Bluebook Is Betting That AI Can Save Accounting From Itself

The Macro: Accounting Has a Demographics Problem Nobody Talks About

Here is a statistic that should alarm anyone who files taxes: 75% of CPAs are expected to retire within the next 15 years. The pipeline of new accountants is not keeping up. Fewer students are enrolling in accounting programs. The 150-credit-hour requirement to sit for the CPA exam is an expensive barrier that steers people toward finance or tech instead. The profession is shrinking at the exact moment when regulatory complexity and reporting requirements are expanding.

This creates a real business problem for accounting firms. A typical firm has partners who manage client relationships and junior staff who do the actual work. When you can’t hire enough junior staff, either you turn away clients or your existing people burn out. Most firms are doing both.

The software that exists today doesn’t solve this. Xero, out of Australia, is solid general-purpose accounting software. Silverfin does practice management for firms. Botkeeper tried the AI bookkeeping angle early and has traction but mixed reviews on accuracy. Vic.ai focuses on invoice processing with AI. None of these are solving the staffing crisis directly. They make individual tasks faster, but they don’t change the ratio of clients to professionals in a meaningful way.

That ratio problem is the actual opportunity. If software can take a CPA from managing 20 clients to managing 200, you don’t need to fix the talent pipeline. You just need the people you have to be dramatically more productive.

The Micro: Nordic Traction and a Self-Taught Founder

Bluebook builds AI agents for accounting firms. The product accelerates research, reconciliations, and monthly closings. The headline metric is a 30% reduction in research time, which sounds modest until you realize that research is the thing junior accountants spend most of their day doing. Looking up tax code. Checking precedents. Cross-referencing client situations against regulatory guidance. It’s important, detail-oriented work that is also perfectly suited to AI.

The company already has 30 leading firms in the Nordics using the product. That’s real traction, not a waitlist number. Nordic countries have some of the most digitized accounting systems in the world (Sweden and Finland went digital-first on bookkeeping years ago), so it makes sense as a beachhead market. The firms there are more likely to adopt AI tools quickly because their existing workflows are already software-native.

CEO Philip Andersson is a self-taught AI builder who founded his first company at 17 and held engineering leadership roles at FirstVet, Mindler, and Epidemic Sound. That’s a mix of healthcare, mental health, and music tech, which sounds scattered until you notice the pattern: all regulated or compliance-heavy industries where software has to earn trust before it scales. Co-founder Filip Stal rounds out the team. They’re a four-person operation out of San Francisco, YC Winter 2025 batch. The team is small for the size of the problem, but 30 paying firms suggests they’re building the right thing.

The Verdict

Bluebook is one of those companies where the market timing is almost too perfect. A massive labor shortage meets an AI capability that’s finally good enough to handle the nuanced, text-heavy work that accounting requires. If they execute, the ceiling is very high.

The thing I like most is that they’re not trying to replace accountants. They’re trying to make each accountant ten times more productive. That’s a much easier sell to the industry than “our AI does your job.” Accounting firms will pay for tools that let them take on more clients without hiring. They will not pay for tools that threaten their business model.

The risks are real but manageable. Accuracy in accounting is not optional. A single tax filing error can cost a firm its reputation and a client real money. Bluebook’s AI agents need to be right essentially all the time, and “essentially” is doing a lot of heavy lifting in that sentence. The 30% research time reduction is promising, but I’d want to know the error rate on the research output. If it’s saving 30% of time but introducing errors that take 15% of time to catch, the net gain is thinner than it looks.

Over the next 30 days, I’d watch for expansion beyond the Nordics. The U.S. accounting market is enormous and the staffing crisis is arguably worse here than in Scandinavia. At 60 days, I’d look for whether firms are expanding usage beyond research into reconciliations and monthly closings, which is where the real time savings compound. By 90 days, the question is whether Bluebook can maintain accuracy at scale with a broader set of accounting scenarios. The Nordic firms are a strong proof point. The U.S. market is the prize. I think they have a real shot at it.