Author: Milutin Todorovic
Graphics: Holly Ur
A promising startup, Otterz, is replacing human accountants in a field that’s lacking client interaction and development.
Accounting Crisis: Why Now?
There won’t be enough skilled labor to match the rising demand for corporate accounting services, with 75% of CPAs already reaching retirement age as of 2019. Such a labor shortage stems from two key trends. The first is that the complexity of the tax law in the United States will increase because, while firms that operate in sectors like manufacturing fit the traditional accounting mold, the pharmaceutical and technology firms do not. In NVIDIA, for example, R&D might be seen as a capital expenditure but is treated as an operating expense. The second is that accountants are spending ~60% of their time chasing queries and data, but only 10-20% on actual advisory and down-to-earth client interaction.
These trends matter in today’s bull market because customers want more human interaction in accounting services, but CPAs can’t provide that since they are so often bogged down in manual work – only Artificial Intelligence can.
The AI Landscape
It is undeniable that AI is the buzzword of the moment, emerging in every aspect of our lives and the market. It’s only a question of whether the firms believe the buzz is merited or just a fad.
The competition in the market can be viewed as a step-like progression. Traditional, fully manual accounting firms like Deloitte, PwC, and other CPA firms have billable hour standards and workflows driven by legacy hardware. At the next level, firms like Zeni, Bench, and Pilot use AI assistants – but humans must click buttons, log in, and reconcile discrepancies in data. The logic of their workflows is still human. At the final stage, Flywheel, Black Ore, and Digits have been using more hypercharged versions of these AI assistants to act as Copilots – but Copilots still require human validation. Thus, there’s no full autonomy.
The AI models these firms use are general-purpose LLMs (large language models) from companies like Anthropic, OpenAI, and xAI. One problem is that LLMs cannot do accounting, because they rely on neural networks to predict the next word, which is inherently probabilistic and can have multiple outputs. The accounting profession requires 100% accuracy & precision and only one output, which means there should be a deterministic alternative. Not even fine-tuning LLMs can provide the same level of expertise that a deterministic AI can.
Another problem of an LLM is hallucination. A Stanford University study found that LLMs hallucinated 58% of the time on legal and accounting queries. In Johnson v Dunn, attorneys were caught fabricating entire citations using ChatGPT. In an accounting case study, Anthropic was asked about 1031 exchange specifics, but its answer was incorrect and generic:
The Autonomous AI in Accounting (Otterz)
Founded by Tapan Ramachandran and Semih Korkmaz, Otterz (Skydeck Batch 16) is an accounting department at the tip of your fingers that is not only trying to automate tasks like P&L dashboards, but make them fully independent from the human hand. Otterz can create its own code or access the web to solve contextual problems with precision and accuracy.
For instance, when asked to prepare a tax return from a single prompt, Otterz creates a 23-step plan, writes code, researches IRS rules, stores knowledge, and finally generates a return. The key is that it doesn’t “guess” the tax form but de facto makes it.
In the same accounting case study mentioned before, Otterz correctly identified the specifications of the 1031 exchange:
A16z’s recent thesis posited that AI is turning capital into labor. Capital buys coffee, engineers, and GPUs. Thus, the transformation will open up new software markets and replace the “per seat” pricing standards that restricted large outcomes. Otterz doesn’t sell accounting software “per seat” and also captures labor budgets, which means it proves the thesis. That is significant since the $300B/year spent software market is infinitesimal compared to the white-collar labor market, which represents the future value Otterz can capture.
Otterz’s low price point also reduces labor time through its semantic network, which further accords with the thesis. This is especially relevant in an era of poor corporate governance and weak accounting standards for many U.S. companies.
Beyond Automation: AGI in Fintech
Simply relying on and scaling LLMs will diminish the scope of their utility as people feed them more data. Artificial General Intelligence can most directly be achieved through autonomous AI, because finance is a forward-looking field based on research and trend observation that demands bigger capacity independent of manual work.
As small businesses continue to search around old boxes and write down accounts receivable by themselves, the growing lack of manpower should prompt them to take a look at Otterz’s website, which explains why a financial accountant can exist right by their side. It’s a solution that will open up the floodgates for small businesses to squeeze through the downsides of barriers to entry and provide an even greater impact on the development of accounting as a profession.
Take-Home Points
- There will be fewer CPAs but more demand for face-to-face customer interaction
- Current market players are utilizing AI as more of an assistant and not to its full capacity
- LLMs are worse because they are probabilistic, which is incompatible with accounting
- Ramachandran and Korkmaz’s Otterz is leading the next gen of autonomous AI