AI in Accounting: Knowing When to Automate and When to Collaborate

Artificial intelligence (AI) has become one of the most talked about innovations in business today. For accountants, auditors, and finance professionals, the discussion usually centers on a single question: Will AI replace us?

The truth is more nuanced. To answer this properly, we can draw on recent work by Erik Brynjolfsson (Stanford Institute for Human-Centered AI), Andrew McAfee (MIT Sloan School of Management), and Michael Jordan (UC Berkeley), who wrote extensively on the tension between automation and collaboration in AI. Their insights, first published in The Atlantic (2024), are deeply relevant to our profession.

Automation vs. Collaboration

Brynjolfsson, McAfee, and Jordan distinguish between two types of tools:

• Automation tools are designed to take over tasks completely. In accounting, these include automated bank feeds, payroll submissions to SARS, or invoice Optical Character Recognition (OCR) extraction. They save time and cut costs, but if they fail, human expertise must step in as the safety net.
• Collaboration tools enhance human capability but still require professional judgment. Financial dashboards, forecasting models, and audit analytics fall into this category. They don’t replace accountants, they amplify our insight.

The authors warn that poorly designed automation doesn’t just fall short; it can actively undermine expertise by encouraging over reliance.

Lessons from Aviation

The article highlights the 2009 Air France 447 disaster, where over-reliance on autopilot (automation) left pilots unprepared when the system disengaged. Human expertise had eroded from disuse.

This resonates with accounting: if professionals outsource too much to “black box” automation, we risk losing the sharpness of professional skepticism, the very skill that protects clients and businesses.

A Story From the Workshop Floor

We’ve seen similar dynamics outside of accounting too. Take modern car repairs.

When you take your car to a dealership today, technicians often rely on a diagnostic machine that plugs into the vehicle’s computer box. But if the computer box itself is faulty, the diagnostic tool can’t provide a clear answer.

I once had a a vehicle with ECU issues. At the dealer, I was told they “suspected” the ECU was the problem, but they couldn’t be sure. They wanted me to authorize a replacement, at the time, a R20,000 part, just to find out whether that was the real issue.

By contrast, seasoned mechanics who trained before diagnostic machines became standard could listen to the engine revs, feel the idle, and tell you almost immediately what was wrong. Their expertise filled the gaps that automation could not.

That experience confirmed something we also see in finance: we are not going to live in a world where AI fully automates expert judgment. Instead, AI will be used to enhance the decisions of experts.

AI in Accounting Today

AI is already embedded in many accounting processes:

• Automated reconciliations.
• Expense categorization.
• Payroll and tax submissions.

But its real promise lies in collaboration, just as Brynjolfsson and colleagues argue. AI can:

• Surface anomalies in audit data.
• Identify unusual tax positions across thousands of returns.
• Strengthen advisory by comparing financial patterns across industries.
• Power predictive cash flow models and valuations.

Take the impairment model as an example. It involves judgment because every debtor’s situation is different. If AI is fed incomplete or incorrect data, it could generate misleading impairment ratios, leading a lender to make poor decisions. Only human accountants can apply judgment, context, and ethical reasoning to interpret such results.

 The Power of Human + AI Complementarity

As the authors show in medical studies, doctors with AI outperform both AI alone and doctors alone. The same applies to accounting: an AI may flag anomalies, but it takes a Chartered Accountant to judge if it’s fraud, error, or legitimate complexity.

Risks of Over-Automation

The article cautions against automation hubris, the belief that AI can seamlessly take over expert work. For accountants, this risk includes:

1. Skill erosion – junior staff lose training if AI handles foundational work.
2. Overreliance – novices treat AI outputs as gospel.
3. Loss of judgment – critical thinking weakens if accountants accept black-box answers uncritically.

Designing for Collaboration

As Brynjolfsson, McAfee, and Jordan conclude, collaboration is often more valuable than premature automation. For accountants, this means:

• Automating the routine: reconciliations, data entry, compliance.
• Collaborating on the complex: tax planning, audit, valuations, advisory.
• Maintaining training pathways so the next generation of accountants can still grow expertise.
• Staying critical of AI outputs, never outsourcing judgment.

Conclusion: Man and Machine

The key insight from Brynjolfsson, McAfee, and Jordan is clear: AI should not be framed only as a replacement. It must be designed to collaborate with experts, not just automate them away.

For accounting, this means AI should free us from repetitive tasks while sharpening, not eroding, our expertise. The future of accounting is not AI versus accountants, but accountants who know when to automate and when to collaborate.

At Qount Accounting, we believe that is the future worth building.

📖 Reference: Brynjolfsson, E., McAfee, A., & Jordan, M. (2024). When Experts Collaborate, They Communicate. The Atlantic.

#Accounting #AI #FutureOfWork #Automation #Collaboration #QountAccounting

More Articles