Australian SMEs warned of hidden AI costs as agents spread

Pretian Squared has warned that Australian small businesses face a new, hard-to-measure AI cost as usage moves beyond basic chatbot access. It says the shift could widen the gap between smaller firms and larger rivals.

The consultancy's assessment focuses on a shift away from occasional free or low-cost AI use toward paid models based on subscriptions, credits, and per-task charges. In its view, that means AI is becoming a recurring operating expense rather than an informal trial tool staff use.

Survey data cited by the firm shows 42% of Australian SMEs now use AI, with marketing and sales content the most common application at 51%, followed by customer service tools such as chatbots at 24%. Separate figures from the Australian Bureau of Statistics put AI use at about 11% of small and micro businesses, compared with 35% of large businesses.

The gap between those figures points to a mixed picture across the market. In many smaller companies, AI use still appears to mean individual staff members using a chatbot rather than company-wide systems tied to revenue or productivity targets.

US spending data offers a guide to how costs may develop. Research from the Federal Reserve Bank of Atlanta found that more than half of US firms expect to spend USD $200 or less per employee on AI, while the top 10% expect to spend at least USD $2,800. Data from Ramp, covering more than 70,000 US businesses, showed median spending of about USD $12 per employee per month, while the top 1% spent more than USD $7,400.

Pricing models

Chris Petzoldt, Founder and Managing Director at Pretian Squared, said the spending range mattered more than the headline price of AI tools because many businesses still could not connect costs to results.

"Most businesses can tell you what they pay for electricity each month. Almost none can tell you what they pay for AI per desired business outcome. If you can't connect the bill to the value you get, you're not managing a cost - you're absorbing it. This is troubling when businesses face growing pressure to lift productivity amid rising operating costs and increasing competition," Petzoldt said.

He said AI pricing was splitting into three broad models. The first is ad-funded access, which offers free entry but may come with limits on use, or with data and attention traded for access. The second is a subscription model charged per user each month. The third is consumption pricing, where credits, tokens or individual tasks determine the bill.

That final model will likely matter most as businesses move toward AI agents and automated workflows. Under that structure, costs rise with usage whether or not the result produces value for the business.

"For a sole trader or a five-person business, ad-funded AI can be a real relief. It keeps capable tools within reach at a time when every other cost is going up," Petzoldt said.

"But nothing is free. You pay with attention and data, and there's a ceiling. Business and enterprise plans are ad-free for a reason. The moment AI becomes part of how you run the business, you're back to paying," he added.

Risk of bill shock

Petzoldt said subscription pricing remained the most predictable model for SMEs because businesses know their monthly outlay in advance. Consumption charging is less certain because agents can run tasks, call tools and use computing resources without any guarantee that the work produced is worth the cost.

"Seats are simple. You know what you'll pay. Agents are different. They run tasks, call tools and burn compute whether or not the output was worth it. That's where small businesses could get bill shock, the same way they did in the early days with mobile data and cloud," he said.

Evidence on agentic AI use among Australian SMEs remains limited. National Australia Bank has found that 35% of SMEs see automation as an opportunity, but there is little large-scale public data showing how many smaller firms have deployed AI agents in day-to-day operations.

According to Petzoldt, larger businesses are moving faster not because they have access to better AI, but because they are making clearer decisions about where to spend and how to tie that spending to commercial returns.

"Large businesses are pulling ahead, and it isn't because they have better AI. They know which tasks are worth paying for, and they improve and reprice their own services on the back of it. That compounds," he said.

He said small business owners should identify whether AI is expected to improve efficiency or create new revenue, track total monthly spending, read pricing metrics carefully before signing up, and set usage limits before rolling out automated agents more widely.

"AI is becoming a permanent line on the P&L - optimally on the revenue side as well. The businesses that tie it to output while controlling cost will be the ones that pull ahead. Everyone else will just get the bill," he said.