KDDI leverages generative AI through user-friendly platforms like the Alli LLM App Market to enhance company-wide productivity. By enabling non-engineers to create AI applications and automating tasks like legal reviews and meeting minutes, KDDI fosters innovation, streamlines operations, and explores future AI possibilities, such as autonomous agents and advanced integrations.
Since the summer of 2023, the Data & AI Center within KDDI Corporation's Corporate Strategy Division has actively pursued the integration of generative AI (including large language models, or LLMs) into its operations. The objective is to make generative AI accessible to all employees—not just tech enthusiasts—enabling company-wide operational efficiency. By leveraging platforms such as Allganize's Alli LLM App Market, KDDI has implemented various initiatives, including the creation of generative AI applications for legal tasks and the deployment of task-specific AI tools.
KDDI sought out a generative AI platform designed for non-engineers, allowing employees without technical expertise to create AI applications within minutes after basic training. This approach democratizes AI development and ensures accessibility across all departments.
Generative AI applications, such as a tool for automatically generating meeting minutes on Teams, have been introduced company-wide. Efforts to promote adoption include workshops, sharing use cases, and strategies to maximize the usage of tools like Microsoft’s Copilot, avoiding potential underutilization.
While tools like Copilot are effective for automating tasks such as scheduling Teams meetings, KDDI has also developed bespoke AI applications for more complex tasks requiring deeper integration with internal systems, addressing unique operational needs.
KDDI’s AI strategy is forward-looking, exploring advanced possibilities like autonomous AI agents and other cutting-edge applications. This approach positions the company to remain at the forefront of AI innovation.
Through a combination of accessible tools, practical training, and tailored solutions, KDDI is embedding generative AI into its core operations. These efforts are fostering a culture of innovation and driving company-wide productivity enhancements.
Doi (Allganize):
KDDI has been using the Alli LLM App Market as one of the platforms for integrating generative AI into its operations. What made you decide to try the Alli LLM App Market?
Kashimoto (KDDI):
When I first came across the pamphlet for the Alli LLM App Market, my immediate thought was, "This looks very easy to use." While generative AI typically involves text-based interactions, which are inherently intuitive, it remains a cutting-edge technology. This often creates psychological barriers, such as concerns that it might feel overwhelming or require programming skills.
What caught my attention about your single-action apps was their simplicity. Users can just describe what they want to achieve, press a button, and instantly create a generative AI application for operational use. The process was surprisingly straightforward. Additionally, I felt the interface was thoughtfully designed with the user in mind, which encouraged us to give it a try.
Doi (Allganize):
Thank you for sharing that. It’s true that earlier AI systems often required technical skills like proficiency in Python or the ability to train models, which created a high barrier to entry. Even though generative AI is easier to use, some people still find it intimidating. We’re delighted to hear your positive experience, as the Alli LLM App Market was designed to be a user-friendly platform for both administrators and end users.
Doi (Allganize):
When testing the Alli LLM App Market, one of the primary goals was to assess its usability for employees unfamiliar with IT systems.
Kashimoto (KDDI):
Yes, our focus centered on two main questions:
Target Tasks: The project concentrated on text-based processes—tasks that are manageable by humans but could benefit from AI to improve efficiency.
Goal: To determine whether non-engineers could independently create AI-powered applications and if these applications could enhance productivity across the company.
To evaluate the practicality and effectiveness of the Alli LLM App Market, we involved on-site employees in the process and gathered feedback through interviews to ensure the platform worked well in real-world scenarios.
Doi:
How did you select the departments for testing the platform?
Kashimoto:
We collaborated with the Legal Department for this test. This department handles a large volume of contracts and operates like a “human chatbot,” fielding internal requests daily—making it an excellent candidate for automation.
While customer-facing departments, like call centers, already use AI chatbots, internal administrative tasks (referred to as "middle operations") often struggle to assess the cost-effectiveness of AI adoption. The Legal Department served as a representative example of internal operations that could benefit from AI-driven improvements.
Doi:
What was the team composition for the project?
Kashimoto:
The team included a mix of participants:
This diverse group allowed us to objectively evaluate the skill levels necessary for creating and deploying AI applications using the Alli LLM App Market.
Before involving the Legal Department, we conducted internal trials to familiarize ourselves with the platform. We created sample applications using the Alli LLM App Market to identify potential challenges and refine our training materials.
The two-month preparation period allowed us to fully understand the platform’s capabilities and provide comprehensive guidance to the Legal Department, ensuring they were well-equipped for the trial.
The structured project approach, which included diverse team involvement and targeted task selection, enabled KDDI to assess the feasibility of integrating generative AI into internal operations. This collaborative effort showcased the platform’s potential to empower non-technical employees and streamline operations throughout the organization.
Doi (Allganize):
After your internal trials, I understand the Legal Department had only one day for their part.
Kashimoto (KDDI):
That’s correct. We wanted to test whether non-engineers could quickly learn and use the platform, so we allocated just one day for their evaluation. After a short tutorial on the Alli LLM App Market, the Legal Department began creating generative AI apps tailored to their work.
Doi:
I heard they found it easy to create the apps. Can you share more details?
