In August 2020, Citigroup made a $900M mistake.
Their team sent what was meant to be a routine interest payment to a group of syndicated lenders. But because of a procedural misstep in their automated system, they paid the entire principal instead.
Some lenders returned the funds immediately, but many didn’t. Citi spent the next two years fighting legal battles for the rest of the money, and took a heavy reputational hit as their negligence made headlines.
Their story underscores a growing issue: AI automation in accounting, while efficient, isn’t perfect. And if it fails, the costs can be astronomical.
What Caused Citi’s Mistake?
The heart of the issue was poorly designed software paired with inadequate financial controls.
Their system required users to navigate multiple autopopulated fields to distinguish between interest payments and principal. On that day, an employee used the system improperly, resulting in a $894M transfer.
While technically human error, the mistake was exacerbated by automation. It was executed after software placed data into a confusing user interface, and sent in full without triggering oversights or warning systems. Plus, a payment that large should definitely have triggered a review.
How Does This Relate to the Rise of AI?
Citi made a mistake with traditional software. However, generative AI is accelerating what can and cannot be automated, and the past holds lessons for the future.
Their story highlights the paradox of automation: the more we trust it to handle complex operations, the greater the opportunity for something to go wrong.
What Accounting Tasks Can Be Safely Automated?
Don’t get us wrong. There’s nothing inherently wrong with automating some accounting work. Indinero is excited about the future of AI, and we help clients automate bookkeeping all the time. But there’s a limit to what you can and should automate.
For starters, we need to distinguish between bookkeeping vs accounting.
- Bookkeepers: Most business owners have acted as their own bookkeeper at some point, so you’re likely familiar with their work: recording expenses, reconciling bank statements, preparing invoices, and other day-to-day tasks.
- Accountants: These professionals build on data bookkeepers collect. They build financial reports for investors and internal stakeholders, make strategic recommendations, and provide insights for long-term financial decision-making.
Simple bookkeeping tasks are ideal for automation, provided that proper oversight is in place. Citigroup integrated some automation into its AP team, an offshoot of bookkeeping, but got in trouble because of weak financial controls.
But the strategic insights and experience of an accountant are too risky to replace with AI.
What Are the Risks of Automated Accounting?
Just ask yourself: would you be confident sharing an AI-generated financial statement with an investor? Or an auditor?
Certain tasks are well-suited to automation, but it’s essential to understand the boundaries. Below are some of the most common and costly pitfalls companies face when automating their accounting processes.
- Legal liability: Even if using an AI tool leads to a mistake, your business is still on the hook legally and financially.
- Hallucinations: ChatGPT is notorious for confident but incorrect statements. This is particularly dangerous in tax and financial reporting.
- Data Leaks: Your data may be used to train future models, or the AI you share with could have a cybersecurity breach.
- Black-Box Output: AI tools can certainly generate reports, but not understanding how they arrived at their conclusions could create problems during audits or due diligence.
- Overreliance: As junior staff become reliant on AI, they may not develop foundational accounting skills, leading to a future expertise gap. Citi’s employee may have been too reliant on their software, too. Nearly $900M in payments should have seemed unusual.
Undetected errors can lead to costly fixes.
Lastly, we want to emphasize that undetected errors can lead to costly fixes. People don’t worry too much about their accounting, until it’s a problem. Anecdotally, we at indinero have heard about companies paying upwards of $50,000 to fix reporting mistakes, and losing $25M in enterprise book value as they address problems a controller or CFO would have caught.
What Can Accountants Do That AI Can’t?
In short, we’re your trusted strategic advisors.
Automation can handle repetitive tasks and assist with report first drafts, but it lacks something crucial: professional judgment. Accountants don’t just process numbers; they interpret them in the context of your business goals, risks, and long-term strategy.
Seasoned accountants prepare investor-ready financial statements that do more than meet compliance requirements. They tell a compelling story. Entrepreneurs preparing for an IPO, an exit, or funding round can greatly benefit from an accountant who knows how to present their financials in the best light.
Beyond reporting, we can help manage working capital, structure debt effectively, and optimize AR and AP to maintain healthy cash flow. We play a critical role in navigating regulatory complexity—an area AI should absolutely not be trusted to help with—as well as navigate tax codes in a way that most benefits your bottom line.
Perhaps most importantly, human accountants bring clarity. We translate complex financial concepts into actionable insights. And when mistakes happen, as they sometimes do, there’s someone accountable for fixing the problem.
Automation Is Powerful, but So Are People
The Citibank incident is a reminder that even huge firms with sophisticated tools can make big mistakes if automation is left unchecked. For small and medium-sized businesses, the lesson is clear: AI isn’t infallible, and human input continues to be relevant.
At indinero, we help growing businesses leverage the best parts of automation without compromising on accuracy, compliance, or strategic insights. Our expert accounting team is ready to handle the numbers so you can focus on growth. Contact us today for a free consultation.
