Komatsu Australia accelerates hyper automation with Power Automate – from licensing to production in 4 weeks
Komatsu is an industry-leading manufacturer and supplier of equipment, technologies and services for the construction, forklift, mining, industrial and forestry markets. For a century, Komatsu equipment and services have been used by companies worldwide to develop modern infrastructure, extract fundamental minerals, maintain forests, and create technology and consumer products. The company’s global service and distributor networks support customer operations, tapping into the power of data and technology to enhance safety and productivity while optimizing performance.
In this article we look at how a team at Komatsu Australia built a Robotic Process Automation (RPA) solution with Power Automate and AI Builder to automate their manual and repetitive processes for fixing invoices. They went from purchasing licenses to an RPA solution in production within a short span of four weeks.
Like any business, Komatsu Australia sends and receives a high volume of invoices from suppliers each year. For the parts department, a small team that manages over 52,000 invoices annually, fixing invoices was a manual and tedious process that put immense pressure given the size of their team. A solution architect, Eric Cheng, looked towards Power Automate and AI builder to automate this process. He used one supplier as a pioneer RPA use case. The RPA solution saved them 300 hours a year in invoicing for this supplier alone, and they’re now expanding it to cover all suppliers.
In this story:
- Meet Eric
- Business scenario
- Before Power Automate
- Power Automate solution
- Summary of impact
- Solution architecture
- Expanding ROI through the Automation CoE Starter Kit
- Creating value together – citizen development program
- Examples of Power Apps built by a citizen developer
- Looking ahead
Meet Eric
Eric Cheng is a Digital Solutions Architect in the Business Technology and Systems team at Komatsu Australia with over 13 years of experience in developing and architecting solutions using Microsoft technologies. He has been one of the early adopters of Power Platform at Komatsu Australia and his team has built several solutions to date.
Drawing on his experience and knowledge of Power Platform, Eric looked towards Power Automate and AI Builder to automate the invoicing fixing process after having success with another similar use case. The user experience of Power Automate for desktop combined with the learning content from Microsoft Learn made it intuitive for Eric to quickly upskill and start building RPA solutions.
As an initial RPA solution, Eric automated a manual and tedious process where a specific customer required invoices to be submitted through their online third-party system. It required branch managers to spend considerable time each week on this manual process. The solution built using Power Automate for desktop saved hundreds of hours a year. The success of this initial solution led to Eric taking on the task to automate another process that involved a much larger volume of invoices.
Business scenario
The parts department receives a high volume of invoices at all times of the day with over 52,000 invoices annually across 259 Original Equipment Manufacturers (OEM) and suppliers. A small three-person invoice processing team is responsible for fixing these invoices, a manual process which requires each invoice to be read, matched against the purchase order, and processed into an IBM AS400 mainframe system used for the management of parts orders between Komatsu Australia’s suppliers and distribution centers.
The team works under tremendous pressure daily to ensure the constant stream of invoices are processed as delays could result in purchase order variances or impact downstream processes such as receiving goods or picking of parts in their distribution centers. A pilot for RPA automation was undertaken to help relieve the invoice processing team and reduce delays and impacts.
Before Power Automate
One of the local suppliers in their system was selected as the pilot candidate to automate the invoice fixing process. Prior to automation, the invoicing team would receive over 1,100 PDF invoices for the supplier in an Outlook shared mailbox.
The manual process was as follows:
- The invoice processing team would regularly monitor the mailbox for new invoices and download them as they arrive for processing.
- Each invoice is reviewed, and information is extracted from the header and line items into a new CSV file using Excel.
- A separate Purchase Order report is downloaded to validate the data and a three-way match is performed between the CSV file data and the report data using a series of Excel macros and VLOOKUP functions.
- The CSV file is then uploaded to their IBM AS400 mainframe via a web portal known as eParts. Each invoice is then processed in the mainframe where any additional charges such as tax, packaging and freight are added before it is sent to their warehouse management system for goods receipting which involves matching the received goods against the purchase orders and generating the packing slips for orders to be picked.
The key challenges of this manual repetitive invoice fixing process were:
- Invoices received at all times of the day from the supplier.
- Small processing team of three working under pressure, was difficult to keep up with the invoice processing when a person is away.
- Invoice processing goes through annual peak periods such as end of financial year which further increased the daily volume of invoices received.
- Delays impacted the warehouse management system and downstream processes.
- Prone to human errors in reading invoices and entering the data into system.
Power Automate solution – Invoice fixing automation
Prior to using the Power Platform, Komatsu Australia had already successfully automated over 18 processes using another vendor’s RPA platform. However, the costs of the vendor RPA platform were significant which presented a challenge for automation scalability. Komatsu Australia turned to Power Automate as it was more cost effective and the automation capabilities were richer, especially when combined with the rest of Power Platform which included AI Builder document processing.
