Azta.ai is a Hungarian startup developing an AI-powered research tool specifically for tax and business law. The company was founded and is led by Máté Márk; the engineering team is largely based in Szeged, while operational work is run from an office in Óbuda.
Key developments
- The product idea emerged in November 2024, when Márk could not find a clear answer to a tax-related question about his sole proprietorship.
- By 2026 the team had grown the system and its user base: the service reports about 14,000 users and roughly 150,000 answered questions to date.
- In April 2026 Phantic Kft., the company behind Azta.ai, closed a €400,000 pre-seed round (the report converts this at an exchange rate of 385 to HUF, equal to about HUF 154 million). The round was led by angel investors including a participant from the Barion pre-seed.
What the product does and how it works
Azta.ai is more than a document search engine. Its database contains laws, decrees, official resolutions, court decisions, Kúria rulings and NAV (Hungarian tax authority) positions. The system stores previous versions of laws going back ten years so that it can assist in long-running cases; the team reports having 1.8 million Hungarian documents.
Technically the product uses several approaches:
- a proprietary search module decides which sources to consult for each question;
- a graph-based knowledge representation maps relationships between documents and legal concepts;
- a Monte Carlo–style search method helps find relevant material efficiently within the multi-million document corpus.
The developers emphasize source citation for every answer and that the system should flag when it is uncertain or will refuse to answer (for example, on criminal law matters). Azta’s website cites 98% answer accuracy; prior to market release the team runs external audits and multi-hundred-question tests to validate accuracy.
Product development timeline
Márk initially tried large language models (ChatGPT, Claude) but encountered hallucinations and inaccuracies. Early versions fed with general online professional material produced many incorrect answers, which led the team to filter sources and gather verified data. Because parts of the National Legislation Database are copyright-protected, the team compiled a legal code from official gazettes and public releases, a process that took about four months.
Key team members include co-founder and Chief Revenue Officer Sándor Attila, lead AI researcher Kővári Bálint, and investor AI Angels Zrt. The majority owner of Phantic Kft. is Márk.
Market, clients and partnerships
- The majority of Azta’s usage comes from small and medium-sized businesses and accounting firms rather than large corporate clients: about 95% of answered questions originate from that segment.
- The startup currently has 600 paying customers and monthly revenue around €15,000.
- Azta cooperates with the Vállalkozók Országos Szövetsége (VOSZ) and the Magyar Könyvvizsgáló Kamara Oktatási Központja (MKKOK). The MKKOK’s training participants receive trial access and discounted purchase options.
Azta offers three subscription tiers: a free entry-level plan (five questions per month), a Standard plan recommended for sole proprietors (HUF 4,990), and a Pro plan (HUF 9,990) with unlimited monthly questions and premium features such as company search and central bank exchange-rate lookup.
International expansion
Technically the team has prepared legal data for 17 European jurisdictions, but currently only Hungarian and Polish users are active. The first foreign market entry into Poland was executed in a rapid, five-day effort: ahead of a pitch event the team collected Polish tax law, localized the site, ran tests with two thousand questions and had initial Polish users by Friday.
Romania is planned as the next market, targeted for 2026. For broader international growth they plan a €3 million funding round; the company states that €700,000 is already committed. The intended allocation is €1.5 million for market entry and €1.5 million for organizational development and brand building, aiming to prepare go-to-market for 10 European countries and later explore a U.S. expansion after processing the full European legal database.
Competitive positioning
Large legal AI platforms such as Harvey and Legora are present in the market. Azta positions itself not to directly mimic them but to focus on under-served niches — specifically local and regional tax practice and business law — and to serve a client base beyond pure law firms.
Advisors and partners note that AI should improve the efficiency and quality of expert work rather than replace professionals. Azta’s approach emphasizes traceable citations, client-supplied documents, and expert review loops to manage legal and financial risks.
Accuracy, risks and limits
The company acknowledges model limitations: large language models can infer by similarity, which in legal contexts is risky (an example given is classifying lemons for VAT purposes). Azta is working to secure professional liability insurance for its answers so that responses could carry insured responsibility in the future.
Leadership goals and near-term targets
Founder Máté Márk has set ambitious targets: build the largest tax and business law knowledge base in Europe within the year and achieve rapid revenue growth. The team believes that with ongoing contracts, local corporate implementations and integration partnerships it is realistic to reach €90,000 monthly revenue by late summer — about a sixfold increase in a few months.
They plan to run a larger funding round by September to support expansion, with a three-year horizon focused on measurable go-to-market execution rather than long-term speculative planning. The company stresses cautious, audited rollouts and aims to be first into as many markets as possible while maintaining accuracy and expert oversight.
