Let us take a glance at what DistroKid has to say
The Unique Journey
DistroKid’s business model is unlike many e-commerce solutions; they empower artists who upload proprietary content across all major music platforms (Spotify, Apple Music, etc) and collect royalties. Three key challenges made managing and scaling customer support difficult:
- Extremely specific/ complex questions: Customer escalations were highly specific and complex, requiring detailed knowledge of content types, streaming platforms, and royalties structures. For example, artists often had issues unique to certain platforms, like Spotify's specific metadata requirements or Apple’s upload limitations. Addressing these queries accurately required a deep understanding of multiple platforms' nuances. To answer, human agents would need to build a wide ranging context and an understanding of how those issues inter-relate, leading to long call times.
- The extreme long tail: The team handled 500+ ticket categories, driven by diverse content types, listing platforms, payment structures, artist collaborations, and geographic laws. The long tail of categories is often described internally within the DistroKid support team as "artists can and will write in about absolutely everything". The complexity of each user request is effectively limitless as each user can have problems with multiple DistroKid products, multiple platform requirements, billing, earnings, and more... all in the same ticket.
- Accuracy: As an earning-enabling platform, it was crucial for DistroKid to provide highly accurate guidance in the very first attempt to maintain brand trust. Any misinformation could potentially lead to financial loss for artists or legal complications, making precision paramount.
To address these challenges, DistroKid partnered with Fini AI to develop a tailored AI-powered customer support solution.
From the start, we were in intense collaboration with DistroKid to understand their needs. It became clear that DistroKid needed three new custom features built:
- Large-Scale Categorization: DistroKid's support team faced an overwhelming volume of complex queries across multiple music platforms. The manual categorization process was inefficient, causing delays and inconsistent responses. Fini implemented an ML categorization model that learned from past data, achieving 99% accurate categorization from day one. This prevented a pressing pain-point of mis-labeling that caused customer frustration as they’d need to speak to multiple agents, leading to longer TAT. They would also need to spend valuable agent hours re-assigning those requests. The model now excels in distinguishing nuanced categories, such as metadata errors versus platform-specific upload failures, creating immediate impact. It is also continuously improving and becoming more specific as knowledge and instruction are updated.
- Continuous Learning Feedback Loop: When dealing with diverse categories and constantly evolving rules from listing platforms, a continuous learning process was crucial. We addressed this with our robust Feedback Loop.If the trained AI agent encountered updated rules it didn't know, it would escalate the query to a human agent, who would provide the updated data source. This information was quickly integrated into the backend, allowing the AI to self-resolve similar issues in the future. If a query was not in the Fini knowledge base, it was flagged for human review and used for AI retraining. Utilizing Fini’s Feedback Loop, the AI agent rapidly incorporated new information, ensuring accurate handling of new queries. This dynamic feedback loop continuously enhanced the AI's capabilities.
- Agentic Flows: As it goes in Animal Farm - “all animals are equal, but some animals are more equal”. Similarly, while all categories are important, some are more sensitive. Our goal was 100% accuracy for these critical categories. We developed a feature called ‘Flows’ for specific queries requiring sensitive handling. By integrating business-driven logic with AI, we reduced unpredictability and ensured precise actions. This enabled DistroKid to handle sensitive topics, such as unfulfilled payment withdrawal requests or time sensitive metadata edit requests with increased speed and accuracy. These high-stakes issues were routed through predefined process flows, collecting necessary inputs from the users and then routing them through set personalized channels, ensuring they were managed with the utmost care and precision instead of blurting same static answers.
Cherry on top was that we were able to automate a whole bunch of very niche use-cases. For instance DistroKid had an issue where artists would request withdrawal and it got messy owing to different platform policies and each artist’s different minimum withdrawal thresholds. With Fini we were able to construct designated flows and get users to resolution in milliseconds now (pfb example).
Controlled Launch: Fini and DistroKid executed a controlled launch over the top 10 categories first, ensuring a smooth transition and sustained customer satisfaction. The AI manager maintained a rigorous accuracy check and worked closely with DistroKid’s CX leaders for continuous improvement. This phased approach allowed for real-time adjustments and ensured the AI system met DistroKid's high standards.
Within a span of 3 months, we saw astonishing results.
- We went from 0% to 99% automation in accurate query categorization
- We were able to reduce 5-10% monthly support cost driven by AI automation. We are on track to save north of 30% support costs by the end of the year.
Conclusion: DistroKid's partnership with Fini AI highlights the transformative power of AI in customer support. By automating half of their query handling and achieving near-perfect categorization accuracy, DistroKid streamlined operations, reduced response times, and enhanced customer satisfaction.
Learn how Fini AI can revolutionize your customer support operations. Visit https://www.usefini.com/ to discover more!
.svg)
.svg)
.svg)
