So, what do you get when you add ChatGPT with Tiller? I call it Tiller on steroids. Not good at creating pie charts and bar graphs, or even if you are good at doing those things. Imagine just talking to ChatGPT and asking it what you want and you will get it. I got this ability by exporting my Tiller sheet to Excel and then uploading Excel into ChatGPT. Ask it to slice and dice your data any way you want, no Google Sheets or Excel experience or knowledge needed other than just basic skills. You will be amazed by what you can get and get it quickly. Review the spreadsheets you get back and if you want to make changes just tell ChatGPT what you want to change and it’s done. I have spent much time with ChatGPT, especially over the last month, and been able to do amazing things; e.g., create tax projections for my kids, do a break even analysis on whether to purchase a second home, create a career and retirement spreadsheet for my son based on some very specific criteria he wanted to consider, creating an exercise schedule by day for me, and having it draft and send emails for me. Have there been some glitches along the way, sure? I think some of it’s just a learning curve as well. You should be able to do all of this on the plus subscription, but you will need a $100 per month pro subscription to access their new finance tool. Using that tool allows you to connect all your financial accounts to that tool, just like you do in Tiller. However, you only get 2 years of P&L data, and anything else you need to verbally mention like your house, car, etc. Although I have not been in the Tiller Community for awhile, I use Tiller every single day to track my finances. Yes, I’m that crazy guy that looks at everything from an accounting perspective, has over 450 categories in Tiller now, and more. After I came on board I brought additional prior detailed data into Tiller and now I have probably over 10 years of very detailed data and some partial data going back to the 1990s. Please let me know you are interested in some on the detail and I will see if I can share some of it without showing you how much I am worth or not worth or how much I make or do not make. Anybody else have any relevant experience here? Cheers, Blake
I’ve been using AI for some coding assistance as well as other general stuff, but I haven’t tried it yet with finances. Glad to hear it’s working well for you! I pulled in a lot of extra data as well, I guess with the thought that some day it might be useful for trends or a historical look back. Having an AI agent that knows it all so I can ask “When did I buy Samsung Washing Machine” and get a quick answer would be nice!
Glad you got the answers you were looking for @Blake.
Personally, I don’t think that LLMs are ready for financial data prime time when it comes to foregoing the backend data modeling it requires to have a half decent chance of answering correctly without the possibility of hallucinating.
There’s also token costs. And token latency. Not to mention that I don’t see ChatGPT or its parent company OpenAI as a benevolent fiduciary or financial adviser with great knowledge- it’s great at guessing the next token but sucks at deterministic math. Deterministic math is needed for accuracy in outcomes regarding your finances.
I only have the free version of ChatGPT. It did help me out with some Google Sheets functions I’m not familiar with, but I absolutely got beyond what it was capable of. (I’m a senior software engineer, I’m used to doing some pretty abstract work.)
I don’t have access to the good AI tools at work, mostly use MS Co-pilot to apply syntax in languages I don’t use regularly and review tricky bits. I do really get frustrated with it sometimes deleting part of the prior work when my context gets larger than what that model is throttled to. This problem might go away with paid subscription, but you also might burn fewer tokens if you use an abbreviated copy of your excel sheet for making changes. Copying the changed areas from the abbreviated copy back to your full copy will also reduce the chances that changes are being made where you didn’t notice them happening.
Good lord! ![]()
yeah second on that are all those helpful for your? or are the bulk of them deprecated? 'cause i had to group the categories after like 50 because having so many just mean it’s difficult to look at a macro level of transactions
I’m not sure I have 450 regular expenses, never mind a separate category for each.
