AI without data strategy is just expensive experimentation

• Datacom research shows only 12% of organisations have rolled out AI across their entire business, despite usage rising to 87%.
  
  

• AI initiatives often fail without a solid data strategy, leading to pilot purgatory instead of scalable outcomes.

• Most organisations aren't starting from zero – they're stuck in a "messy middle" of partial modernisation that creates hidden complexity.

Investment in artificial intelligence (AI) is accelerating across Australia and New Zealand, but results are not keeping pace. 

While organisations are continually exploring use cases and running pilots, many are "spinning their wheels" – struggling to move beyond AI experimentation to real, scalable value, says Rich Williams, General Manager of Data and Analytics at Datacom. 

We sat down with Williams to explore why often, the core issue isn't their AI capability, but rather their underlying data readiness.

Research from Datacom's State of AI Index shows despite AI adoption rising to 87% of all organisations, just 12% have rolled out AI across their entire business, with barriers such as skills shortages, data integration issues, and shadow AI continuing to slow progress.

In my experience, what's holding many organisations back isn't a lack of AI capability, it's the lack of solid data foundation and strategy. At the end of the day, AI without a robust data strategy is just expensive experimentation, leading to pilot purgatory instead of scalable outcomes.

It's easy to assume your organisation is making great progress with AI when you see proofs of concept delivering promising results, new tools adopted quickly and teams exploring use cases across the business.

Yet in truth, there can be a growing gap between AI potential and operational reality. When it comes to actually scaling AI, you can soon find that data sources conflict and outputs aren't trusted. Issues around data ownership and governance can also hinder progress.

Look closer and there are often telltale signs that you're treading water. From a technical perspective, warning signs include multiple proof of concept cycles and continued reliance on manual processes. From a business perspective, red flags include no sustained ROI and a lack of measurable improvements in costs, profitability or market share.

At this point, pushing ahead with more proofs of concept is futile. You need to take stock and address the underlying issues.

Most organisations aren't starting their AI journey from absolute zero – they are already partially modernised. A headstart on modernisation sounds like an advantage, but the risk with modernisation isn't falling behind, it's being stuck halfway.

Partial modernisation creates complexity if it means you're dealing with fragmented data environments, multiple platforms, duplicated pipelines, inconsistent definitions of key metrics and unclear ownership of data. As they say, good decisions made on bad data are just bad decisions you don't know about yet.

This "messy middle" is where most AI initiatives stall.

How to know if you’re stuck in the messy middle:

• Critical data still isn’t accessible end-to-end, meaning teams rely on spreadsheets, workarounds or manual intervention to complete AI pilots.

• Data quality issues undermine trust in outputs, with inconsistent, incomplete or poorly defined data making it hard to move from promising experiments to production use.

• AI initiatives aren’t aligned to cyber security, privacy or governance requirements, so pilots stall when they face enterprise risk, compliance or approval processes.

• Business units have conflicting priorities and definitions, creating misalignment on what success looks like and preventing scalable, organisation-wide adoption.

• Technology decisions are driven by platform cost or feature sets rather than business outcomes, resulting in solutions that are technically interesting but hard to justify or operationalise.

• Proofs of concept continue without a clear path to production, with no agreed ownership, operating model, success metrics or plan to embed AI into day-to-day processes.

AI success is typically determined long before models are deployed. In my experience, success depends on whether underlying data is trusted, well-defined and connected across systems – secure and governed but still accessible and available when needed.

Without this solid and resilient data foundation, AI results become inconsistent, difficult to scale and high-risk. This often goes hand-in-hand with an absence of business alignment and buy-in at all levels, which further hampers progress.

Many organisations make the mistake of failing to integrate their data strategy with AI initiatives, treating them as separate entities. AI and data strategy teams often work separately, leading to duplicated efforts, misaligned priorities and ineffective implementation.

Issues as simple as confusion over terminology and requirements can result in misaligned expectations, unclear outcomes and wasted investments which don't meet the needs of the business.

To make matters worse, lack of clarity in defining success metrics for AI projects often results in perceived failures, despite potential value being generated from the underlying data.

In short, AI strategy without a robust data strategy is just experimentation which will never reach maturity.