Basis, an AI agent platform for accountants that recently
The site,
Basis co-founder and CEO Matt Harpe told Accounting Today about how his billion-dollar startup has been honoring the accounting profession by naming the conference rooms at its New York offices after accounting and finance pioneers like Luca Pacioli and Alexander Hamilton while also
"There's a lot of talk out there about AI, and people are very excited, and I think rightly so," said Harpe. His company provides AI technology for accounting, tax and audit, and he estimates its technology is being deployed in about 40% of the top 25 firms in the U.S., and over 25% of the top 150 firms. Earlier this month, Sikich, a Top 50 Firm based in Chicago,
"To our knowledge, we're the most deployed AI platform in accounting firms," said Harpe. "The way that we do it is we try to bring together both amazing accountants and some of the strongest machine learning researchers and engineers in New York."
Firms are using it in their accounting, tax and audit practices. "For all those firms, they are actively using it across at least one of their practice areas," he added. "We're not in all the practice areas for all those firms. Generally the way that we approach it is we want to partner with them to help them become the most AI-native practice possible. We only really work with firms that are deploying across the entire practice. It's not something where one person or 10 people can use it. For all those practice areas, we think of Basis being essentially the core of their AI strategy in that practice area."
New York AI
The company has all 120 of its employees working in its New York offices, where it competes with AI giants like OpenAI and Anthropic for talent.
"Our goal is to be the home for applied AI in New York," said Harpe. "There are fantastic engineers and researchers in New York. We're very lucky that a lot of them have chosen Basis. They want to work on frontier challenges in applied AI, not building another chatbot or just bolting something onto an existing product, but actually figuring out how we can build the most sophisticated agentic systems that actually are accurate enough to be used in real production workflows. The other part is they want to contribute to doing real work in the real economy. I think people find that exciting, that these agents are actually doing real things. They're not just something that people are asking silly questions on the side. If you look at the number of companies that are doing that, most of them are not headquartered in New York. I think we're really one of the few organizations that is at the frontier across any domain, not just in accounting, but across any domain, in terms of building these AI systems."
All of Basis's R&D efforts happen in New York, not in Silicon Valley like many AI companies. "Obviously we're not the size of OpenAI or Anthropic from a headcount perspective, so there's an opportunity for people to take on real responsibility in building these systems, as opposed to being one of many," said Harpe. "That's appealing from that perspective."
Unlike those AI giants, his company is focused on the accounting profession. "Our mission is to figure out how we can actually make the AI as capable and accurate for accounting as possible, and when you do that, it ends up being useful," said Harpe.
Making AI reliable enough for accounting
He believes AI can reliably automate tasks for accountants. "Our objective really is to give every accountant at the firm this AI that can go out and take on a bunch of the most manual parts of their job and essentially report back to them and complete these things on an ongoing basis," said Harpe. "I think we have a strong track record of actually driving outcomes. AI can be relied on because it's accurate enough and it's capable enough to take on these complex workflows with a high level of accuracy."
Generative AI large language models have been accused of "hallucinations," making up deceptively plausible-looking information from the massive amounts of data they vacuum up from the internet and social media. Harpe believes the models are getting better now.
"I actually think the trust part is so important in accounting," he added. "Accounting is all about trust. That is a core part of what the profession is dedicated to doing. You want to systematize and structure all the economic activity that happens in and around these organizations, and then know you can trust that synthesis. That is the job of accounting, so trust is very important, and I think that most AI companies are not actually focused on accuracy enough. The thing that we prioritize above everything else is capability and accuracy of AI."
Basis has a Deployed Intelligence team that helps accountants build more trust in AI at their firms by demonstrating that accuracy.
"We have this program that we go through where we have them take workflows that they are very familiar with, and they know in and out," said Harpe. "We have them use the AI to start working on those workflows, so they can see directly what it can do and what it can't do. That helps them build intuition and trust. It's as if you were onboarding a new person on your team, and you give them some things, see how they do, and then move on to the next thing. We help them build up that trust as part of that process."
