Mind the AI innovation gap Podcast

A listening tour of Australian boardrooms found that executives were enthusiastic about the potential for artificial intelligence (AI) to unleash an economic bonanza but they wrestled to translate their aspirations into beneficial, tangible outcomes.

Returning from a recent five-state tour, Steve Anderton, Brennan's head of digital solutions, revealed to iTnews a landscape marked by “limited adoption – but a lot of interest”.

Hampered by ill-defined goals, vague or non-existent success metrics and exuberant overconfidence, a widening chasm had opened between AI champions and sponsors – most notably, CFOs that held the purse strings. It was exacerbated by a feeling that board directors, owners and investors were rushing AI into production before essential groundwork was laid, Anderton found.

“Hype was driving a sense of urgency to look into these technology sets [but] there was a lack of readiness … and the business case wasn't stacking up,” Anderton said. “There was a disparity between the alignment of the technology and the outcome [which made] releasing funding really difficult.”

This gap between aspiration and execution meant just five percent of AI pilots graduated to production, Anderton said, citing Brennan's research partner, Adapt. 

Anderton attributed the low conversion rate to AI technologies not being in a fit state, dysfunctional AI working groups, and failure to shore up platforms and infrastructure to support scale. Pilots also often failed to justify investment or were conducted in isolated environments, unprepared for production realities.

Notably, 60 percent of CFOs surveyed said they lacked confidence their own businesses were writing effective AI use cases that warranted investment.

To resolve these tensions, Anderton advocated ‘micro-innovation’ to bridge this perception gap and kickstart innovation. He said it started with cross-functional teams of decision-makers and technologists rapidly prototyping, iterating and proving value with agreed success metrics to sidestep overinvestment in risky initiatives that delivered limited or ill-defined benefits.

“So when you get to the end of that prototyping phase in a matter of weeks, not months, you're able to say the business case will stack up … but that takes a cultural shift in an organisation,” he said.

Anderton advised AI champions not to “eat the elephant all at once”. Instead, pick a high-visibility use case with low investment and risk that added significant impact.

Key to scaling AI into the real world? Firm up your infrastructure

To progress an AI pilot or prototype into the real world required underpinning IT and governance foundations, Anderton said.

Enterprises must have strong data platforms, comprehensive data governance, and Modern cloud-based platforms to rapidly innovate and scale.

The proliferation of “Shadow AI” where users adopt readily available tools such as ChatGPT or Microsoft Copilot without oversight, posed added risks such as data leaks.

“These tools are so readily available that users are just signing up and it's starting to proliferate into business”, leading risk-averse management to shutter AI projects until they understood the risks, Anderton said.

Conversely, a robust and flexible Modern infrastructure with strong governance wrapped around a data platform, processes and policy sped the adoption of AI in the enterprise. This could include deploying a governance, risk and compliance platform such as Microsoft Purview to define safety ‘guardrails’, he said.

“It is the ability to swiftly bring in new data sources, to ingest them into your data platform, to model that data and present back to stakeholders [and] users that drives value back to the business.

“You don't want to be waiting around to provision infrastructure or services to support these … cloud-native solutions; tooling is absolutely essential to underpin that innovation.”

People still the missing link to forge an AI future

Addressing cultural challenges, such as workers’ rational fear of losing their jobs, was also critical, Anderton said, but not necessarily for reasons often cited elsewhere.

“It's not that AI will take your job but, potentially, someone who can use AI is likely to take your job”, he said underscoring the need for effective change management and training with any AI rollout.

A lesson Anderton learned from his nationwide listening tour was the urgent need for Australian enterprises to adopt a “bimodal approach”, building innovative AI on a firm technology and governance base.

“Don't get too bogged down in the technology,” Anderton cautioned.

“Secure your foundations but put a framework in place that attaches to the outcomes and value” you seek to attain. Then, investment should naturally follow the measurable and expected benefits, speeding AI adoption from aspiration to reality, he said.

Listen to the full Podcast below or visit the iTnews page of your favourite podcast platform.

To learn more about how to kick your enterprise AI innovation program into high gear, visit Brennan’s website at brennanit.com.au.