Combinely Wants to Be the AI That Actually Understands Your Tax Return

The Macro: Tax Firms Are Drowning in Repetitive Work

I spent a week talking to accountants last year about their workflow during busy season and the recurring theme was brutal. Not “we need better software.” It was “I am working 70-hour weeks doing tasks that feel like they should be automated but aren’t.” Document collection alone eats an absurd number of hours. Chasing clients for their W-2s, matching uploaded files to the right line items, following up on missing schedules. It’s tedious, error-prone, and happens thousands of times per engagement.

The accounting software market is massive and well-served at the surface level. Intuit has TurboTax and ProConnect. Thomson Reuters has UltraTax. Wolters Kluwer has CCH Axcess. These are deeply entrenched products that firms have used for decades. But they’re fundamentally tax preparation software. They help you fill out the return. They do not help you with everything that happens before and after the return gets prepared.

That’s where the gap is. The actual time sink in a modern tax practice isn’t the computation. It’s the communication, the document management, the quality review, the back-and-forth with clients who don’t know what a K-1 is. AI is theoretically perfect for this kind of work. Pattern matching against prior year returns, extracting data from uploaded documents, answering routine client questions, flagging inconsistencies in draft returns. The question is whether anyone can build an AI product that accountants actually trust with sensitive financial data.

Trust is the real barrier here. Accountants are professionally conservative people. They have to be. One wrong number on a tax return can trigger an audit, a penalty, or a malpractice claim. Any AI product in this space needs to be right not most of the time but essentially all of the time, with clear audit trails showing exactly where every number came from.

The Micro: Ex-Deloitte and Google Building for the Firms They Know

Combinely is an AI coworker for tax teams. Not a chatbot bolted onto existing software, but an integrated platform that handles document collection, client queries, and tax return production and review. The pitch is that it saves accountants roughly 90 minutes per day during busy season, which for anyone who has lived through a tax season, is the difference between a manageable workload and burnout.

The company was founded by Tom Invernizzi and Arthur Granacher. Tom is the CEO, coming from Deloitte and PWP, which means he’s lived the big-firm tax grind firsthand. Arthur is the CTO, with engineering experience at Google and Shazam. The combination is the classic “one person who understands the problem, one person who can build the solution” pairing that I tend to find compelling. They’re a five-person team, part of YC’s Spring 2025 batch.

What separates Combinely from a generic AI assistant is the workflow integration. The product includes an Outlook add-in, an Excel add-in, and a web interface. It plugs into the tools accountants already use rather than asking them to switch to a new environment. The tax research feature provides AI-powered guidance grounded in actual tax code, not hallucinated legal citations. The Excel add-in verifies calculations in place. These are small details that signal the team understands how accountants actually work.

They’re SOC 2 Type II certified and GDPR compliant, which is table stakes for handling financial data but still worth noting because plenty of AI startups ship first and worry about compliance later. Combinely’s client list includes firms like Andersen and DFK International, which are not small shops. Having enterprise clients at this stage suggests the product is performing well enough that serious firms are willing to put their name next to it.

The product covers three areas: tax, audit, and accounting. That’s ambitious for a five-person startup. Most companies in this space pick one vertical and go deep. Combinely is spreading across three, which could mean they’ve built something genuinely flexible or that they’re stretching too thin. The fact that they started with tax and expanded outward is encouraging. Tax is the hardest of the three to get right, and if the core engine works there, audit and accounting are natural extensions.

The Verdict

I think Combinely is well-positioned in a market that’s about to get crowded. The AI-for-accounting space is heating up. Keeper Tax has built a popular tool for freelancers. Bench automated bookkeeping before pivoting. April is going after consumer tax prep with AI. But the enterprise accounting firm market, the one Combinely targets, is still relatively open. The incumbents like Thomson Reuters and Wolters Kluwer are adding AI features, but they’re doing it slowly and within the constraints of legacy architectures.

The risk is that the big incumbents eventually catch up. Thomson Reuters has more data, more integrations, and more customer relationships than Combinely will have for years. If UltraTax ships a genuinely good AI assistant, switching costs could keep firms on the incumbent platforms. Combinely’s advantage is speed and focus. They can iterate faster, respond to user feedback faster, and build workflows that the big platforms won’t prioritize.

Thirty days, I’d want to see expansion metrics. How many firms are using it daily versus trying it once? Sixty days, the question is whether the Outlook and Excel add-ins are driving adoption or sitting unused. Ninety days, I’d want to understand average time savings per user. If accountants are consistently saving that 90 minutes, the product sells itself through word of mouth. Tax is a referral-driven industry. One partner tells another partner at a conference, and deals happen. The foundation looks solid. Now it’s about proving the time savings at scale.