Australian Businesses Struggle to Measure AI Investment Amid Rising Costs

As the adoption of artificial intelligence (AI) surges across Australian businesses, the challenge of measuring the return on investment is becoming increasingly complex. Despite a significant drop in the cost of AI processing tokens, companies are finding themselves spending more than ever on this transformative technology.

Anna Volkova, head of people and culture at HR software provider HiBob, emphasized the difficulties in gauging the financial impact of AI. She noted that while her company has achieved some operational efficiencies, the rapid evolution of AI capabilities makes it challenging to adapt and assess effectiveness. “We’ll probably start to see impact to our bottom line in maybe two or three years,” she stated, echoing sentiments shared by many other businesses in the sector.

This perspective aligns with comments made by Assistant Minister for Technology and the Digital Economy Andrew Charlton, who remarked on the staggering $5-8 billion annual expenditure on AI in Australia. He projected that this figure could escalate to between $20 billion and $40 billion annually within the next decade. “There’s a distant sucking sound,” he quipped, referring to the perceived outflow of value from the country due to AI investments.

Meanwhile, spending on AI has become a prominent line item in company budgets, with the Commonwealth Bank reporting an increase in technology expenditure from $2.3 billion to $2.4 billion in the latest financial year. Although they noted around $200 million in gross benefits from AI use cases, the broader question of whether these investments justify their costs lingers.

On the global stage, concerns are mounting over the billions invested by U.S. tech giants in AI and whether these investments will yield substantial returns, raising fears of an AI-driven stock market bubble. As the AI landscape becomes more intricate, firms are grappling with how to allocate their IT budgets effectively. According to Stu Scotis, AI leader at Deloitte, companies are now faced with balancing traditional IT expenses with new AI costs, which are becoming unpredictable.

At the heart of this dilemma is the concept of 'tokens,' which represent the computational effort required for AI tasks. The price of these tokens has plummeted in recent years, spurred by growing competition, enhanced chip efficiency, and increased data center capacity. Epoch, a research institute, found that the cost for top-tier AI models has drastically decreased, with the price for processing one million tokens falling from $15 in November 2023 to just 17.5 cents as of February last year.

However, as the cost of tokens declines, the demand for more complex computations is driving overall spending higher. Mr. Scotis explained that while human labor is compensated in dollars, digital workers—AI models—are paid with tokens, necessitating a focus on optimizing value from AI outputs. He reassured that the cost of AI does not automatically equate to cheaper solutions compared to human labor, especially for complex tasks.

As companies navigate this evolving landscape, many are adopting varied pricing models for AI usage, including seat-based payments for employee access to platforms like Microsoft’s Copilot. This approach often results in fluctuating costs, as the frequency and complexity of AI tasks can lead to increased token consumption.

Despite the decreasing token costs, some organizations are implementing strategies to manage AI expenses, such as capping the number of tokens employees can use within specific time frames. Jon Whittle, an AI expert and former technical lead at NASA, cautioned against a haphazard approach to AI implementation, suggesting that businesses often lack strategic direction, which complicates cost management. He pointed out that even AI systems struggle to accurately predict the token requirements for tasks, which can vary significantly.

In summary, while the cost of AI processing is decreasing, the overall financial commitment from businesses is escalating as they strive to integrate AI into their operations. The challenge remains to harness this technology effectively to truly assess its impact on the bottom line.

Source: www.abc.net.au