Published in partnership with Novigi

Senior superannuation fund executives must learn to separate the AI reality from the hype that currently surrounds it or run the risk of committing capital to big projects that don’t achieve  results.

Novigi senior partner Alex Moynihan says personal AI tools like ChatGPT and Gemini are “incredibly powerful for content generation, pattern recognition [and] decision support”, and that the pace of development has been relentless.

“You only have to look really at the last 18 months around how fast these tools are evolving and how powerful they’ve become,” he says.

“If you compare ChatGPT 5.1 versus 3.5 released at the end of 2022, it’s just a whole different world in terms of capability and accuracy.”

He expects the productivity tier of AI to continue to improve and deliver significant value to individuals and teams. But real AI-driven transformation for organisations is also emerging elsewhere.

AI is beginning to automate, accelerate and reshape end-to-end workflows, as demonstrated by AI-enabled software delivery, commonly known – and misunderstood – as “vibe coding”.

Moynihan says that the “silent movement” beneath a very noisy surface is seeing AI-coded solutions that combine agentic tooling with enterprise-grade engineering discipline to deliver outcomes rapidly, completely changing how technology solutions are delivered.

 The emerging agentic stack will deliver almost everything, Moynihan says: “provisioning infrastructure, engineering the database, coding the application, and developing an excellent front-end user experience”.

“If I were to make one immediate recommendation to anyone, it would be to sit down with a tool like Replit or Cursor for a day, and your eyes will just open in terms of what this will do to the enterprise more broadly,” Moynihan says.

“These tools are turning software delivery on its head. It is not difficult to see how this capability will transfer across all organisational functions, very quickly.”

What lies beneath

One impact of these agentic development tools is to empower the business to be the driver of innovation. Instead of wading through multi-week sprints, organisations can use AI to generate code overnight, test it during the day, and iterate continuously. The acceleration this produces “is just incredible”.

“We are building enterprise-grade applications using these tools,” Moynihan says.

“While I started my career as a developer, it was many years ago, and this has enabled me to demonstrate value quickly through prototyping before commitment. This technology is still in the early stages, and right now it is the worst it will ever be, and we’re already delivering enterprise grade solutions.

“The way I see it is that these tools are ultimately going to enable the full end-to-end software delivery life cycle to be automated and to be in the hands of those that own strategic outcomes and business outcomes.”

In this guise, AI is more than just productivity boost. Moynihan says it changes who controls the process: end-users also become the owners of a solution because they have collaborated directly in its development.

This can transform not only how change is delivered but also how it is adopted by its  co-creators.

A recent project saw business users testing, shaping and directing features daily, then adding ideas to a Kanban board built into the application itself.

“At the end of a two-week sprint, two important change management outcomes were observed,” Moynihan says.

“First of all, the business owners took ownership of the solution as they co-created it… and secondly, through co-creation, the change management, training and adoption was effectively completed in those two weeks.”

What’s real, what’s not

But senior leaders still face a harder question: how to distinguish real value from inflated, blue-sky promises.

Every meaningful deployment of AI comes with undeniable – often unavoidable – challenges, which can include incomplete or inaccurate data, security considerations, bias, hallucinations or model drift impacting confidence in output, misalignment of processes and systems with regulatory obligations, constraints of legacy IT systems, or inaccuracies that can undermine confidence in an entire project – for example, a home-loan AI agent offering customers an interest rate not matching the bank’s official rate.

“In a heavily regulated industry such as superannuation, that’s a no-go,” Moynihan says, and an organisation’s leaders should expect any serious conversation about AI – be it with external consultants or the organisation’s own in-house teams – to acknowledge and engage directly with these sorts of risks.

Moynihan says AI is not a silver bullet that can fix all of an organisation’s inefficiencies . It must be tested, grounded, validated and integrated, and the surrounding risks must be managed with the same rigour applied to any other enterprise-grade system.

Executives should think carefully about where AI genuinely fits in the range of potential solutions that the organisation has available to it.

The idea of an organisation-wide “AI strategy” is superficially attractive – it might satisfy a board that is alert to the potential of the technology but not well-versed in its application; and it may tempt an organisation to try to apply AI to every problem it encounters, even when there are simpler technologies that could deliver faster, cheaper and more reliable results. But an “AI-first” mindset becomes “a solution looking for a problem”.

Business leaders should be strategy-led and “agnostic of the enablers”,” Moynihan says. They should begin by being clear about the strategic outcome they want to achieve, then break it into measurable objectives, identify the pain points that stand in the way, and only then determine which enablers – rules engines, workflow systems, process redesign, automation, people, data analytics or AI – are most appropriate for each problem.

Where the focus is on driving efficiencies and effectiveness through existing workflows and processes, a root cause issue such as a poorly designed form introducing downstream pain points, for example, doesn’t need the kitchen sink thrown at it: “You don’t need AI to do that for you,” Moynihan says.

A hint that AI may not be the most appropriate technology to use might lie in whether a process is deterministic. If it produces predictable outputs, is rules-based and structured, an enterprise workflow capability may solve it more effectively than AI.

But AI is “really good at that sort of probabilistic kind of generation, interpretation, inference and so on, of unstructured and rich data sets”, Moynihan says, so it is best to understand all of the potential solution enablers and select the right one for each task, rather than treating every problem as a nail and AI as some kind of magical hammer.

Reshaped, end-to-end

Before Moynihan joined Novigi this year, when it acquired the firm he co-founded, TurningPoint Advisory, he had already witnessed how AI had started to reshape end-to-end organisational processes rather than merely accelerate or streamline existing ones.

This is the early sign of the so-called “second wave” of AI adoption, where processes and systems are reimagined and completely redesigned with AI as a foundational technology, not bolted on.

Re-imagining processes is inevitably more disruptive and hence risky to a business than just driving efficiencies and effectiveness through existing ones. Moynihan says that as organisations adopt AI at a deeper level, they will blend AI-generated solutions with enterprise guardrails, traditional integration architectures and established governance.

Moynihan says we’re only at “base camp” when it comes to the transformational potential of AI, but even so, super funds must have confidence in the claims made for it by potential delivery partners.

“If it’s communicated as all upside, be very wary,” he says. Likewise, it should be a red flag if there are promises of a massive return on investment with only a vague outline of how that can be achieved.

Moynihan is a believer in the ‘show, don’t tell’, approach, and if he were a client he would want demonstrated evidence of how the application of AI has worked in similar situations before. “I’d want them to walk me through exactly how they did it, and probably most importantly, talk about the challenges,” he says.

“Being able to describe and demonstrate that experience end-to-end, and prove the value that was delivered, I think is critical. Too often, we can find ourselves in the conceptual conversation about what AI is going to do, instead of the practical outcomes that it can allow us to achieve right now.”