I, personally, don’t see this as a Tiller Killer, but it’s going to have an impact on some of the other apps in the market like Simpifi and Monarch. It’s only being offered in the Pro version of ChatGPT, so $100 a month. Nobody is paying that for personal finance. But, if it turns out to be secure and there are assurances that they’re not training their LLM’s on our banking information, it could be a strong player in the market.
I do think Tiller is going to have to come up with their own AI strategy for customers though. And quick. I can certainly pay for my own subscription to an AI tool and use it in my Tiller sheets, which I believe is a better way to go. I really don’t want OpenAI having access to all my financial information…
Current models of AI are good at producing chat and meeting summaries. They are probabilistic. They can guess accurately what to say or how to respond. Math needed for financial advice is not probabilistic but deterministic. AI is terrible right now at that math. The reason you have to opt out of the training is that ChatGPT wants to change that. But they want you to pay them 100/month for a guessing game. No thanks. The optimum model for financial deterministic math is the database. Spreadsheets are good too but they are flat file. They have their limits. Whereas a database can be optimally designed (thanks to Kimball modeling and data warehousing & star schema techniques) to produce accurate deterministic math and reliable front ends (such as Power BI) which is why Chief Financial Officers use them as their chief strategy tool. You are right to be skeptical of ChatGPT’s aggressive reach into your bank data.
Interesting. If someone is already a Chat Pro user, then it could be a no brainer to cancel your financial SW sub, if you have one.
However, I’m not going to pay $100/mo if my primary use case for ChatGPT is to get access to this financial tool. That isn’t remotely competitive in the market if I’m not also going to be a heavy user of ChatGPT.
My wife has a Plus sub for her small business and I piggyback on her account when I need to, but don’t see need to upgrade from that. I do think AI is going to be a major disruptor in this space though.
edit: However, if/when AI gets smart enough to act as your Financial Advisor on top of being your primary financial software, then $100 per month is a damn bargain. Robo advisors already exist and have for years at the major firms; AI will just accelerate that. High net worth folks are still going to want a human to talk to though so I think the disruption here will be in the low to middle net worth segments.
You hit on a really important distinction here regarding what AI is marketed as versus how it actually functions. LLMs cannot predict the financial future; they are next-token prediction engines, meaning they are only as good as their underlying data structure.
A lot of people confuse AI with ‘robo-advisors,’ but right now, robo-advisors are mostly just automated scripts that rebalance basic 60/40 or 80/20 stock-to-bond relationships. Even the ones trained on technical market indicators aren’t any better than the ‘gurus’ who programmed them.
To actually forecast financial instruments successfully requires a specific, proven methodology—like the contrarian approaches David Dreman has chronicled. Unless an AI model is specifically trained on that level of market psychology and fundamental logic, its forecasting authority is zero.
From a technical standpoint, if you ask an LLM to forecast a portfolio by just feeding it raw spreadsheet rows without a proper deterministic semantic model underneath it, it will eventually choke on token limits. When it does, it doesn’t do the math—it just hallucinates its ‘best next guess’ to give you an answer. I definitely don’t want a hallucination managing my portfolio, and I’d caution anyone else against relying on it either.
As @casilverthorn.96 noted it’s something we’re exploring. We’ve transitioned away from the GPT noted in the original post she shared and have built out an MCP server that you can hook up to Claude or the AI tool of your choice and have shared it with some of the people who signed up for the waitlist and are gathering feedback to determine next steps.
@MelonCollie what would be the primary use case you’d hope an AI integration could solve for you that would make it worth paying for? (Not $100 / mo, but also not free).
That’s a great question Heather. It’s really THE question, isn’t it?
I think AI brings the ability to ask questions of our data. No more building a sheet to find all the transactions in the last 3 months that are categorized as “travel” and have the tag “Europe.” Just ask it. Or, ask the AI to build you a dashboard that looks and feels exactly the way you want it to (and that’s where Tiller has an advantage over the other guys like Monarch and Simplifi.)
I can easily pay for Gemini and ask questions of the data, and have it build the dashboards for me, But I would rather know my data was not heading into the gaping maw of Google’s LLM’s. If it costs more than Gemini or the others, I don’t think it makes sense. Maybe Tiller’s value add is just some really creative prompts, or the training on how to make the most of the new tech.
The first use case that comes to mind for me (and one that I’ve considered attempting in the near future via other tools) is something to the effect of: “Build me a custom dashboard in a new sheet that shows ____ with these specific features.”
I can find a lot of what I need via the templates and community solutions, but not everything, and not custom tweaked to my exact preference (and I don’t consider myself a spreadsheet expert, so I struggle with tweaking premade templates sometimes).
An interesting video by Rob Berger talking about his experience trying the new ChatGPT finance connection. It wasn’t all positive by any means. I still like the idea of downloading my transactions into my own spreadsheet and using AI to build out dashboards and views exactly the way I want them. I just don’t like the idea of connecting my bank accounts directly to OpenAI.
Just to add to my earlier post, I’m in no way going to hand my financial data over to one of the public AI offerings (especially OpenAI). I might consider paying for a Tiller add-in for this under the assumption the AI tool is fire walled on Tiller’s server. I’m not sure if that degrades the ultimate quality of the AI tool, or not.
