AI is coming for accounting after coding

The central argument is that accounting is harder to automate safely than coding or legal work because it demands near-perfect accuracy under relatively fixed rules. Contracts rely on formal language and can often remain usable despite a weak clause; software can tolerate some bugs. Accounting errors, by contrast, can distort a company’s entire financial picture. The example given is Campfire, which reportedly raised $65 million to build an accounting system from scratch but achieved about 95% accuracy. Across 100,000 ledger entries, that rate would leave roughly 1,000 errors - enough to make human review necessary and potentially undermine the accounts.

The argument also emphasizes the consequences and accountability attached to financial mistakes. Losing reliable financial records could threaten a company, while businesses may be reluctant to entrust their numbers to a system they cannot independently understand. Although human accountants also make errors, they can be sued and held legally responsible, giving them an advantage AI lacks. The piece distinguishes automating bounded bookkeeping tasks from replacing accounting judgment and responsibility, and says competition from major AI labs will push margins down. It further points to Sarbanes-Oxley requirements for human finance officers and controllers as evidence that accounting still depends on accountable people.