My partner and I share a credit card. We also share a bank account that autopays it every month.
Most months, this works fine. Everything is roughly 50/50, autopay handles it, nobody thinks about it.
Then we took four trips in five months.
New Orleans for Christmas. Palm Beach over Thanksgiving. A birthday trip to Savannah with my mom and sister. A road trip to Durham with our curling club.
Each trip had different people, different split arrangements, and different “who covered what” dynamics.
By March, I was staring at five PDF statements, a shared bank account running low, and zero confidence about who owed what to whom.
The kind of mess spreadsheets can’t clearly communicate
130+ transactions across five months. Dining, travel, subscriptions, home expenses, and four separate trips mixed together.
Every trip had a different split.
New Orleans was just us - 50/50.
Savannah’s Airbnb was split with my sister, but the hotel was different because my mom had her own room.
Durham had group expenses friends already paid back for, plus our own charges mixed in.
On top of that, our shared bank account pays a monthly installment for new windows. I’d put in extra cash twice to keep it from overdrawing.
I needed one answer: how much does each person put in before next month’s payment hits?
I tried doing this in a spreadsheet. Twice.
Both times I got tangled in my own logic halfway through. The split rules were too inconsistent to hold in my head while also entering data.
The real work was the conversation
I dropped all five PDFs into a conversation with JFDIBot and said “help me figure out what everything is tied to.”
First, it read every transaction and sorted them chronologically into one ledger. Not by statement period - by date. That alone was useful. Charges from the same trip were scattered across two different billing cycles.
Then it became a back-and-forth.
JFDIBot would categorize something. I’d correct it. “That’s not a local dinner, that’s from the Palm Beach trip.” “The Savannah Airbnb is split two ways with my sister, not three.”
Each correction updated the model it was working from.
Here’s the part worth paying attention to: the bank account reconciliation.
I explained the setup - autopay pulls from it, windows come out monthly, I’d topped it up a couple times.
Then I asked one question: “How much does everyone need to put in before next month’s payment?”
JFDIBot mapped all the inflows and outflows, factored in the different splits, added a buffer, and gave me a single number for each person.
I’d already spent an hour trying to do this in a spreadsheet. JFDIBot had it sorted in about 20 minutes.
Why a thinking partner beats a formula
A spreadsheet would have been fine if every expense split the same way.
The problem was the exceptions.
Four trips with four different group compositions and four different splitting arrangements, layered on top of recurring shared expenses and a bank account that multiple people contribute to.
What JFDIBot gave me was a thinking partner that could hold all the context - every transaction, every trip, every splitting rule - while I just answered questions.
“Was Palm Beach just you two?”
“Is the second Airbnb charge also Savannah?”
“Does your mom split the Airbnb or just the hotel?”
The output was a Google Sheet with a full chronological ledger and a summary tab showing exactly who pays what. But the real value was the conversation that produced it.
When your problem is a conversation, not a spreadsheet
If you have a problem where the data is messy but knowable (PDFs, statements, receipts), the rules are inconsistent (different splits for different contexts), and you keep getting lost in your own logic when you try to do it manually -
That’s a conversation.
Feed in the raw data. Explain the rules as you go. Let the AI hold the complexity while you make the decisions.
The logic was never hard. I just couldn’t hold it all at once.