My passion is, and has always been, how technology can make our profession more efficient, leaving us with higher-value work such as serving and advising clients, not toiling in spreadsheets and manually moving data between systems. In the 2010s, this was through deterministic code and SQL queries; today, it is through AI agents working alongside us as virtual team members.
The pace of innovation in AI has felt exponential. Model intelligence is surpassing what we thought possible just a year or two ago. The latest models are solving decades-old mathematical problems that human mathematicians have relentlessly tried to solve. Because of this pace, there is a ton of excitement about where and how this technology can be used. A few things have become clearer and clearer to me, forming a sort of set of “AI use” maxims.
The highest and best use of AI is “doing the work”: giving AI agents a properly secured sandbox and letting them work their magic on messy files or unstructured data to get that information into the GL, a database, etc.
The lowest and worst use case of AI is substituting for people in your personal life or fully substituting for high-trust advisers. Human relationships are built on selflessness, care, and trust. Call me old school, but if you insert an AI agent into our friendship, you are being selfish with your time, showing that you don’t care to communicate from the heart, and diminishing trust by bringing a vendor into our relationship. It is weird, dystopian, and not for me.
Open-source and local models will play a significant role in our profession over the coming years… more on this below.
In addition to the maxims above, a few concepts deserve a bit more explaining.
I am extremely excited about the increase in open-source technology in accounting and finance teams and within firms. More choice and competition benefit everyone, and for too long, this group of professionals has largely been at the mercy of bigger firms and their products because the alternative choices were limited and the cost to build their own was too high.
This bottleneck is no longer in place, and I’m excited to help push open-source tooling in this space and collaborate with other like-minded professionals. The frontier AI labs have built incredible products that provide an incredible amount of value. But the future of intelligence will thrive as users have more choice. That means choice in the intelligence model, choice in the harnesses used to run the models, and choice in the hardware the models run on. Problems rarely have one-size-fits-all solutions. Having a variety of choices will allow a greater number of problems to be solved and uplift everybody.