In every industry and profession, business leaders are making decisions about their AI strategy. And in almost every case, the strategy is “more.” According to Gartner, global AI spending is expected to increase by 46% in 2026. But more AI isn’t always the right answer. In the financial profession, AI without guardrails can lead to disaster.
Earlier this year, a CFO tested several leading LLMs on their ability to complete accounting tasks. While Claude Opus 4.7 led the field, the model was only able to successfully complete 80% of the tasks; every other model fared worse.
Would you turn your books over to an accountant who makes a mistake on 20% of their decisions?
There is a place for AI in finance and accounting. When deployed carefully and correctly, AI can help accelerate workflows and increase the value humans are able to deliver to their customers and coworkers. The strategy? Point, don’t fix: use AI to analyze large volumes of data and identify problem areas for humans to address.
The need for human intervention
When an accountant or financial professional takes on a new client or project, their first task is to make sense of the challenge. That can involve sorting through massive amounts of data, aiming to uncover potential problems or identify strategies that could lead to savings.
For human financial professionals, that first moment can be tedious and time-consuming. The problem is sorting and discovery: recognizing which areas require attention and, more importantly, which ones don’t. This is an ideal use case for AI, which can analyze reams of data in seconds and uncover signals that would otherwise take hours for a human to recognize.
In my work, I’ve seen AI deliver value in this way in minutes. In one case, we ran a new client’s books through our AI system and within minutes flagged a $785,000 discrepancy in reported net income — a gap that had gone undetected across two separate accounting systems. What looked like clean books on the surface turned out to require a full account-by-account rebuild before they were reliable for tax or compliance purposes. AI didn’t replace the expert review — it told us exactly where to look.
This is where human intervention becomes critical. A $785,000 discrepancy requires real expertise to ensure it has been resolved properly. An AI bookkeeper might be able to provide a solution; is it worth the 20% chance that the solution is wrong? You can rely on AI to sort through the mass of information, find the problem, and point to it. You need humans to come up with the fix and implement it correctly.
Where AI adds value
AI can handle some bookkeeping tasks but not others. Where do you draw the line?
Accounting and finance professionals should think about these three questions when making decisions about their AI strategy.
1. Where can you uplevel human activity? Clients pay their accountants and financial advisors for their expertise and guidance. Business depends on knowledge and relationships. AI can deliver the most value when it allows human experts to spend more time on these high-value tasks. Look for opportunities to automate the repetitive, time-consuming tasks that pull your people away from high-value work. Odds are, there’s an AI solution that will save both time and mental toil.
Upleveling human activity can also reinforce the quality of the AI itself. When human experts engage with AI, that expertise feeds into the system and refines the outputs going forward. This is more than pointing at problems. It’s steering toward better outcomes. When human experts define the context, set the parameters, and refine the outputs, that knowledge becomes part of how the AI performs. The expert layer shapes what AI looks for and, over time, how well it finds it
This strategy should also apply to how you review and adjust AI activities. Some tasks can be reviewed by junior employees, while others require CPA-level oversight. The difference should be based on factors like risk, dollar value, and confidence scores. This structured approach provides an actionable framework beyond simply hoping that human employees catch AI errors.
2. Are you competing on price or compromising on precision? For many businesses, AI is attractive because it allows them to cut costs. When AI automation saves hours and hours of human work, it can allow organizations to reduce headcount and save on salary. That in turn allows them to lower costs for their customers and become more competitive in the market.
However, those cuts can come at a cost. Reducing headcount, particularly the expert humans that ensure the quality of your work, threatens the quality of your work. Every decision you make about AI implementation should be viewed on a spectrum: are the savings worth a potential tradeoff in performance?
3. How do you establish trust? If your AI tool makes a mistake in your accounting, that’s the last mistake it will ever make. Lost trust in the financial industry is impossible to overcome. This is why the question of where to draw the line is so crucial, and why so many companies are taking on dangerous risks by automating too many of their tasks with AI. Every time you turn over a task to an AI tool, ask yourself how likely it is that the tool makes a mistake and what the downstream impact would be if it did. If the answers make you uncomfortable, that’s a task that requires a human.
The same principle applies to data security, particularly in an industry with strict compliance requirements. If an AI tool mishandles your customer data, you could be facing consequences ranging from lost business to regulatory and legal action. “Move fast and break things” doesn’t work for financial professionals.
At the same time, demonstrated human expertise can build trust with skeptical customers. AI tools that are built by human experts, evaluated by human experts, and operated by human experts are worthy of a customer’s trust.
We’re living through one of the most exciting periods of innovation for the financial and accounting sector, and we shouldn’t let fear stand in the way of progress. Financial professionals can see significant value by deploying AI, but only when combining the tools with real human expertise.
Experts in the loop. Compounding improvements. Point, don’t fix. These are the principles that will underpin a successful, responsible AI strategy.
====
Todd Fox is the Chief Technology Officer at Arvo Tech. Todd brings over two decades of engineering and product leadership experience from scaling technology companies.
Sign in to get access to this free resource, and all of our whitepapers and reports.
Download this content today!
Register Now Already registered? Click here to Log In
