As Japan's working population shrinks, the work style reform have become a critical management issue. Toshiba Digital Solutions is implementing work style reform projects at the company level. In one of these projects, in April 2018, we created the RPA Promotion Working Group and began applying RPA to various in-house operations with the theme of improving operation efficiency through the use of ICT. The core of this working group was our Reform of Management & Information Systems Division.

Generally speaking, there are two main objectives for using RPA. The first is to apply RPA to complex operations to improve productivity, reducing working hours (person hours) and overtime. The other is the elimination of human errors such as input errors, recognition errors, management errors, and the like.

Furthermore, RPA can also be used at night, on holidays, and on weekends, freeing enterprises from time constraints and increasing work frequency. For an implementation of RPA, it must first identify its operation procedures, which contributes to operation organization and standardization. Furthermore, applying RPA to routine work performed at specific times also has the potential to free staff from time restrictions and work stress.

Operations that are suited to the application of RPA have the following characteristics.

• ●
  Repeated operation according to consistent rules
• ●
  Operation that handles structured data
• ●
  Operation that spans multiple applications, such as those used in enterprise systems, information systems, and cloud services
• ●
  Operation that consists of organized operation processes and standardized operations
• ●
  Operation that is prone to human error

When deploying RPA throughout a company, it is important to assess whether you will actually be able to produce the results you desire, and what issues will be faced during this deployment.
In Phase 0, we applied RPA in a pilot test to three operations performed using enterprise systems and individual applications.

RPA for our sales processing reduced monthly workloads by an average of 70 to 80%, as well as avoiding entry errors.
Based on this result, we confirmed that the use of RPA produced significant reduction effects.

From the latter half of 2018, in Phase 1, we establish a structure for promoting RPA within the company. The RPA Promotion Working Group consists of four teams. The Management Team handles operations such as budget management, budgetary policy formulation, and reporting. The Support Team handles inquiries and problem reports related to RPA from throughout the company. The Operation Team maintains and manages RPA operation platforms and provides development skill training. The Development Team develops robots when requested by internal departments, conducts related interviews, and defines requirements. Promotion members selected from each department are responsible for identifying the needs in each department, coordinate with partners to carry out robot operation and development, and efficiently promote RPA. Of course, RPA systems coordinate with enterprise systems, information systems, and other in-house systems to perform processing, so coordination with information system departments is essential.

In parallel with the establishment of this promotion system, we also asked each department to submit recommendations of operations they wished RPA to be applied to. We received over 200 submissions for RPA candidate operations. Our Management Team lead the effort to determine which of these operations were well suited to RPA. They conducted interviews with departments and used a unique decision-making procedure to divide operations into four ranks of RPA application priority, S, A, B, and C, based on the efficacy of RPA and our current workload (Fig. 1).

Of the 200 submissions received, 13% of the operations were assigned to Rank S, 15% to the next rank, Rank A, 9% to rank B, and 46%, almost half, to Rank C. The remaining 17% were deemed to be outside of the scope of RPA at the time of evaluation. Some of these were operations that required reconsideration, including the use of methods other than RPA.

As these results show, RPA is no magic bullet; it is vital that its efficacy be closely considered and that it be deployed skillfully.

From April 2019, we began work on applying RPA to 10 Rank S operations. One of these operations was the automated input RPA for quotation reply. In this operation, PDF files in a designated folder were converted to text using OCR^*, saved in a specific file format, and then the information in the saved file was entered into a quotation reply field. We applied RPA to this process.

This work was performed frequently, almost 400 times a month, but by applying RPA, we were able to reduce the workload by an average of 80% per month (Fig. 2).

The 10 operations we applied RPA to, including this automated input RPA for quotation reply, are already in the production phase and have become essential tools to the people responsible for the operations.

In Phase 1, our RPA Promotion Working Group worked on applying RPA to Rank S and Rank A operations, which involve workloads of over 10 hours per month. Individual departments will be responsible for applying RPA to smaller-scale operations, such as Rank B operations. In other words, department-led RPA is developed and deployed. This is Phase 2 of our RPA promotion initiative.

First, in order to eliminate employees' image of robots as magic bullets, we prepared an in-house portal site for sharing information regarding RPA. We are providing training and sample robots to enable individual departments to develop their own RPA in Phase 2 and beyond.

One of our tools for developing RPA for Rank S operations is a server robot that operates within servers, but the tools used in development by individual departments are client robots that operate within individual PCs. Each department uses time sharing to remotely log into a dedicated PC with installed RPA tools and perform development.

RPA deployment is making progress, especially in regional and branch offices with few employees, because this approach eliminates the need to install tools in individual departments, and enables the offices to implement RPA based on their own priorities.

From April 2020, we have also deployed tools with concurrent licenses, which multiple people can log into and use simultaneously. We are working to prepare environments in which every department can develop RPA comfortably.

In Phase 3, which began in fiscal year 2020, we are working on creating a system that enables the robots created by individual departments to be used across department lines. As part of that effort, we have launched a robot collection site. Information regarding the robots created by each department is registered in the site so that the robots can be used throughout the company. We are also considering sharing the existing robots developed by the RPA Promotion Working Group. In the future, we will continue to develop operation robots that can be used by multiple departments, as well as calling for robots to be used throughout the company.

Furthermore, to improve the flexibility with which operations can be selected for RPA, we have set up the registration web site that the person in charge of each operation can register the operation wanted to apply RPA. Based on the needs for all departments identified by promotion committee members in each department, the person in charge of each operation who is familiar with the business can register the target operation for RPA, so that it is possible to collect detailed information to apply RPA.

We've looked back on our past efforts and gained several insights.

The first is that RPA is the first step in taking inventory of operations and improving their efficiency. Through our project, we've been able to discover operations that have been carried out as a matter of course but whose objectives are unclear.

The second thing we have learned is the importance of creating FAQs^* regarding RPA tool usage methods, usage procedures, robot sharing, and the like. Individual departments are developing and deploying RPA themselves, so our RPA Promotion Working Group has received many inquiries. FAQs are essential for reducing corresponding workloads, promoting the development of RPA, and organizing knowledge regarding RPA.

The third is that there are cases of operations for which RPA is effective despite the operations not involving a large number of person-hours of work. Regional and branch offices handle regular operations with small staffs. Even if each individual work process takes little time, applying RPA can greatly improve overall work efficiency.

One of our future tasks, in addition to deploying RPA across multiple departments, is investigating how to combine RPA with AI to achieve further operation efficiency improvements. In doing so, it will be important that we decide on operation rules, such as the scope of what can be left up to robots. To determine the shape of these rules, we will continue to deploy RPA and gather new knowledge.

RPA is one way of improving operation efficiency, but unlike past operation efficiency improvement measures, it enables operation staff to make changes to processes themselves. We believe that it can contribute to our work style reform objective of creating an environment in which all employees exert autonomy and initiative in their work.

That said, we have yet to reach the stage of full-fledged deployment and utilization of RPA in our company. We have just completed what could be called the "marketing phase," creating a company-wide movement towards leveraging RPA.

In the future, we will further develop the functions of the RPA tools themselves. We will utilize these to further improve operation efficiency and quality, and promote our work style reform.

We will then provide the results to our customers in the form of solutions and services. Please expect by all means.