AI Enterprise Ringback & Customer Service Solution

Enterprise ringback tones had been one of the company’s early core businesses. The traditional process was straightforward but heavily manual: consultation, requirement clarification, copywriting, music and voice selection, recording, post-production, upload.

As the market entered a long-tail phase, the business persisted, yet repetitive service kept consuming manual effort. For this maturing business, I redesigned the manual service flow: consultation, copywriting, and audio production became an AI-driven workflow, while humans kept handling complex cases and final delivery.

The real problem was not “how to connect an AI model” but how to break a process built on human experience into concrete tasks AI could handle.

I split it into distinct capabilities: business consultation, ringback copywriting, copy revision, audio generation, and human handling of complex cases. This turned AI from a chat window into an actual working node in the business workflow.

I designed the solution around a Multi-Agent architecture with three core capabilities. Business Consultant handles enterprise ringback consultations — products, pricing, processes, service methods, and FAQs. Copywriting Agent generates ringback copy from the company name, industry, business scope, products, and services; it can produce several style variants and lets users keep refining. Audio Agent converts the final copy into usable audio, supporting Mandarin, male/female voices, plus background music, volume, and speech speed.

I also built a business knowledge base covering products, packages, pricing rules, service processes, FAQs, company information, and upload requirements. The system follows AI + Human-in-the-loop: AI handles standardized, repetitive work; humans handle complex consultations, payments, official uploads, and special needs.

From the public account, users follow the service flow below from consultation to delivery — turning several manual steps into one continuous AI-assisted experience.

I independently handled most of the work from business analysis to deployment: process analysis, product design, agent architecture, prompt design, knowledge-base design and maintenance, multi-agent orchestration, third-party integration, workflow testing, real-user validation, deployment, and upkeep.

Technology was only part of it. The harder calls were: which tasks should AI handle, and which must stay human? How should AI outputs follow real business rules? And how could the whole flow become usable for real customers?

After launch, the solution entered real business use. Users reach the AI service through the public account and complete the full loop from consultation, copy generation, and copy revision to audio generation and preview; the system still handles real user consultations today.

The project became a working AI enterprise ringback and customer-service solution, continuously refined from actual usage.