Industry:Retail

Region:North America

Client:Major Retailer and Accelirate

AI-enhanced Robots Process 93% of Invoices Automatically for a Major Members-only Retailer

Major Retailer and Accelirate customer story hero image

7,000

invoices processed monthly

30

seconds now spent on processing an invoice that used to take 3-5 minutes

160+

hours saved monthly

95%

confidence score for invoice processing with ML model

RPA consultant Accelirate implements UiPath Document Understanding to help automate Invoice Processing for a major retailer and helps the employees focus on more important tasks.

The accounts payable (AP) process plays a vital role in any company’s finances, so efficiency is critical. Many AP teams in large organizations, however, know the reality of trying to manage hundreds or even thousands of invoices every month. 

For a large U.S. members-only wholesale retailer, invoice processing had become a real burden. A particular challenge was the large volume of invoices it received from gasoline and freight vendors, which arrived at a rate of 200 to 500 per day on an average. On peak days, that number could reach 700.The sheer quantity was a challenge. But making it worse were the manual, labor-intensive steps required to handle each invoice. The AP team’s staff had to open individual emails containing attached invoices key into the company’s internal system to find a supplier’s ID, and manually extract the data from the invoice. From there, a staff member would have to enter the information into the internal accounting system so the vendor could get paid.

This had to be repeated for each invoice—a process that took anywhere from three to five minutes per invoice. This was quite a cumbersome process and unsustainable without automation over the long term. The company decided to seek help from Accelirate, an automation solutions company with experience in a wide range of industries, including retail, manufacturing, banking and financial services, healthcare, and energy.

“The customer understood that this part of the business was a prime candidate for automation,” says Ahmed Zaidi, Co-founder and Chief Automation Officer at Accelirate. “It was a vital financial process that, if automated, could eliminate tedious manual tasks and free up staff time for more value-added work.”

Building a solution with UiPath Document Understanding

Accelirate, a leading RPA vendor and UiPath partner, created a solution combining RPA with optical character recognition (OCR) and UiPath Document Understanding, which helps robots learn how to extract, interpret, and process data from a wide range of document types. The solution would include pre-trained machine learning models for invoices. Accelirate retrained these via UiPath AI Center to capture additional bill-of-lading information that is specific to the trucking services used by gas and freight vendors. They worked with the UiPath Document Understanding team to help train the ML model to identify and retrieve the bill-of-lading fields.

The Accelirate solution includes six steps:

1. A bot reads emails containing invoices. If the email contains one or more attached files that are not invoices, the bot flags the email, which is then forwarded to an AP department staff member for manual evaluation. If the email only contains invoice-related attachments, it saves the files to a file share where they are queued up for Document Understanding.

2. The bot splits the invoice. The bot digitizes the invoice using OCR to find and determine the locations of the header and footer for each page of the invoice. The bot then splits the invoice into its individual pages, with each page displayed in correct order.

3. The bot extracts data using Document Understanding. The bot uses OCR and machine learning to extract information such as invoice date, number, amount, and due date. Accelirate used machine learning and AI Center to train the out-of-the-box model for invoices on the numerous formats of invoices submitted by different vendors. Similarly, the bots learned to search for bill-of-lading (BoL) information. The bot automatically checks to see if the associated BoL information is on file; if not, it notifies the AP team for further action. 

4. The bot moves the data to a queue. Depending on the results of the invoice extraction, the data is moved to one of two queues. If the confidence level for the extracted data is above the 95% threshold, the invoice is sent automatically to a reconciliation queue for payment processing. If the confidence score on the information is below the threshold, it is sent to a queue where an AP team member uses the UiPath Validation Station to inspect the invoice. Confirmed information from Validation Station is uploaded to AI Center to help continuously retrain models, making them more efficient as more data is fed to it. 

5. The bot repeats steps 1-4 until every invoice has been reconciled or moved to the reconciliation queue.

6. The bot generates a report. After reconciliation, the bot generates a report detailing count of items in each queue, how long the bot ran, and how many items were processed. All reports are archived for record keeping. 

The initial model training was done on 174 pages of documents from 30 vendors. Extra training with more than 800 pages of documents from an additional 25 vendors followed that. The training was designed to address virtually every type of invoice that might be submitted to the company.

A high level of confidence for the automation—and time saved for the staff

Zaidi says using the UiPath technologies helped Accelirate and the customer rapidly achieve a high degree of confidence that the bots were accurately processing the invoices.

Once the customer started using it in production, 93% of the invoices were going straight through to the reconciliation queue without needing any manual inspection,” he says. “Of those invoices, there was a 95% confidence score for the extracted data. The high degree of confidence in the quality of data obtained with UiPath Document Understanding created a lot of confidence for the solution.

Ahmed Zaidi • Co-founder and Chief Automation Officer, Accelirate

Moreover, the machine learning done with UiPath Document Understanding includes a feedback loop via Validation Station that makes the model more accurate over time. This helps ensure that the accuracy of data will continually improve over time.  

Zaidi notes that once the customer’s AP team understood the efficiency gains delivered by the solution, they enthusiastically embraced it as a major productivity enhancement.

Now, instead of someone having to spend up to five minutes processing a single invoice, the AI-enhanced robot handles that same document automatically in about 30 seconds.

Ahmed Zaidi • Co-founder and Chief Automation Officer, Accelirate

He adds that only a small fraction of the previous number of invoices need manual evaluations at Validation Station.  

We estimate that using the newest AI solution within UiPath Document Understanding,” Zaidi says, “they can turn at least 20% of their AP department to other, more valuable activities.

Ahmed Zaidi • Co-founder and Chief Automation Officer, Accelirate

Related case studies

Contact Center

Case Study

Finastra
“Agents love it” – the inside story of an automation-powered contact center
Read the case study
Dexcom's journey

Case Study

Dexcom
Dexcom's journey to efficiency with AI-powered automation, unlocking 200,000 hours
Read the case study
How citizen developers power up Wesco's automation program

Case Study

Wesco International
How citizen developers power up Wesco's automation program
Read the case study

Ready for your own case study?

Speak to our team of knowledgeable experts and learn how you can benefit from RPA.