The Problem
A partner at an accounting firm described their most painful recurring workflow to me. When his clients sent him PDF bank or credit card statements, before the firm could do anything useful with them, every transaction on every page had to exist as a row in an Excel spreadsheet. Clients who would request catch-up work could be coming with statements from multiple different years across several accounts, arriving all at once and usually late.
When the firm encountered these scenarios, they would manually enter the transactions from the PDF statements into Excel. This literally meant having the PDF statement open on one half of their screen, the Excel spreadsheet on the other half, and manually typing out or copy-pasting the date, description, debit/credit amount, and the balance for each transaction in the PDF statement. This would take hours, sometimes days, and in extreme cases weeks.
It wasn't only the amount of time that it took which was an issue. Since the work was repetitive and mundane, mistakes in the data entry were inevitable and in accounting, even one wrong number is unacceptable. If mistakes were made and the team noticed that the numbers weren't adding up properly, they would have to analyze every line of data entry they had done to identify where the mistake is. This would cause an even larger time sink and make them fall behind on other responsibilities.
While he was explaining this workflow to me, he said:
“I hate this process so much it makes me want to pull my hair out.”
My Role
I ran the discovery conversation and identified what was the most painful workflow the firm had that needed to be solved to save a significant amount of both time and money.
I designed the end-to-end workflow, then built the automation entirely myself. This meant choosing the technology stack, which included Make.com, Azure AI Document Intelligence, Claude API, and Power Automate.
It also included building features like error handling, duplicate detection, automatically sorting multi-year statements into separate Excel files, handling statements from different banks, giving the user the ability to reprocess a statement in case something went wrong, and fixing bugs that real statements surfaced.
Once the product was shipped and being used by the firm, I tested the automation with the partner continuously and kept refining it to work with the statements his firm actually processes.
I tracked the outcome by keeping count of the number of statements and transactions they processed.
Approach
I knew that the people working in the firm who were responsible for performing this workflow were not particularly tech-savvy. Therefore, I wanted to create a solution that was as easy to use as possible with the least amount of effort required.
When I had met with the partner at the firm and he was explaining his workflow to me, he showed me that they store all of the PDF statements that they receive from clients and the Excel files that they create in a shared OneDrive account that all employees have access to.
They also had a specific structure for the way they organized their folders and followed naming conventions as well.
When I was brainstorming possible solutions, I thought what if instead of making a web application of some sort, which they would have to create an account for and sign into, I create the solution directly inside of OneDrive so they wouldn't have to leave the platform they're already comfortable using.
I asked myself "what would be the most ideal experience if anything was possible?" The answer I came up with was the user dragging and dropping the PDF statements into the folders in OneDrive that they already store them in. The automation could detect this, start processing the statements automatically in the background, and create an Excel file with all of the transactions inside of it. The Excel files would be created in the folders they usually are in.
The user would get an email notification that their Excel file is ready to be reviewed and with that, they'd be able to bypass any sort of manual entry whatsoever and reduce the task to a simple drag and drop, which was the easiest possible experience I could think of.


Key Design Decisions
No New Software
This was one of the first ideas that popped into my head when thinking about the solution for the firm's problem. Since I wanted to keep the user experience as simple as possible, I figured what could be simpler than keeping the experience in the tool they use every day and are already comfortable with? I wanted the experience to simply be dragging and dropping files into folders in OneDrive.
I also wanted the automation to detect when a file was uploaded and automatically start processing that file. This way, the user wouldn't have to think about anything and it would make training other people to use the automation an easy experience as well.
Splitting Statements at the Fiscal Year-End
One of the requests the partner at the accounting firm made was for the automation to be able to automatically split transactions that span across two different year ends into different Excel files. For example, if one credit card statement had transactions in December 2025 and January 2026, then two separate Excel files would need to be created, one for 2025 transactions and the other for 2026 transactions. They would also have to be organized into different folders.
This request ended up significantly changing the build for the automation but it helped me follow one of the pillars of this product which was that I didn't want the user to do any sort of manual data entry whatsoever. If the automation didn't support this feature, then the transactions from 2025 and 2026 would all go into the 2025 Excel file and the accountant would have to manually cut all the 2026 transactions, create a 2026 Excel file, and paste all of the transactions in it. Across hundreds of statements, this time cost would add up.
Detailed Folder Structure
Another decision I made to make the experience as simple as possible was tailoring the automation to work with the folder structure they were already using.
They were very particular about the way they organized their files. For example, when Excel files were created for chequing and savings account statements, they had to be separated into different folders. Same goes for credit cards. If one of their clients sent PDF statements for multiple credit cards, each credit card would need its own Excel file.

It was clear they had been using this system for a while and the entire staff was already used to it. To make sure that the transition to using the automation was as smooth as possible, I added logic that would detect whether a particular statement that was uploaded was a chequing account statement, a savings account statement, or a credit card statement. The Excel file would then be created in the correct folder depending on what type of statement it was. If there were different credit card statements that were uploaded for the same client, the automation automatically extracted the last 4 digits of the credit card and put it in the Excel file name to ensure that transactions from each credit card statement are extracted into their own respective Excel files.
This allowed the files to be organized in the same way they were before the firm started using the automation, with the only difference being that now instead of manually entering all the transactions from the PDF statements into Excel, they were just simply dragging and dropping the PDF files into the folders.
Matching the Excel Format They Already Used
Similar to how I tailored the automation to work with the firm's folder structure in OneDrive, I also tailored the Excel file to be formatted in the same way the firm already does it.
This meant Excel files for bank statement transactions have five columns which are Date, Description, Withdrawal, Deposit, and Balance and credit card Excel files have four columns which are Posting Date, Transaction Date, Description, In/Out.

This was important because it kept the experience of reviewing an Excel file the same for the firm and they didn't have to get used to any new formatting. Small decisions like these maximize efficiency, since they preserve the familiarity of the old workflow and remove any friction a new experience might cause.
Outcome
The firm has been using the automation for around 6 months now and they've processed over 1,000 PDF bank and credit card statements and over 20,000 transactions. The best part is that they've been able to process these numbers in a fraction of the time using the automation compared to them having to manually do it.
After the first month, the partner at the firm texted me the following message:
“Dude this has been perfect. Converted 400 lines in minutes and I didn’t have to pay anybody to type it out. The monthly fee to you guys already got covered.”
A few months later, the same partner told me that there was a case where a client approached them with PDF statements from the past two years and a very tight deadline to process all the transactions. The partner said that before, it would take them three to four days to process all of that and they would have to work overtime. Using the automation, they processed everything in 20 minutes and still charged the client for three days of work.
The firm went from an outdated, inefficient workflow to a modern solution that has saved them countless amounts of time and money.