Digital Transformation: Missed opportunity or AI-enabling superpower?
In the era of AI adoption and acceleration, what role should digital transformation play?
For years, digital transformation has been a favourite boardroom topic across organisations large and small. Now AI has kicked the door in. The question is not whether AI belongs in the conversation; it is whether digital transformation still has a starring role or has been demoted to the warm-up act.
This is why AI is now a standing agenda item in boardrooms, including those responsible for major assets and infrastructure. Fair enough too. But while this relationship to broader digital transformation is often fuzzy, that fuzziness matters because it is where organisations miss the chance to turn AI from an expensive novelty into a durable advantage.
Devon Middleditch
Head of Digital Engineering & Technology NSW
Digital Transformation:
At its best, digital transformation is a disciplined program of work: reviewing operational workflows, assessing supporting technology, and lifting the capability of the people expected to use it all.
The ambition is sensible enough: take those lessons, align them to organisational strategy, modernise ways of working, introduce enabling technologies, build cultural readiness, and train the workforce to operate differently. The prize was equally familiar: better efficiency, lower risk, improved profitability, and, crucially, a stronger data foundation underneath it all.
In the engineering design, construction, physical asset, and operations sector, these programs were usually analysed to within an inch of their lives. Proofs of concept were run, initiatives were carefully staged, and implementation teams were assembled to support rollout at scale.
Programs of this kind were typically well planned, engaging a broad church of stakeholders across the organisation.
That level of scrutiny helped pressure-test the mission and, in many cases, produced outcomes that were undeniably robust.
Interestingly, though, these sectors are notoriously slow to digitise. There are many local exemplars, but scaling them across the sector has proven difficult. In NSW, there have been two significant contributions that have helped lift the bar.
Firstly, the highly successful TfNSW Digital Engineering Framework, which became the early benchmark for how to digitise and deliver digitally across infrastructure lifecycle for Government. Secondly and most recently, the Infrastructure Digitisation and Data Policy from INSW, which pivots focus from pure transportation digital and data hygiene to cast the net wider across whole of governments physical asset base.
Case Study in Nudging the Status Quo:
Under pressure to improve performance, efficiency, and perhaps look a bit innovative while doing it, many organisations seemed to lose patience with transformation itself. Rather than closing the gaps or resetting the program, they scaled initiatives back, disbanded teams, and drifted back to business as usual.
Why fix what is not obviously broken, particularly if the invoiceable hours are still flowing? Different motivations were clearly at play.
Meanwhile, the technical cohort had already moved on. While the formal transformation machine was still clearing committee gates, practitioners were experimenting with machine learning and, soon enough, generative AI, turbo-charging the automations they had built over previous decades, albeit with tools that suddenly looked prehistoric by comparison.
These early adopters found something materially different. Generative AI - particularly large language models - could create new content such as text, images, and code, not merely classify or predict from existing data. Unlike traditional models that map inputs to outputs, LLMs infer patterns from vast datasets and generate contextually relevant responses on demand. For organisations that felt less like incremental improvement and more like a new operating lever entirely.
Elsewhere, this newly mainstream technology was gathering momentum for a simpler reason: it was easy to use. Natural-language interfaces gave non-technical staff a shortcut into a world of automation that had previously been gated by specialist skills.
AI and the masses became fast friends. Between 2023 and 2025, even the most unlikely products were suddenly wearing an AI badge like it was a VIP lanyard. Everyone wanted to be at the party - and who could blame them? The promise was irresistible: efficiency, speed, cleaner code, fewer errors, leaner teams. Tonic for the modern operating model.
The Big Shift:
Companies adopted rapidly, invested heavily, and reshaped their operating models in what felt like a long weekend, while scrambling to scale the underlying infrastructure. Often, there was no proof of concept, a limited strategy, and little to no plan for change.
In short: just “AI the hell out of it” and sort the details out later.
Fast forward to today and the appetite, adoption, and investment have been enormous - matched only by the number of underwhelming outcomes.
The early hype, oversell, and inconsistency have started to dent the promise. Trust is wobbling. The euphoria is wearing off. The AI tin still looks shiny, but more buyers are finally reading the label.
