NTT Docomo deploys PKSHA AI agent that asks customers clarifying questions

Summary

NTT Docomo has deployed PKSHA's ChatAgent conversational AI in its online billing and contract support channels. The system shifts the burden of clarifying intent from the customer to the AI. Before directing users to the correct procedure, ChatAgent asks follow-up questions about specific lines, time periods, and other details. PKSHA also offers a paid 'response stability' add-on to keep delays within acceptable levels during high traffic periods.

Why it matters

Self-service in telecommunications is a high-volume, high-friction touchpoint. In the past, ambiguous customer queries (e.g., 'billing change') often led to lengthy FAQ searches or handoffs to human operators. Docomo reverses the interaction by having the system ask the user questions. This allows routine requests to be routed efficiently, enabling human agents to focus on genuinely complex cases. This move also signals that Japanese carriers are moving generative AI-based customer service from the pilot phase to production environments. This step will likely pressure competitors to take similar actions.

Highlights

• The AI reduces self-service friction by asking clarifying questions about lines, timing, and plans, rather than expecting customers to formulate precise queries.
• The paid 'response stability' option pre-allocates computational resources to maintain consistent response times during peak query periods.
• PKSHA claims its platform has handled over 750 million cumulative conversations. The company plans to expand the agent to sectors beyond telecommunications.

Critical take

The article reads more like a PKSHA vendor press release than an independent news report. It lacks data on resolution rates, customer satisfaction (CSAT), or cost per contact. The 750 million conversation statistic is cumulative across all clients and not specific to Docomo. The 'stability' option is essentially a temporary workaround for scaling issues. Furthermore, asking two or three follow-up questions is a common technique in chatbots, not a generative AI breakthrough.