Basil Thinks AI Can Make Tax Research Less Miserable

The Macro: Tax Research Is Stuck in 1997

Every accountant I have talked to describes tax research the same way: necessary, tedious, and shockingly manual. The tax code in the United States is roughly 10,000 pages. The IRS publishes thousands of pages of guidance, rulings, and notices every year. State and local tax codes add another layer of complexity on top of that. When a CPA needs to answer a specific tax question, they are searching through a mountain of text, cross-referencing regulations, and hoping they did not miss a recent update that changes the answer.

The tools that exist for this are old and expensive. Thomson Reuters Checkpoint and CCH IntelliConnect (from Wolters Kluwer) have dominated the tax research market for decades. They are essentially searchable databases of tax law with some annotations. They work, but using them feels like using a library catalog from the early internet era. The search is keyword-based. The results are overwhelming. The interface assumes you already know roughly where to look.

These are not small companies. Thomson Reuters generated over $7 billion in revenue last year. Wolters Kluwer is similarly enormous. They have no incentive to cannibalize their existing products with something better. That is the classic innovator’s dilemma, and it leaves a gap for startups willing to rethink the experience from scratch.

The accounting industry is also under pressure from a persistent talent shortage. There are fewer CPAs entering the profession than retiring from it. The firms that survive will be the ones that make their existing staff more productive, and tax research is one of the biggest time sinks in the profession. If you can cut a three-hour research task to thirty minutes, you have effectively hired a fractional employee for the cost of a software subscription.

The Micro: AI That Reads the Tax Code So You Do Not Have To

Basil is building AI-powered tax research for accountants. The core idea is straightforward: instead of searching through thousands of pages of tax law manually, you ask Basil a question in plain language and get a sourced answer with citations to the relevant code, regulations, and guidance.

The product is aimed squarely at CPAs and tax professionals, not at consumers trying to do their own taxes. That is an important distinction. TurboTax and H&R Block own the consumer end of the market. Basil is going after the professional tools segment where Thomson Reuters and Wolters Kluwer have been collecting rent for years.

The company is part of the current wave of vertical AI tools that take a specific professional workflow and make it dramatically faster. We have seen this pattern succeed in legal research (Casetext, acquired by Thomson Reuters for $650 million), medical documentation (Abridge, Nuance), and contract analysis (Luminance, Ironclad). Tax research is a natural next target because the source material is structured, the questions are specific, and accuracy is verifiable.

What makes tax research particularly suited to AI is that the answers are not subjective. There is a right answer buried in the code. The challenge is finding it efficiently and making sure you have not missed a recent ruling that changes the analysis. An AI system that can ingest the full corpus of tax law and surface relevant provisions with citations could be genuinely transformative for practitioners.

The risk with any AI tool in a regulated profession is trust. Accountants are personally liable for the advice they give. If Basil produces an answer that is wrong or incomplete, the accountant bears the consequences. That means the citations and sourcing need to be perfect, not just good. Every answer needs a trail back to the actual text of the law. Hallucinations are not a minor inconvenience here. They are a malpractice risk.

The Verdict

I think Basil is targeting one of the best verticals for AI-powered research. Tax law is large, complex, mostly text-based, and served by incumbents who have not innovated in years. The professional user base is motivated to adopt tools that save time because they bill by the hour and every hour spent on research is an hour not spent on advisory work.

The question at 30 days is adoption velocity. How many CPAs are actually using this, and are they using it for real client work or just kicking the tires? At 60 days, I want to see accuracy benchmarks. What percentage of Basil’s answers are correct and complete compared to traditional research? If the answer is anything below 95%, accountants will not trust it. At 90 days, the question is whether the big firms notice. If a Big Four accounting firm starts piloting Basil, the game changes entirely. If it stays a solo practitioner tool, growth will be slower but still meaningful. The market is there. The incumbents are sleeping. The execution bar is just very, very high.