BUSINESS | ARTIFICIAL INTELLIGENCE

You Hired AI to Make You Efficient, but it Profits When You're Not

ALI GREEN | JUNE 13, 2026 | 4 MIN READ

When you ask AI to put together your next speech, presentation or strategy paper, it doesn't just deliver a first draft.

Unprompted, it may suggest a shorter version, a stronger version or a different angle. It may offer to challenge your assumptions, expand a section, generate alternatives you can compare, or ask a follow-up question you hadn't considered.

Every back and forth feels productive, like you’re getting closer to a better outcome. And, in many cases, the final result is often far better.

But every interaction also keeps you in the conversation, and I think this is one of our biggest blind spots.

One of the promises of AI is that it reduces the cost of work. In many cases, it absolutely does. Blank pages become first drafts in seconds, reports that once took hours can now be completed in minutes, and tasks that once required specialist expertise are suddenly within reach of much larger teams. But the companies providing these tools are not passive players. We've adopted AI to help us do more with less, yet the businesses behind these platforms also need to make money and ultimately the more we use their AI, the more money they make.

That doesn’t make AI bad. But it is a commercial reality worth understanding before we start making major decisions about productivity, hiring, workflows and future growth based on projected AI savings.

Most leaders have heard the term "tokens". Far fewer understand what they are or why they matter.

For those still trying to wrap their head around them, tokens are simply how AI companies measure usage. Every time you interact with the platform, whether that's asking a question, uploading a document, generating a report, iterating an idea or refining an output , you're consuming tokens. Think of them like electricity. The more you use, the more you pay.

For smaller businesses using fixed monthly subscriptions, this cost may show up instead as usage caps and the need to wait before using the tool again unless upgrading to a premium tier. Either way, the underlying point is the same: usage has an economic value, and the platforms have every reason to encourage more of it.

That is what makes AI different from a simple software subscription. The thing creating value, the conversation itself, is often the same thing creating cost. The more interaction, the more consumption. The more consumption, the more revenue for the AI company.

And once you understand that, a lot of the current AI conversation starts to look different.

Recently, Axios reported that an unnamed company may have spent as much as US$500 million on Anthropic's Claude in a single month after failing to implement limits around employee usage.

To most of us, a half-billion-dollar AI bill in a single month is impossible to comprehend. In reality, it may have been worth every cent. It may have generated many times that amount in revenue, future revenue or organisational value. Equally, introducing employee usage limits might reduce the bill, but also reduce capability.

Whether it was a good investment isn’t the point. What struck me most is that the large figure only represents what Anthropic charged, not the true cost of AI to that unnamed business, which would have been considerably higher.

What an invoice like that doesn't capture is everything happening around the technology. It doesn't capture the time staff spend learning new tools, attending training sessions, building prompts, testing outputs and comparing models. It doesn't capture the work required to develop and deploy the governance frameworks required once AI becomes part of how the organisation operates. Nor does it capture the hours diverted from other priorities to iterating, rewriting and checking whether AI-generated work is accurate, useful and of a high enough quality to send internally or externally. And unlike many technology projects, these costs don't disappear once implementation is complete: new models emerge, existing platforms change, governance frameworks evolve, teams need retraining. The learning curve doesn't end, it simply moves.

Of course, the real value of AI does not come from asking it to draft the occasional email. It comes from embedding it into workflows, systems and processes so that it actually changes how work gets done. That requires time, training, oversight and often significant organisational change.

At this point, AI is no longer simply a tool sitting as a software expense in the budget, it has become part of the organisation’s infrastructure.

And infrastructure is expensive to replace.

Every week I see another company confidently forecasting or celebrating AI-driven savings. And they might be right. When used correctly AI provides enormous opportunity for Australian businesses. But every time I read one of these announcements, I find myself wondering “How have they calculated these savings?”

The tools we are bringing into our organisations to reduce costs belong to companies whose commercial success depends on increasing usage. That doesn't make the tools less valuable. But it does mean leaders need to understand the economics before claiming the savings.Pricing is unlikely to hold, either. AI companies are still fighting aggressively for market share, and the cost of running these models remains enormous. It would be optimistic to assume today's economics will remain unchanged forever.

So before you sign off on the next AI business case, restructure a team or forecast future savings, ask yourself:

Are we measuring the time spent learning, implementing, governing and managing these tools, or only the usage and subscription fees?

If our usage doubled next year, would we know why?

If our preferred platform increased its pricing, how exposed would we be?

And, if a better platform emerged tomorrow, what would it actually cost us to leave the ecosystem we've built around today's provider?

Before we claim the savings, we need to understand the costs.

_____________________

Enjoyed this article?

Share it with someone who came to mind while you were reading it. Most readers discover Inside Brief because someone they trust sent them an article.

_____________________