KDDI and Azure OpenAI Service cut middle office sales preparation work by 74 percent

Establishing a center of excellence to promote generative AI use and business innovation

KDDI is evolving the power of connection while refining its core 5G communications business. The company’s KDDI Vision 2030 medium-term business plan aims to forge an ideal world, which includes promoting data-driven methodologies and implementing generative AI across society. In 2023, the enterprise released KDDI AI-Chat, an AI chat service for 10,000 employees, to expand the use of generative AI internally. Another goal is to better leverage in-house experience in customer proposals.

However, according to Rui Kimura, Senior General Manager of the KDDI Data & AI Center, “A bottom-up approach in every department could not adequately accelerate the use of generative AI.” To overcome this obstacle, Kimura obtained the backing of executives and launched the KDDI Generative AI Center of Excellence (KGA CoE) in 2023.

Kimura explains, “The CoE consists of teams called ‘pods.’ Each pod covers a separate area, such as system environments, governance, and external solutions under the Data & AI Center. We also have several use case pods, which create and validate situations for internal use.”

“Middle Office Innovation” is one such pod. Middle office operations are miscellaneous operations that exist between the front (sales) and back offices. This includes creating materials for customer proposals, scheduling meetings, and other general tasks.

Identifying time sinks and presenting solutions in three phases

The Business Design Department is pivotal to the Middle Office Innovation pod’s work. Tatsuya Takahashi, Leader of KDDI’s DX and AI Business Group, explains the challenges of middle operations below.

“Preparations for sales take up a lot of time. Our research showed that contact time with customers was only about 23 percent of all sales work. We manage countless products and spend a lot of time gathering information to make customer-specific proposals. We set a goal of using generative AI to make proposals that leverage our internal knowledge.”

Upon launching in April 2024, the Middle Office Innovation pod conducted interviews at head office to identify the causes of sales preparation time sinks. The results showed that almost 50 percent of time was spent on collecting and organizing information to create materials. Additionally, creating and reviewing the materials occupied 30 percent and 20 percent of time respectively.

The Middle Office Innovation pod decided to solve this problem in three broad phases: streamlining document searches; simplifying document creation, review, and sharing processes; and evolving document creation with semi-automated tasks.

The first phase began in May 2024 and a beta version launched six weeks later. Following consistent improvements, a system for the first phase was completed in September 2024.

Vast majority of users spend an average of 74 percent less time collecting information

Yosuke Matsuura, Deputy General Manager of the Business Design Department, provides more details on KDDI’s system, an overview of which is shown in the figure below.

“Our system enhances efficiency with AI-generated slide images and summaries and Azure AI-powered search. Microsoft Entra ID ensures security. Preprepared SharePoint materials and auto-completion functions speed up searches. A results page displays previews, marking the most relevant slides for easy information discovery.”

KDDI employed agile development for the system, with a one-week sprint. During the development period, the company conducted user interviews and a survey to get feedback on the system. According to the survey, 80 percent of users were able to gather information faster, with time savings of about 74 percent per person. Displaying an image of a slide and a summary as search results enabled users to more intuitively identify the materials they needed.

“You don’t discover the effects of generative AI until you use it,” says Okazawa. “As such, our development team worked closely with the KDDI departments that had planned and used the system, actively incorporating feedback from interviews.” Takahashi responds by adding, “We’re appreciative of their instant responses to our feedback.”

Accelerating the sales cycle and expediting meetings

In late 2024, KDDI initiated the next phase of its project, focused on enhancing the efficiency of document creation.

The development team is working on functions for creating KDDI presentation materials and short summaries. Masayoshi Kobayashi of the Business Innovation and Promotion Department comments on the expected gains below.

“Streamlining information collection and using generative AI to create PowerPoint documents accelerates the sales cycle. Generative AI can replace humans in routine tasks, such as preparing KDDI presentation materials and summary slides. Besides the benefits for sales reps, these summaries can lift the speed and quality of reviews and meetings.”

With verification of the second phase underway, Takahashi says, “We’ll start pushing external sales when we’ve seen a certain degree of results.” He’s also planning to encourage other headquarters to use the system to build and share knowledge more widely.

“In the past, we’ve set key themes, such as middle business innovation, for everyone to work on, but we aim to promote citizen development going forward,” says Kimura. KDDI will continue to nourish the use of generative AI among workers by employing the technology to suggest future initiatives and business challenges to solve.

According to Kimura, “We also use Microsoft Power Platform, which we can combine with Microsoft Copilot to create diverse tools without having to develop from scratch. Our development departments are already using GitHub Copilot. We’re excited about leveraging our own experience as 'client-zero' in these areas in proposals to our customers.”