To reduce the pressure on the invoice processing team, unattended automation with Power Automate was utilized. Unattended enables a process to be fully automated in the background with Power Automate desktop and cloud flows, without needing human interaction or decision making. This was perfect as it helped eliminate manually monitoring the Outlook shared mailbox for invoices, validating invoices against purchase orders, and entering invoices into their mainframe system to initiate downstream processes.
Within four weeks, Eric developed and went live with the unattended RPA solution. Application Lifecycle Management (ALM) best practices were also followed as several environments were set up for the development, testing, and enterprise deployment of the solution.
The solution utilizes Power Automate and AI Builder to automate the processing of invoices end-to-end with the mainframe system, as follows:
Automation of the invoicing fixing processing with Power Automate.
Power Automate cloud flows
- The cloud flow runs on a schedule and retrieves emails from an Outlook shared mailbox.
- When an email with a matching condition is retrieved, the cloud flow will download the attached invoices as PDF files and send it to the AI Builder document processing model.
- The results returned from the AI Builder document processing model is saved into a SQL database.
- The cloud flow then verifies all invoice data against the purchase order details in an Azure Synapse data lake including details such as supplier number, purchase order line-item numbers, part numbers, etc.
- The cloud flow then triggers a desktop flow to create a CSV file of the invoice data which is loaded into the IBM AS400 mainframe system.
- The cloud flow then triggers a second desktop flow which processes the invoices into the IBM AS400 mainframe system. This includes adding any additional charges such as taxes, freight, packaging, etc.
- Lastly, processed invoices are archived into a separate table in the SQL database for audit history.
AI Builder document processing
- A custom document processing AI model was built, trained, and published.
- A sample size of 10 invoices were used to train the AI model with a 99% accuracy score.
A custom AI builder model was built to analyze and extract data from PDF invoices received and a Power Automate cloud flow saves the extracted data into SQL database which then executes corresponding desktop flows.
Power Automate desktop flow (1)
- The desktop flow queries a SQL database and generates a CSV file in a predefined schema.
- The desktop flow then launches an Edge browser, navigates to and logs into the eParts web portal.
- The desktop flow then browses to the CSV file and uploads the file into the eParts web portal.
Top screenshot – CSV file is uploaded into the eParts web portal, bottom screenshot – Power Automate desktop flow steps for the CSV file upload.
Power Automate desktop flow (2)
- The desktop flow will open a terminal window and logs into our parts mainframe system.
- The desktop flow will traverse across various screens and enter data as part of the invoice fixing process.
- The desktop flow will generate and email an exception report to relevant stakeholders.
Top screenshot – The terminal window of the mainframe system for processing the invoice data, bottom screenshot – Power Automate desktop flow steps for the data entry on the mainframe system.
Summary of impact
By automating the invoice fixing process, Komatsu Australia realized several key benefits:
- Automated 1,200 invoices across 1,190 SKUs and is projected to provide 300 hours of efficiency gain per year for just the one supplier alone.
- Reduced errors by eliminating human data entry.
- Enabled 24/7 invoice processing, including weekends for overseas suppliers.
- Freed up capacity to the processing team of 3, allowing them to focus on monitoring the process and managing exceptions.
The success of the automation for the pioneer local supplier provides a blueprint for the automation of the remaining suppliers of the process. By applying it to their top three OEMs, the savings will be 300 hours per month rather than annually.
“With this one single supplier, we’ve automated over 1,000 invoices annually and realized an efficiency gain of 300 hours per year.”
Solution architecture
The following is a representation of the high-level solution architecture that supports the automation of the invoice fixing process for their parts department. Power Automate for desktop and cloud flows orchestrate the automation end-to-end, with AI Builder performing the extract of the invoice data stored into the SQL database. All invoice data is then verified against the purchase order data in an Azure Synapse data lake which is then generated as a CSV to upload into the eParts portal followed by the processing in the IBM AS400 mainframe system used for the management of parts orders.
High level solution architecture that combines Power Automate and several services in Power Platform and Azure for invoice fixing in the Parts department. .
Expanding ROI through Automation CoE Starter Kit
Prior to the development of the invoice fixing RPA solution, the Business Technology and Systems team sought out to lay the foundation for automation enablement. It was important for Komatsu Australia to have a foundation to build upon to scale their automation with Power Automate for desktop and cloud flows. With the proven success and impact of automation in the invoice processing area, they’re expanding their usage of Power Automate across the organization and exploring reducing the dependency on the existing RPA platform.