Kashimoto:
The apps they developed included a legal document review tool and a text summarization app. Some of these were completed in as little as 10 minutes, proving that even non-engineers could use the platform effectively.
Doi:
They created these apps right after the tutorial?
Kashimoto:
Exactly. What stood out was the focus of the three participants from the Legal Department. They worked quietly and efficiently to build apps. Typically, when introducing new tools, we get a lot of questions like, “What does this function do?” or “How do I use this feature?” But this time, there were no such questions. They quickly grasped the platform after a simple tutorial, which was a first for me.
This ease of use highlights not only the intuitive nature of generative AI but also the thoughtful design of the Alli LLM App Market. Its user-friendly interface makes it surprisingly simple for non-engineers to create functional applications.
Doi:
Thank you for sharing that. From the beginning, we’ve prioritized no-code platforms with simple interfaces, such as flow-based AI chatbot builders. The feedback from your Legal Department reflects our commitment to making AI accessible to everyone.
Kashimoto:
Afterward, we interviewed the Legal Department participants, and they reported no psychological barriers or difficulties using the platform. They also felt confident in setting prompts for future generative AI applications, providing strong validation of its usability.
Kashimoto (KDDI):
Beyond the Legal Department’s specific use cases, we implemented a meeting minutes app across the company. This app converts Teams meeting transcripts into well-structured minutes. When we invited employees to test it, over 80 people immediately signed up for early access.
Doi (Allganize):
How is the app performing in practice?
Kashimoto:
It’s performing very well. Previously, we had an in-house app for generating meeting summaries, but we sought to improve its quality. While Microsoft’s Copilot offers similar features, not all employees—such as contract or temporary staff—have access to Copilot licenses.
The meeting minutes app developed through the Alli LLM App Market is accessible to everyone and produces high-quality summaries, meeting our need for a company-wide solution.
KDDI is continuously exploring new applications for generative AI. For instance, the company is developing user-friendly legal summaries tailored for employees who may not be familiar with complex legal terminology.
The success of initiatives like the meeting minutes app highlights the potential of generative AI to drive operational efficiency. By prioritizing accessibility and impactful solutions, KDDI is making generative AI an integral part of its operations, benefiting diverse teams across the organization.
Doi (Allganize):
It seems that evaluating the effectiveness of no-code generative AI platforms was a key focus for you.
Kashimoto (KDDI):
Yes, we also tested other platforms like Dify. While Dify is a no-code tool, it does require some engineering knowledge, such as library setup. It’s a fantastic tool for technically skilled members, offering a lower barrier compared to coding from scratch. It’s particularly useful for engineers looking to streamline their workflows.
Doi:
Indeed, Dify is well-suited for users with coding experience, offering broader capabilities.
Kashimoto:
Exactly. We also explored Copilot Studio, which initially appeared as user-friendly as the Alli LLM App Market. However, as we delved deeper, it became apparent that advanced settings and integrations required coding knowledge. For tasks like detailed configurations or troubleshooting, support from engineering-skilled team members was often necessary.
Doi:
By contrast, the Alli LLM App Market is designed for non-engineers while still allowing seamless system integrations.
Kashimoto:
That’s one of its standout features. It supports non-engineers while maintaining the flexibility to enable deeper functionalities when needed. This dual approach makes it a compelling tool, and we’re excited to explore its potential further.
Doi:
What are your future plans for generative AI in the Data & AI Center?
Kashimoto:
We’re exploring multiple directions. One promising area is AI agents—systems that can handle abstract tasks without detailed prompts and proactively request additional information from humans. These agents would better understand overall context and collaborate more effectively. We’re keen on finding ways to implement such capabilities.
Doi:
AI agents are a major focus for us as well, and we’re steadily implementing features in this area. For example, we’ve developed functions where AI selects the optimal LLM model for a task or auto-generates generative AI apps based on user text input. We’d love to collaborate further on this.
Kashimoto:
That sounds promising. Another potential application is using AI for creative reviews, much like how the Legal Department leverages generative AI. Creative reviews can be time-intensive, so having AI perform an initial check could significantly enhance efficiency.
Doi:
We’ve already launched apps like pharmaceutical compliance checks, which could be adapted to suit your creative review needs. Additionally, we’re heavily investing in Retrieval-Augmented Generation (RAG) to enhance accuracy and usability in enterprise contexts. We’d be happy to support your efforts in this area as well.
Kashimoto:
We’ve noticed your recently published Japanese RAG Leaderboard—it’s very impressive. Our goal is to provide user-friendly RAG tools, and we’ll benchmark our efforts while considering your solutions.
Doi:
We appreciate your feedback. Is there anything else you’d like to see from Allganize?
Kashimoto:
One of the most valuable aspects of your platform is the ability to quickly test the latest LLMs in real business applications. As new models are released, Allganize integrates them rapidly, allowing us to efficiently evaluate their impact through the Alli LLM App Market. This capability is essential for driving productivity improvements across the company.
Doi:
We’ll continue to improve the Alli LLM App Market to meet your expectations.
Thank you for sharing your valuable insights today.