To be fair, the speed of adoption has also revealed something useful: organisations can move a lot faster than they usually claim - if they really want to. But this period has also been a masterclass in riding the hype cycle (Source) at an industrial scale.
Why has AI adoption been so frictionless?
So, you must ask: What has made the adoption of AI so frictionless compared to the headwind experience of digital transformation?
Part of the answer is that many people had spent years ‘experiencing’ digital transformation without ‘feeling’ much transformation at all. Then AI arrived in plain English and started helping with recipe ideas, holiday plans, meeting notes, code snippets, and strategic brainstorming.
It was fast, plausible, and strangely democratic. Untested in many cases, yes - but immediate, and that matters.
In broad terms, this is because AI arrived and was implemented as a user experience before organisations had time to develop fit for purpose governance frameworks:
- It was not heavily scoped inhouse before people started using it.
- Staff generally did not resist, and many welcomed a simpler way to get work done.
- Middle managers often saw it as a practical shortcut rather than another transformation burden.
- Budget discipline and formal compliance frequently arrived after adoption had already started.
- Downstream stakeholders leaned in with the same enthusiasm, creating momentum that traditional programs rarely enjoy.
- And because experimentation felt cheap and accessible, organisations briefly behaved as though failure might finally be acceptable - at least until the bills arrived.
Now, because so many people are using AI in the workplace, organisations, particularly government departments, are scrambling to get a governance framework in place. The users are demonstrating a value proposition, and bureaucrats are having to retrofit a governance framework to make it work.
Deviation of Approach:
The deviation in approach is now impossible to ignore. Organisations shortcut the hard work of defining purpose, pressure-testing the use case, and designing the operating model - then act surprised when the AI program fails to deliver what was written on the tin.
There are several reasons AI implementations disappoint: limited analysis of the problem to be solved, unclear objectives, weak success measures, and a tendency to confuse activity with impact. But the elephant in the room is even more basic:
Underlying data quality.
Why the Data Matters:
Organisations need a coherent data framework and disciplined data hygiene. Without both, AI tools may look impressive in a demo yet remain unreliable, difficult to scale, and ultimately hard to trust.
Without an intentional data framework, supportive systems, sound standards, and better ways of working, organisations do not have an information asset - they have a digital attic. Plenty stored away, very little easy to find, trust, or use consistently. AI agents currently struggle in that environment for the same reason humans do.
At that point, many implementations are not just at risk of failure - they are quietly engineered for it.
Effective implementations do the opposite. They remove friction, improve data flow, and make ingestion into AI tools deliberate rather than accidental. That is how platforms convert raw data into useful, consistent, reliable information - and how organisations generate outcomes they can repeat at scale.
Digital Transformation as an AI Superpower?
Which brings us full circle.The real superpower is not AI in isolation; it is the combination of AI with the strongest habits of digital transformation and digital engineering: clear purpose, good implementation discipline, and a comprehensive data framework that supports consistent creation, storage, processing, and use of information.
Pair that with the right leadership shift and a modest amount of cultural honesty, and successful AI can support a leaner, more adaptive operating model. But people still make or break the implementation. Done well, the result is the closest thing to an enterprise superpower: data treated as an asset rather than a by-product.
From that vantage point, organisations can finally start to “AI the hell out of” the right parts of the business - solving actual problems, taking hearts and minds with them, and realising something much closer to the promise on the proverbial AI tin.
The cost of action vs. inaction.
Most organisations want a better outcome but are often stuck on a deceptively simple question: what kind of AI do we need? where do we start? and what will it realistically deliver?
If an organisation is still wrestling with the why or the how, the immediate challenge is clarity - enough clarity to weigh the cost of action against the much quieter, and often larger, cost of inaction. The real discussion therefore, is not about AI, but about productivity.
So, yes, there is investment involved in assessing the current landscape. Yes, subject matter expertise is usually needed to identify the opportunities and navigate the shift. But that cost is modest compared with the expense of carrying on exactly as you are, while expecting somehow to arrive somewhere better.
So, the real question is this: Are you ready to lift the lid on your organisation and use digital transformation principles as the vehicle to supercharge AI in a way that sticks?
Real discussion is not about AI. It is about productivity.