Case Study: We Built an AI Assistant to Automate Expense Reporting

For most fast-growing businesses, expense reporting is a universal headache. It's a chaotic mess of paper receipts, forgotten purchases, and a painful month-end scramble to get everything into a spreadsheet.

A client of ours—a thriving marketing agency—was facing this exact issue. Their creative team was brilliant at their jobs but terrible at tracking expenses. The founder was losing hours every month chasing down receipts and manually entering them into their books, taking her away from her real job: running the business.

We saw an opportunity not just to streamline a process, but to build them their own "AI Assistant" that could handle the entire workflow, turning a 10-minute task into a 30-second one.

The Problem: The "Shoebox of Receipts"

The agency's old process was entirely manual:

• An employee would make a purchase (e.g., software, client lunch).
• They would (hopefully) save the receipt.
• At the end of the month, they'd submit a pile of receipts to the founder.
• The founder would manually read each receipt, categorize it, and type the vendor, date, and amount into a spreadsheet.
• Finally, this data would be sent to their bookkeeper for reconciliation. This process was slow, prone to errors, and universally hated.

The Solution: An AI-Powered Expense Workflow

We designed a system that feels like magic to the end-user but is powered by a robust, intelligent automation. The only thing an employee has to do is take a picture of a receipt.

Here’s the architecture of the "AI Assistant" we built:

• Step 1: The Drop Point (Google Drive)
  We created a shared Google Drive folder. An employee simply snaps a photo of a receipt and uploads it from their phone. That's their only task.
• Step 2: The Brains (Make.com)
  Our workflow engine, Make.com, instantly detects the new file. It's the central hub that connects all the other services in the workflow.
• Step 3: The AI Magic (OpenAI GPT-4 Vision)
  Make.com sends the receipt image to OpenAI's powerful GPT-4 Vision model. We give the AI a simple prompt: "Extract the vendor name, the transaction date, and the total amount from this image." The AI "reads" the image and returns clean, structured data in seconds.
• Step 4: The Human Checkpoint (Slack)
  The extracted data is posted in a private #expenses Slack channel with the receipt image and two buttons: "Approve" and "Reject." A manager can verify the data is correct with a single click.
• Step 5: The System of Record (Airtable)
  Once approved, Make.com instantly creates a new, clean record in an Airtable database. The bookkeeper can now access this organized, pre-approved data at any time for easy import into their accounting software.

The Results: From Hours to Seconds

This AI-powered system completely revolutionized the agency's expense management process.

• 95% Reduction in Processing Time: A task that took 5-10 minutes of manual work now takes less than 30 seconds.
• Elimination of Manual Data Entry: The founder no longer spends any time typing out receipts.
• Real-Time Expense Tracking: The agency has an up-to-the-minute view of expenses as they happen, not just at the end of the month.
• Happier Team: Employees love the simple "snap and upload" process, leading to better compliance and faster reimbursements.