused approach to data and AI, focusing on outcomes while building the right level of governance.

We work with boards, executives and technology leaders to clarify where value sits, establish ways of working that support consistent use of data, and build trusted foundations for AI that can scale over time.

The focus is not just on capability, but on making sure it delivers something tangible. 

Contact our team to discuss how your organisation can better use data to support decision-making and manage risk.