About Nagano Prefecture:
Area: 13,560 km²
Population: 2,008,353 people (As of June 1, 2023)
Nagano Prefecture is located in the central part of Japan and is renowned for its majestic mountains of the Japanese Alps and the expansive landscapes of Shinshu Plains. Skiing and snowboarding are popular winter activities, and visitors can indulge in local delicacies such as Shinshu soba noodles and Shinshu apples. With a wealth of historical tourist sites, it's a prefecture where one can enjoy a heartwarming blend of nature and history.
We spoke with Mr. Sadakiyo Aida, who is the section chief of the Digital Transformation, Department's Planning and Promotion Division in Nagano Prefecture, and Mr. Gorou Izuru, the chief examiner about the background, effects, and their impressions of the Crew demo.
Q. Can you provide some background on the Crew software POC (Proof of concept)t?
A. The Crew software demonstration experiment was conducted in Nagano Prefecture to explore the potential of online generative AI services, such as ChatGPT, in enhancing business efficiency. While these AI services are recognized as valuable digital tools, they also raise concerns regarding information security and the commitment of our staff to their responsibilities. Given these considerations, we found it imperative to understand these services' characteristics thoroughly, leading us to carry out verification tests for various generative AI tools.
Q. Why did you select Crew over the other ChatGPT solutions available?
We chose Crew because it uniquely offered the capability to generate responses by referencing multiple documents within an organization, a feature not commonly found in other services. Additionally, knowing that Crew’s parent company, Monex Group (8698 on the TYO), a domestic financial institution, had adopted Crew reassured us of its security measures. Moreover, among the solutions available for municipalities, Crew distinguishes itself with its feature-rich management dashboard, which includes the ability to register prohibited words, among other essential functions.
Q. How did you implement and use the Crew software in your workflow?
We integrated Crew software into our system to enhance our document processing capabilities. Leveraging Crew's features, we were able to streamline our workflow by efficiently referring to multiple documents and utilizing its advanced AI capabilities to draft new documents. Furthermore, Crew assisted us in generating Q&A sections that adhere to specific ordinances and regulations.
Q. With Crew, there's a feature that allows users to upload and reference documents to provide answers. Which types of documents have you uploaded and utilized?
We have uploaded and referenced documents related to regional administrative paperwork and accounting using Crew. Notably, the accounting documents are extensive, and if printed, they would stack up to a height of about 10 centimeters.
Our Reference manual was a cumbersome, 10 CM thick physical copy.
Instead of sifting through a hefty 10 CM thick manual, you can simply upload our accounting-related documents to Crew. For instance, if you ask, "I am considering purchasing goods worth more than 1 million yen. What are the necessary procedures?" Crew will swiftly provide a concise response, such as: "When purchasing goods, ensure you follow the outlined procedures. The supporting documents you attach should detail the purpose, item, specifications, etc." This allows you to pinpoint the important procedures effortlessly. Crew aids in efficiently highlighting the primary steps after which we only have to perform a quick review..
Q. How has implementing the Crew software tool affected your operations?
A. Before using Crew, when dealing with accounting documents, it would take several minutes, sometimes over an hour, to sift through and locate relevant sections. Now with Crew, these tasks can be completed in mere seconds to a few minutes. Moreover, Crew allows for quick referencing to original provisions within the same short time frame. As a result, we estimate a time-saving of about 90% for our staff members on these tasks compared to our previous methods.
Q. By using Crew, you eliminated 90% of the work. Are there any additional time-saving benefits?
Absolutely. One significant instance is our experience with the documentation on the standardization and commoditization of municipal information systems. Using Crew, we've managed to expedite responses, which has dramatically reduced our work hours. Previously, a process that took about an hour is now reduced to mere seconds or minutes with Crew. These materials, provided by the Digital Agency to municipalities, are often extensive, spanning 70-80 pages in PDF format, containing both text and detailed diagrams. In the past, just skimming through these documents took an hour due to their complexity and the technical terms they contained. Now, with Crew, the process is much more efficient, and there's no longer a need to constantly search or recall complex terms while understanding the content.
Q. Sounds like a tough document. What specific question did you present to the Crew?
I asked Crew, "Please list the key points regarding the Fit & Gap for standardization". Normally, if an employee were to read and research this list, it would take a considerable amount of time. However, with Crew, I received the desired response almost instantly, and it was especially helpful that it showed the source of the reference.
Furthermore, the documents distributed by the government are frequently updated, and it's time-consuming to check what exactly has been changed with each version. If you wanted to understand the changes and the progression from version 0.1 to 0.2, for instance, it's hard to immediately discern the differences by just looking. Given the high page count, it's not like we remember every detail from the previous version, making comparisons cumbersome. Therefore, by uploading the latest version to Crew, I found it convenient to receive answers directly from the most recent document.
Note: "Fit & Gap" refers to the process of comparing the functions that a service provides with the functions that the service user demands, to determine the fit and the discrepancy.
Q. What additional applications do you envision for the Crew software tool?
A. I envision Crew being used to streamline the document creation process. Previously, crafting a document required selecting reference materials, searching through their contents, and reviewing them thoroughly, which took about two hours. With Crew, the software identifies relevant sections, references them, and generates the text, reducing a two-hour task to just a few minutes. While humans will still need to do the final check, Crew is highly effective as a starting point or for summarizing key points.
Q. What other Crew features have you utilized?
A. We utilized the prompt templates available within Crew. In contrast to other services where we had to manually search for or copy prompts, Crew offers them conveniently organized by use-case right on the screen. With a single click, we could apply these prompts, enhancing our user experience.
To effectively employ a generative AI in business, the quality of its response significantly depends on the prompt given.
Prompt: A directive or instruction given to a generative AI to elicit a response.
It's vital for the user to familiarize themselves with the AI and hone their instructing methods to ensure clear and accurate AI responses. Thankfully, Crew provides multiple prompt templates by default (At the time of this interview, there are 19 different prompt templates available), simplifying the process for new users. Additionally, there were English prompt options which we found intriguing to experiment with.
Q. Do you have any request for future improvements for the Crew software tool?
A. Absolutely. I recommend enhancing Crew's capability to provide more accurate responses when it comes to specialized areas like legal and financial topics. The intricacies of law often require a deep dive into government and ministerial documents, which can't be understood from a mere surface-level statement. While these complexities can be challenging even for humans to interpret, a more refined ability for the Crew tool to process and compile information from these intricate documents would greatly expand its applicability across various administrative departments.
We indeed want to develop a unique model that the administration can easily use in the future. Thank you very much for your time!