Before introducing a new solution to our clients, we test it on our systems. In this example, we've been experimenting with AI in our accounting and budgeting departments recently.

The Problem

Our aim? To reduce manual labor, automate tasks where possible, and see if AI lives up to the hype.

Here's what we found.

Our Digital Accountant

Meet "Avery," our Digital Accountant -  a virtual persona we created to take charge of our accounting operations. Avery is programmed to manage various tasks and has an email address where we direct all invoice-related communications.

Avery understands English, Ukrainian, and Polish, the three languages we use in our accounting documents. Plus, she's available 24/7, does not procrastinate, and starts working immediately when a task arises.

Technology Stack 

For this project, we used a combination of Airtable, Microsoft PowerAutomate, Google Vision, and OpenAI.

  1. Airtable served as our primary system. It's a flexible cloud-based database platform that allows us to manage our accounting data effectively on our size of operations;
  2. Microsoft PowerAutomate helped us automate workflows, making data collection, organization, and management more efficient;
  3. We used Google Vision for OCR (Optical Character Recognition) to convert documents into editable text;
  4. OpenAI was crucial for intelligent text processing. After Google Vision turned our invoices into text, OpenAI extracted relevant data like invoice numbers, dates, and amounts.

Handling Incoming Invoices 

Previously, processing incoming invoices was manual and time-consuming. Our team had to read countless emails, convert attachments into readable formats, and input the data manually into our accounting system or Internet banking. It was particularly challenging when invoices came in as scanned documents or photographs.

But not anymore!

Now, our email client filters and automatically send regular invoices to Avery. For one-off invoices, we manually forward them to Avery.

Once Avery gets the invoice, she processes the emails, converts documents into editable text, and extracts needed data like invoice numbers, amounts, and dates with the help of OpenAI. This data is then entered directly into Airtable for review and approval.

Avery also categorizes the invoices according to the budget accounts. If she detects a potential budget overrun, she alerts the management team.

Handling Outgoing Invoices 

Avery also plays a crucial role in transforming our workflow in outgoing invoices. However, we use a methodical, rules-based approach tailored to each client rather than relying on AI for this job.

For every client, we create a custom form detailing their billing principles.

Then Avery uses the time reports and other relevant information to create a draft invoice when billing a client. Before any document is sent to the client, we check it for errors. Once the invoice is approved, Avery drafts an email with the invoice attached.

When payment is received, Avery generates an official invoice through an API call to our external tax reporting system. She also saves all invoices to a cloud drive for easy access at the month-end when we send this data to the accounting company, which takes care of our tax reports.

Processing Bank Statements 

We had to use a different approach for bank statements due to security rules. Like many other traditional European banks, our bank does not allow easy access via API to the data. However, we found an efficient and secure method, yet more manual than everything else in our system.

We make CSV files from our Internet banking and send them to Avery.

Avery uses AI to sort out different types of payments based on their descriptions and other details. Then she classifies transactions without invoices, like bank charges, money moved between our accounts and tax payments. Also, she does a good job matching the payments and invoices. 

Error handling

Does Avery ever make mistakes? Sure, she does. After all, no system is 100% foolproof. But the beauty of Avery's AI is that she learns from each mistake, improving over time. Moreover, her error rate is significantly lower than manual processing, and it's much easier for our team to verify her work than to handle the entire process ourselves.

Final Thoughts

Even as a small company, the impact of Avery on our workflow has been significant. She has taken over the tasks we didn't enjoy doing and has become a reliable, tireless part of our team. Larger companies could see even more significant benefits from using Artificial Intelligence.

Avery is always ready to work and requires no breaks or holidays. Unlike humans, she doesn't get tired or irritable, maintains a consistent performance level at all times, and improves constantly.

We believe this is a glimpse into the future of work. AI won't replace all our jobs, but it's likely to reshape how we work. Those who adapt to working with AI will be better positioned in this changing landscape.

So, are you considering how AI might assist your operations? We're here to help. So, get in touch with us.