Eric and his team have rolled out the Automation Centre of Excellence (CoE) starter kit (preview) to manage and have visibility across the automation development lifecycle. The Automation CoE starter kit has enabled the following:
- Built-in predictive scoring and return on investment analytics enabled the team to quantitatively prioritize the backlog and visualize the health and performance of the automation using a Power BI report.
- The reporting capabilities have helped the team build business use cases for senior management to understand the value delivered to their organization across the automation pipeline.
- Ability to track metrics and monitor the automation in real-time has helped ensure the automation is operating efficiently and without issues.
Prioritized backlog of automation solutions and ROI real-time insights with the Automation CoE Starter Kit.
“Being able to visualize ROI in real time is beneficial as it provides a quantitative justification for each automation. This ability allows senior management to see the value delivered to the business across the automation pipeline.”
Creating value together – citizen development program
As part of a “Creating Value Together” initiative, Eric and his team have started to build their own community of citizen developers at Komatsu Australia. Their first cohort includes eight employees from different areas of the company such as the distribution centre, sales, and operations. These citizen developers have no prior experience developing software solutions. The diversity in business aptitude will help accelerate the development of solutions in alignment with their business processes. They’ve setup a Yammer channel and Teams group where citizen developers are encouraged to interact with each other and learn through discussions. These interactions have helped continue to increase the learning momentum and grow interest in building low-code solutions. All citizen developers are also encouraged to complete certifications to increase their expertise as they start to scale with Power Platform.
Collaboration amongst the citizen developers using Microsoft Teams and Yammer.
Examples of Power Apps built by a citizen developer
Matt Burton, a senior controller in the New Construction Experience team, is in the process of building two apps as part of the Creating Value Together program.
Scanning and identifying equipment attachments
 |
As an industry-leading manufacturer and supplier of equipment, Komatsu Australia receives numerous equipment parts from vendors that need to be delivered across their branches for customer sales. One type of equipment part is Komatsu Genuine Attachments (KGA), such as a quick hitch, bucket, or wear package for excavators and loader machines. When these KGAs are received at the distribution center, the team at the center are required to code “attachment tags” and link these tags to the KGAs for stock management.
Matt is developing a Power Apps canvas app using the built-in NFC tag reader capabilities to enable employees to identify the KGAs at their branches when the attachment tags are scanned. Additionally, by scanning the KGAs, it quickly gives visibility of the stock on hand in comparison with the data in their stock management system. |
Changing the destination for new equipment
 |
Business Managers needed an effective way to change the inbound port of entry for new equipment before it leaves an OEM site. Matt is developing a mobile app with Power Apps to simply this process and provide Business Managers the flexibility to update destinations from their phones. Upon submission, cloud flows are triggered to notify the Logistics team and request approval. Based on the submission type, two different actions are processed.
For every change in destination that occurs, each one is logged into SharePoint list to maintain a history of changes. |
Looking ahead
Scaling the community across the company
The long-term goal for Komatsu Australia is to roll out their “Creating Value Together” citizen development program with the aim of organically growing their own no-code/low-code developer community. The pilot citizen developers will become champions within their teams to encourage others as the program is rolled out across the company. Eric and his team are committed to having the right mechanisms in place to monitor and support the citizen developers to ensure the growth of their citizen developer program.
“A large part of our citizen development program is to create and foster a community of practice because in the future state, it isn’t IT who is necessarily going to be supporting these solutions, it’s the citizen developers themselves and the business.”
Aligning Power Platform with ALM and monitoring best practices
To support their long-term RPA growth with Power Automate for desktop, Eric and his team are in the process of setting up their solutions with Azure DevOps for Application Lifecycle Management (ALM). ALM is already in place with many of their existing code-first applications and the same best practice is to be applied with their Power Automate RPA solutions. This is to ensure there are approvals and that checkpoints are performed prior to going live with each RPA solution. Another tool they are using is Azure App Insights for monitoring their enterprise application and there are plans on incorporating this with their Power Automate RPA solutions. The long-term goal is to use the Microsoft ecosystem for solution development, monitoring, and governance for digital enablement at Komatsu Australia.
“We can see the capabilities increasing, the maturity of the Power Platform rapidly accelerating – it’s delivering a lot of great value for our organization in terms of efficiencies.”
Related links
- Microsoft Power Platform Automation CoE starter kit
- Komatsu Industries optimizes machinery up-time and predicts remaining useful life of components using Azure Machine Learning
- Komatsu Group’s ‘workplace safety’ approach offers insights, business improvements using Microsoft Teams
- Mining equipment manufacturer helps safeguard vital intellectual property with Microsoft Information Protection
- Smart factory base KOM-MICS visualizes and improves wider production sites with Azure
- Komatsu achieves breakthrough performance gains and savings with Azure SQL Database Managed Instance
