Something went wrong while submitting the form. Please try again.
Annette Garcia joined Basis in 2024 as our fifth full-time employee. Today she shares how she got here, talks through her work at Basis, and explains how accountants like her are building the future of the profession at the frontier of applied AI.
Annette: My journey to Basis started at the University of Florida, where I studied accounting. I did not enter college thinking I was going to be an accountant. I started pre-med and quickly realized I did not want to work in a hospital. Accounting felt like a practical way to understand business. If you can read financial statements and understand the numbers, you can take a lot of different directions in your career.
After that, I earned my master's in accounting while starting my career at Kaufman Rossin, where I worked as an auditor. I was reviewing financial statements for large private companies across many industries: manufacturing, distribution, real estate, healthcare, finance, produce, and so on.
I really enjoyed audit because it is a way to study how different companies actually do their accounting. You see why practices vary from industry to industry, why processes look different, and why accountants make certain judgment calls. It is a great way to observe and learn.
After a few years, I wanted to go in-house, so I found an opportunity with OpenStore. I spent a lot of time there working with engineers to build internal tooling and reporting by pulling data from different software systems. That was when I realized how much I enjoyed the technical side of the work.
So when I discovered Basis, it was the perfect next step. I wanted to join a startup at an earlier stage. I wanted to take on more risk. I wanted to solve harder problems. I could see the team was serious and wanted to work hard. And they were building AI software for accounting: an open field where we get to make order out of chaos. That was very intriguing to me, and it has only become more so over the years I have been here.
Annette: Let's take a step back. On the one hand, accountants are leaving the profession. On the other hand, the global economy is demanding increasingly complex accounting for more sophisticated financial transactions and business operations.
Our overall goal at Basis is to help the accounting profession meet that demand through more and better accounting. With agents, accounting can happen faster and more deeply, which gives people room to think about larger issues and ultimately make better business decisions with sharper information.
Annette: Here is a small example. Today, an accountant might only capitalize fixed assets above a certain threshold. If a company is buying chairs, tables, and computers, the accountant might say, "I am going to capitalize these as fixed assets only if each is at least $5,000." Anything below that may not be worth the time to reconcile, track, and depreciate. This is normal. Accountants have plenty of other things to do.
Now, with AI, that constraint changes. If an agent can look at the transaction, apply the right policy, run in the background, and keep following the same instructions month over month, why not account for more of it? Why not have more fixed assets tracked and depreciated over time with a lower capitalization threshold, say $1,000? That does not mean judgment disappears, but it changes what is possible.
So overall, we can do more accounting, and do it better.
Annette: We make sure that Basis agents produce work that meets the standard of professional accountants. We are the translators of accounting jargon, knowledge, and workflows. Our job is to ensure what we build actually works for accountants and the clients and companies they serve. We are accountants who care about making other accountants' lives better.
It took over a year after I joined for us to figure out a title for it, but this team emerged from a desire to work very closely with ML engineers, the people building the agents. It is hard because AI is still so new and there is no playbook. The role is essentially agent manager, context engineer, intelligence architect, and quality assurance engineer all at once. We are asking accountants to rip apart their career paths, enter something entirely new, take a big bet, and hopefully have a lot of fun doing it.
The product team owns the vision for what we are building and how it will show up in Basis: how a module should look, what features belong in it, and what a workflow should feel like. They ask APO questions such as: What do accountants want? What do accountants not want? What flow of buttons in the UI would feel natural to an accountant? Where in a workflow should certain accounting information be presented? They expect us to bring user empathy and accounting judgment into product decisions.
The engineering team builds the infrastructure and code that support accounting workflows in Basis. Then APO comes in to train the agents how to actually think like accountants. In a way, APO functions like an engineering team because we are still writing code; it is just code in English rather than in Python.
Annette: There is no reason why someone cannot go to a regular chatbot and ask for accounting work to be done. But accountants like things done in a particular way. They like consistency. They like having an audit trail. They like predictability. They like knowing how systems work. And because AI can be non-deterministic, it is hard to get risk-averse accountants to trust it.
That is the real problem. It is not just whether an AI model can produce an answer. It is whether an accountant can understand how it got there, whether the answer is expressed in terminology they recognize, whether the same preferences are respected month after month, and whether the output holds up when someone reviews it later.
Ask a regular chatbot to build a lease amortization schedule twice and you might get two wildly different answers. Ask Basis, and it respects your preferences and already knows how to construct that schedule to a professional standard. The accounting expertise is built into our agents, so you get consistent outputs without having to spell out every step yourself.
Ultimately, it is about building a system that makes accountants feel like Basis is another staff accountant on their team, not a contractor being asked to do something for the first and last time ever.
Annette: Writing is number one. Everything we do boils down to written language: how you communicate with agents, with others on the team, with users. Effective writing means conveying your message so clearly that it is self-contained and a reader with absolutely no prior context - whether a human or an AI agent - can understand what you wrote and have no open questions.
Another thing you need is agency and curiosity. You have to be okay with starting from a very low point in your knowledge and growing through a steep learning curve. Doing the work is the only thing that builds intuition for how agents work and how the Basis system operates.
This ties back to one of our operating principles as a company: Seek truth together. Sometimes there is no clear right answer. You get to propose what you think is right. You get to fail. You get to fail again. Then you can win. But you need a very high tolerance for that process. You have to accept that getting to the right answer takes patience, and you have to be willing to hash it out together as a team.
Annette: One thing I am excited about is extending the Basis platform from CAS to tax and audit. We started with CAS because it is probably the truest reflection of traditional accounting. Basis has to know how to do core accounting before it can perform audits or handle taxes. Now that we are confident Basis agents can do core accounting well, we can build on that and expand into audit and tax.
Audit and tax each bring their own complexities. Tax has a huge regulatory surface area that changes constantly. Audit requires a lot of professional judgment about what to test, what evidence is enough, and how to support a conclusion. Teaching agents how to navigate those decisions is a different problem than the core accounting work we started with. Our team needs to build new muscles to win in those domains.
