KDDI Sets Out AI Deployment and Customer Experience Strategy
- •KDDI held its summit on October 27–28, outlining how AI can support customer experiences and enterprise deployment.
- •At au stores, high-value plan changes were about three times as common as online; AI cut search time by 80–90%.
- •KDDI began offering an AI vulnerability management service on October 28; its core AccelWaves network is planned for fiscal 2028.
KDDI held KDDI SUMMIT 2026 on October 27–28 at Takanawa Gateway Convention Center and online, under the theme “Bringing the potential of AI to life so each person can shine.” At the opening keynote, President Koji Matsuda said trust, attachment and emotionally resonant experiences would remain reasons customers choose companies even when AI compares prices and specifications. The next day, Director Hiroaki Hosoi outlined KDDI’s plan to integrate AI into business operations and systems and support their operation after deployment.
Matsuda cited examples of AI agents moving deposits to the financial institution offering the best terms, or canceling a fan club membership because the member had logged in less often. He said such decisions may be financially rational but could weaken people’s sense of security and connection to things they value. He said the marketing principle that customers pay for experiences they value remains unchanged, and companies need to identify which strengths to assign to AI and which tasks people should handle.
Matsuda said customers changing smartphones at au shops switched to higher-value plans at about three times the rate of online customers. AI search assistance reduced the time needed to find an answer by 80–90%, and halved 200,000 inquiries from shops to support desks. KDDI Avatar Canvas, a service for practicing customer interactions with an AI avatar, begins business service on November 2, 2026; AI Roleplay will also be offered that day to staff at au Style and au shops nationwide. Matsuda described combining people and AI to enhance existing strengths as “buffing with AI.”
At the keynote discussion, Naoshi Fujikura, president and CEO of Universal Music LLC, said trust and excitement are reasons customers choose a company. He said music marketing can draw on past data, but following it does not guarantee a hit; people’s intent and enthusiasm for what they want to deliver matter.
Hosoi grouped corporate AI use into three types: AI supporting employees’ daily work through tools such as Copilot and Gemini; AI assisting specialized work through business-focused SaaS; and “AI inside systems,” integrated into company data and business systems. He said the third area can create business competitiveness, and listed cloud data migration, system modernization, AI analysis, agent development, operational efficiency and impact measurement. KDDI plans to work with KDDI Aile to assess companies’ differing progress and initiatives across departments.
In a related talk, Shugo Takagi, president of KDDI Aile, said implementation requires integrating company-specific data into AI, embedding it in daily work and making results visible so gains continue. Companies moving ahead with adoption may realize they need to make veteran employees’ tacit knowledge and specific context usable by AI, prompting them to return to data preparation. KDDI’s in-house development began with five people in 2013; planning, development and operations staff have worked as one team, repeatedly building and checking prototypes. AI-driven development is speeding up the cycle of showing prototypes and improving them.
Hosoi demonstrated a voice AI responding to customers who want to cancel. KDDI continues adjustments and trials with a view to using it in au’s contact center, aiming to improve overnight service and difficulty getting through, and to improve employees’ working conditions through measures against customer harassment. At the Takanawa office Lawson store, staff and planning and development teams tried measures such as timed discounts in response to fewer Friday visits and waste from unsold goods. Takashi Kimura said they observed that some customers who used a sale once returned on other Fridays. The team is also testing a system in which AI suggests prices based on inventory and visitor numbers, while staff make the final judgment using their experience. It is emphasizing a feedback loop that sends AI the reasons and context behind rejected suggestions so they can inform later decisions.
Kimura said frontline insights can become company-specific data and create differentiation even when companies use similar AI. KDDI is still exploring company-wide rollout and links to evaluation systems, and plans to start with people ready to make improvements and build successful examples. A press briefing also addressed how to change labor-based contracts and revenue models as AI-driven development reduces work hours. Kimura said new contract formats have not been established; Takagi said the company is exploring business models combining communications infrastructure and ongoing services. For customers building in-house capabilities, KDDI sees demand for continued AI development know-how and outsourced work they cannot handle on their own.
Hosoi said security measures need to continue through five processes: visibility and diagnosis, improvement and optimization, implementation, operation and incident response. K
The fastest, most reliable AI news, straight to your inbox