A big part of that trust involves preventing confidential client information from being sucked into the public-facing AI systems.
"It comes down to trust as well," said Harpe. "Clients trust the firm. That relationship is very sacred, and now you want to make sure that you don't end up putting that data in a place that's then going to get exposed in some way. It's very important to have really clear security policies and clear AI policies, since you can essentially guarantee to firms this data will be fully siloed. We actually don't just silo data firm by firm. We actually silo it client by client, so data that's existing for one client can't physically be leveraged in another setting. We have really strict security guarantees around that. Firms have become more sophisticated in thinking about what their policies are going to be when they're working with partners to make sure that the data is being used properly."
He's found that firms now have detailed questionnaires asking technology providers like Basis . about how they handle specific situations and the cybersecurity certifications they hold.
"People are definitely getting more sophisticated and making sure the proper controls are in place," said Harpe. "We are religious about making sure that the integrity of the data is upheld. No data can be trained on public models, and no data can potentially get from one client instance to another client instance."
AI's impact on jobs
Many accountants are concerned about losing their jobs to AI, and while the tools can be useful from an efficiency perspective, they can also be perceived as a threat.
"I'm a big believer that there is a lot more accounting that needs to be done in the world that is not done," said Harpe.
He cited his previous experience working on developing machine learning systems for the health care industry. "To my understanding, there's no hospital in America that can actually tell you how much it costs them to provide a procedure to a patient," said Harpe. "If you're going to get an MRI in a hospital, the hospital doesn't actually know how much it costs them. This is a problem, because if you're a hospital, one of your main objectives should be how to reduce the cost of care over time. Either you can increase your margins or provide more cost-effective care to patients. The fact that we can't do that, in my view, is an accounting challenge. The job of accounting is to structure and systematize all the economic activity that happens in and around an organization, and after spending all your time focusing on the compliance activities, there are all these other things you can't get to. I'm a big believer that if we can free up people from spending time on these very manual aspects of the job that everyone looks at and says why can't we have a computer doing this, then it will allow people to spend time on many more things. I actually think it's actually possible that we'll have more accounting jobs in five years."
He compared it to the impact of AI on software engineering, and how such employees are still very much needed. "Software engineering is ahead of accounting in terms of the adoption of AI," said Harpe. "No engineer in our company is writing a meaningful amount of code anymore. Pretty much none of them write code, but of course, that used to be a core part of what they did all day long. But our engineering and our machine learning teams are as busy as ever. The reality is it allows us to build things that we previously wouldn't have considered building because people are more capable."
When asked what his company is building with the funding it has received so far, he said most of it goes to hiring and computing costs. After initially focusing on outsourced accounting technology as its bread and butter, Basis has expanded into tax and audit, and it's dedicating the funds to continuing to build up teams that can support firms across the country in those areas. Basis has so far received over $130 million in funding from investors including Accel, Khosla Ventures, Google Ventures, BTV, NFDG, Box Group and former Goldman Sachs CEO Lloyd Blankfein. They are looking beyond the recent volatility in tech stocks.
"We have investors who are really long term oriented, so we are focused day in and day out on doing the long-term building that we think is important, and not on the day to day," said Harpe.
Basis's Ode to Accounting website is also about the long term, relying on
"One of the things that we care a lot about is making it clear that accounting is actually very important, and accounting has a very important role to play in this coming AI era," said Harpe. "We were working on Ode to Accounting for the better part of a year, and it's all about the history of accounting and how we got here. The first artifacts of human writing were accounting. In our view, the underpinning of how all market economies work is the ability for us to be able to understand what's happening in each granular organization's decisions. Sometimes the profession loses track of that, and in this next era, as the economy is going to get more complicated, it's actually important to recognize that accounting is core to giving us the legibility to engage in these more complicated forms of economic interaction where you have AIs that are interacting with each other."