I’ve not settled down entirely on my Tiller categories, but when I am doing everything manually I have very few categories. Fixed, His, Hers, Ours.
| Category | Group |
|---|---|
| Air For Tires | Automobile |
| Andrews Autobody | Automobile |
| Car Wash | Automobile |
| Clay’s Auto Service | Automobile |
| DEQ | Automobile |
| Dick’s Auto | Automobile |
| DMV | Automobile |
| Gas - Chevron | Automobile |
| Gas - Fred Meyer | Automobile |
| Gas - Other | Automobile |
| Gas - Shell | Automobile |
| Jiffy Lube | Automobile |
| Les Schwab | Automobile |
| Maddies | Automobile |
| Phoenix Autoworks | Automobile |
| Subaru Maintenance | Automobile |
| Tanasbourne West Chevron | Automobile |
| Tire Factory | Automobile |
| Trimet MAX | Automobile |
| Valvoline Oil Change | Automobile |
I think you get the picture now. I only use about five of the 20 above. I always said go detailed versus broad on the categories because if you do not then you need to do it all over again if you change your mind in the future. Cheers.
That makes sense Blake since I couldn’t wrap my head around the sheer number of categories you have. If you are using only about 20% (extrapolating 5 out of 20 sample size) of them then you have approximately 110 categories in use. According to Ashleigh’s presentation the other day on categorization Tiller supports up to 200 categories and 20 groups. So you are well within that figure, least on the categories front. Personally I have 30. I think half or so are the default categories and groups Tiller starts with. Although it’s unclear whether the limit of 200 categories and 20 groups is active categories and groups. Still maybe if you aren’t crashing the browser cache or coming up against token limits or crashes then maybe it doesn’t matter how many active categories you have. ![]()
Wow, you get detailed with categories; you list 4 just for gas. Is there a reason you are so detailed rather than just using the broader gas as a category? I only have 3 for auto: insurance, fuel, repairs/maintenance. I don’t find all the details helpful; maybe there is something different to consider with how you do it.
I could see this as a workaround for the way descriptions often vary, you could have a pivot table off of the category to pull a consistent description of the transaction. I still would combine gas stations though, and I’m using a regex to grab a pivot-able key myself.
The beauty of excel is that is very customizable, but do you do that because your descriptions are unreliable. I think that’s generally done is use autocat to get the description to where your current category is and if you need the full details it’s in the full description column
Is the Tiller MCP available to be added to ChatGPT as a custom plugin via custom MCP url? Does anyone have that MCP server url by chance?
Thank you! I got it successfully connected to ChatGPT
For others to note, I had to use ChatGPT in the browser for the OAuth to work when setting up the custom MCP plugin (I got an OAuth error when I first tried in the ChatGPT desktop app)
Really interesting discussion! I’ve been exploring similar ideas around making Tiller financial data easier to understand and analyze.
I actually built an app called Dollar Insights to help with that. It’s been an interesting project, and I’m always curious to hear how others are approaching these challenges.
Happy to share more if anyone is interested. I’d also welcome your thoughts!
Best,
Mic
@m.svoboda09 Michael, do share information about the app you built. I am interested and am sure others will be too.
Thanks for asking, Clint. It’s called Dollar Insights. I built it because I wanted to see my Tiller data visually without building pivot tables. It connects to your Tiller sheet read-only and draws dashboards for cash flow, spending, budgets, net worth and retirement. Everything runs in your browser, so your data never reaches my servers or any AI service, and the math is plain, repeatable calculation, not an AI model. That was partly in response to David’s point above.
Full disclosure: it’s a paid app ($8.99/mo, not affiliated with Tiller). There’s a free demo with sample data, no sign-in needed: dollarinsights.net
I’d really value feedback from this group. Which view do you wish your sheet gave you that it doesn’t today?
Best,
Michael
Hi @m.svoboda09 Michael,
I looked at your demo and was very impressed. You have provided a great visual and drill down view into the Tiller data. I really like that. I have my Tiller template in Excel so can’t try out the demo with my own data. I like how you provide some key reasons for the view of data and what it means.
Speaking of @freshman.david he has a really insightful list of key metrics along with the meaning and insight. The list can be found here https://freshman-finance.com/the-cfo-numbers
His solution is really nice too and provides for large volumes of data and reporting.
Anyway, I’m just starting to figure out what I need in a visual dashboard. I do like what you and David have created.