The Inside View

AI Can Write Your Email. It Still Can't Do Your Reconciliation.

Over the past few months, AI has found its way into almost every conversation I've had with Finance Directors, CFOs, and finance leaders. Given the amount of noise surrounding the technology, that's hardly surprising. What has been surprising is how consistent the feedback has become.

Despite headlines suggesting AI is rapidly transforming every corner of business, the people actually using it day to day tend to describe a much narrower range of practical applications. The pattern is remarkably consistent. AI is exceptionally good with language, structure, and communication. When it comes to numerical accuracy, verification, and anything carrying genuine business risk, confidence drops away very quickly.

That gap between what AI is good at and what people want it to be good at is becoming increasingly important for finance leaders to understand.

Where AI Is Already Delivering Value

One area where finance professionals seem to have reached near-universal agreement is communication.

Whether it's drafting emails, adjusting tone, summarising meetings, preparing presentations, or creating first drafts of reports, AI is already saving meaningful amounts of time. Need to soften a difficult message, prepare a more diplomatic version of an email, or produce multiple alternatives before communicating with a stakeholder? The technology performs remarkably well.

Several finance leaders I've spoken with admitted that AI has become a routine part of how they communicate. Not because they can't write themselves, but because it often helps them get to a stronger result more quickly. In some cases, people even described becoming better communicators as a result of using it. Regular exposure to alternative wording, sentence structures, and communication styles inevitably improves your own ability to write and present information.

For these types of tasks, the productivity gains are immediate and relatively low risk. If an AI-generated email needs tweaking, the consequences are minor. If it helps someone communicate more clearly, the value is obvious.

The Work Finance Professionals Actually Want Automated

The irony is that the tasks finance teams most want AI to perform are often the ones they trust it with the least.

Very few finance professionals wake up hoping someone will automate an email. The real prize has always been the more technical work: reconciliations, consolidations, reporting, analysis, calculations, and the countless manual processes that consume time across most finance functions.

Yet this is the area where confidence remains noticeably lower.

The issue isn't that AI can't assist with numerical work. In many cases, it can. The issue is that finance operates to a different standard than many other disciplines. Being mostly right isn't enough. An answer that's 95 per cent correct may be perfectly acceptable when brainstorming ideas or drafting content. It's a very different story when producing management accounts, forecasting financial performance, or preparing information for senior stakeholders.

Finance professionals are trained to verify information, challenge anomalies, and investigate things that don't look right. AI can support that process, but most leaders I speak with aren't yet comfortable handing it over entirely.

Confidence Can Be Misleading

One theme that came up repeatedly during these conversations is that AI's greatest strength can also be its greatest weakness.

It's extremely persuasive.

When AI provides an answer, it typically does so with confidence. The structure is polished. The language sounds authoritative. The conclusion often appears logical and well reasoned.

That creates a risk.

A recent example came from outside the finance profession when a Tasmanian government department reportedly relied on AI as part of a legal process and found itself dealing with inaccurate citations and references. The output looked credible. The underlying information wasn't.

Finance professionals immediately recognise the danger because their world revolves around verification. Numbers don't become accurate because they're presented confidently. They become accurate because they've been tested, reconciled, reviewed, and challenged.

The same principle applies to AI-generated outputs.

The risk isn't necessarily that AI makes mistakes. Humans make mistakes too. The risk is that people stop checking because the answer looks convincing.

Most Organisations Are Finding the Same Balance

What's interesting is that many finance leaders appear to be arriving at a similar conclusion.

The organisations gaining the most value from AI aren't treating it as a replacement for people. They're treating it as an assistant.

It helps create a first draft. It accelerates routine tasks. It improves efficiency. It reduces administration.

But a person remains responsible for applying judgement, validating the output, and making the final decision.

The exact numbers differ depending on the individual and the organisation, but the general sentiment is remarkably consistent. Most leaders are comfortable with AI assisting meaningfully in the process. Far fewer are comfortable with AI owning the entire process unchecked.

There seems to be a growing recognition that the technology is at its most powerful when combined with human expertise, rather than used as a substitute for it.

The Bigger Question Which Is Now Being Asked

The most interesting part of these conversations wasn't actually about AI's capabilities.

It was about productivity.

For years, technology has been sold on the promise that automating work would free people to spend more time on higher-value activities. AI is simply the latest chapter in a much longer story of digitisation and automation.

The question is: if all this time is being saved, where is it going?

Speak to enough leaders and an uncomfortable pattern starts to emerge. Tasks may be faster. Administration may be easier. Reporting may be more efficient. Yet that doesn't automatically translate into higher productivity.

In many cases, the time simply disappears.

The assumption has always been that reducing effort would naturally create more output. In reality, that only happens when organisations deliberately decide what they want people to do with the capacity they've created.

A Leadership Challenge, Not a Technology Challenge

This is where the conversation stops being about AI and starts being about leadership.

Technology can remove effort. It cannot automatically redirect effort.

If a task that previously took two hours now takes thirty minutes, somebody still needs to decide how the remaining time gets invested. Should it go into stakeholder engagement? Deeper analysis? Commercial partnering? Strategic projects? Customer conversations?

Without clarity around that expectation, the benefits of automation often fail to materialise in the way leaders hoped.

The work becomes easier, but the output remains broadly the same.

That's not a failure of technology. It's a failure to adapt expectations alongside the technology.

The organisations that achieve meaningful productivity improvements are usually the ones that redesign work at the same time they implement new tools.

Final Thought

The conversation around AI often swings between two extremes. Either it's going to transform every aspect of work overnight, or it's little more than a passing trend that will fail to deliver on its promises.

The reality, as usual, sits somewhere in the middle.

AI is already creating significant value inside finance teams. It's helping people communicate better, work faster, and remove some of the administrative burden that has traditionally absorbed so much time.

What it isn't doing, at least not yet, is removing the need for judgement, verification, accountability, and experience.

Finance has always been a profession built on accuracy. The tools may evolve, but that principle remains unchanged.

For now, AI can write your email.

It still can't do your reconciliation.