AI solves challenges as directors consider risks

The failed promises and emerging perils of artificial intelligence (AI) dominated discussion at the AICD board and Brisbane stakeholders’ dinner in September. Keynote speaker Aldo Gutierrez GAICD, who leads the data and analytics advisory practice at McGrathNicol, outlined implementation barriers to AI and also important considerations for companies considering moving into AI. The Brisbane meeting is the fifth in an ongoing series of board events being developed to focus the AICD board on engaging with key stakeholders. The AICD board meets in capital cities across the country regularly. As a national organisation, it is critical that the board regularly connects with members and stakeholders.

AI has huge potential to transform industries and to solve complex global challenges — but much of the benefit is yet to be realised, Gutierrez told the dinner in Brisbane at Customs House. “Unfortunately, AI hasn’t yet been able to deliver the potential that is being promised to everyone,” he said.

Implementation barriers

Referring to Gartner’s AI in the Enterprise Survey, which polled 632 leaders highly involved in AI, Gutierrez said there were six main barriers to the implementation of AI. Heading the list was estimating and demonstrating the value of AI, because while AI was increasing efficiency, it wasn’t reinventing workflows.

“The value AI could be providing… has been capped,” he said. “We need to find the reason why and do something about it. The benefits seem really good on paper, but we haven’t really seen them materialising in the work that we do every day yet. Over three in four respondents say AI tools have decreased their productivity and added to their workload in at least one way.”

Other barriers to the use of AI also included lack of talent or skills within the organisation to manage it, lack of strategic alignment, and lack of high-quality data.

Lack of trust in AI was also cited as a significant barrier, given past actions by technology giants. For example, Google’s AI-generated overviews have dispensed advice on smoking while pregnant or adding glue to pizza that is misleading at best, and dangerous at worst.

Assessing the hype

Gutierrez also described the cycle for technology transformation, which tracks the different stages any emergent technology follows — even the replacement of steam engines by electrical generators in the early 20th century. Such technologies follow a predictable course through a series of stages — from the “quick wins and hype” to the “successful transformation”.

According to Gutierrez, AI is currently at the peak of the hype cycle, with the inevitable shift to the second stage in the technological transformation stage.

“There will be constraints, there will be costs, and there will be a lot of different issues that we probably never anticipate — it will be a reality check,” he said.

The good news is that at this next inflection point, people realise they need to take charge and reorganise the system, rather than get dragged along by the big promises, or avoiding the technology altogether due to the perceived downsides. At this point, people find better and different ways to do things, developing data strategies, data governance programs and making necessary changes to adopt new technologies more effectively.

“Eventually, it gets to the point where the transformation is a reality,” said Gutierrez. “If we anticipate the cycle, things are going to be probably a little bit easier for everyone in the future.”

First movers

Gutierrez outlined three important considerations for companies considering moving into AI. The first was to establish a data strategy — which 66 per cent of large companies have failed to do so far, according to figures from BigID.

The second was finding good talent. “It is not only the person who can code the most amazing algorithms,” said Gutierrez. “It is the person or the group of people who are going to make the connection between the technology and the business. If you find one of those unicorns, life is going to be a little bit easier.”

Data governance was the third piece of the puzzle – and covered internal concerns around privacy, security and compliance, as well as external forces arising from customers and regulation.

“We’ll reach a big milestone when everything is so automated and the data has so much quality that the insights are received by executives at the same time as the analysts,” said Gutierrez.

Yet more than half of all large Australian companies have no formal data governance process and no reporting to their boards, he pointed out. “We can regain control over AI’s development through a robust governance framework with a special focus on data governance,” he said.

Productivity gains

AICD chair Naomi Edwards FAICD addressed the challenges involved in balancing the opportunities generative AI (GenAI) with sober considerations of its risks. “What is the precipice we walk on?” asked Edwards.

As a director and chair of the risk committee of TAL Australia, Edwards noted that the life insurer had recently shaved a day a week off its entire claims function by partnering with Microsoft’s leading AI technologies like Copilot for Microsoft 365 and training it on the company’s five million customer records.

“It’s incredibly important just to see that very simple emergence of a productivity gain happen so quickly and in such a safe way,” she said.

Edwards noted that in freeing up staff from having to look up “over 800 documents about definitions for a sickness or illness claim” means they can now focus on helping customers face-to-face — for example, by securing them places in rehabilitation programs.

Director resources

Edwards highlighted a range of resources developed by the AICD to support the director community. These included a seven-webinar Governance of AI series.

In June, the AICD also released four resources to assist directors in their governance of AI.

Developed with the Human Technology Institute (HTI) at the University of Technology Sydney, the resources include a Director’s Introduction to AI, which provides foundational knowledge, the Director’s Guide to AI Governance, a concise snapshot of the “Eight elements of safe and responsible AI governance” and a handy checklist for SME and NFP governance.

The AICD is also developing a Director’s Guide to Governance with external experts, and an experiential short course due to launch later this year, which will help directors deep dive on AI with external experts.

Director updates through member events such as the AICD board and Brisbane stakeholders’ dinner are also important ways of engaging with members and their questions and concerns around AI, she said.