Holy Stone uses AI to handle routine support work

Founded in 2009, Holy Stone sells drones, baby products and children’s toys online. With operations in Shenzhen, Fujian and Xiamen, it has been especially successful for global brands. It specializes in developing and selling quadcopter drones and is a well-known brand in Amazon’s drone category. Under the motto “small city, big dream,” it aims to make HolyStone a globally competitive company where employees are happy.

Holy Stone4 min read

Holy Stone customer story
35%Staffing effort saved
50%Efficiency gain
10%Reduction in lost business
Table of contents
  1. Background: a dedicated AI agent makes serving customers across time zones easy
  2. Customer engagement: AI intent recognition personalizes replies and removes repetitive daily work
  3. Troubleshooting: large language models make support friendlier and more accurate
  4. Purchase inquiries and negative review follow-up: human agents & AI working together automatically

Background: a dedicated AI agent makes serving customers across time zones easy

As a toy drone brand, Holy Stone’s customer service team is responsible first and foremost for serving large numbers of consumers across time zones and regions. As the company has grown, its toy categories and business lines have expanded and it has entered new markets in China and abroad, steadily increasing the pressure on customer service.

For a company committed to an excellent shopping experience and quality service, responding promptly, minimizing errors and serving customers well are essential. Facing these pressures, Solvea worked closely with Holy Stone’s customer service department to make sure a dedicated AI agent serves Holy Stone’s customers reliably and well.

Customer engagement: AI intent recognition personalizes replies and removes repetitive daily work

“Replying to these tickets used to take three and a half hours. Now it takes just half an hour to review them.”

Day-to-day operations work at global brands is frequent and mechanical, and Holy Stone is no exception. Every time a new campaign launched, the customer service team spent a long time answering similar questions.

When campaigns such as coupons, social media promotions and discounts go live, a traditional chatbot can handle part of the mechanical work of explaining how to take part. Over time, though, it becomes clear that consumers have other needs at the same time. Imagine asking about a drone problem and getting instructions on how to claim a coupon: answers like that inevitably frustrate consumers and can even turn them against the brand.

Solvea AI uses large language models to identify the intent of each customer message first, sorts messages accordingly, and responds once the consumer’s specific intent is clear. It generates natural, pleasant replies and avoids asking again for information the customer has already provided, handling this well-defined scenario more effectively and in a friendlier way.

A senior customer service advisor on Holy Stone’s customer team said: “Replying to these tickets used to take three and a half hours. Now we just spend half an hour reviewing them. What AI sends to consumers is basically correct, and when it can’t answer, it alerts a human agent right away. It frees up our hands and takes a lot off our minds.”

Holy Stone: Implementation progress illustration

Troubleshooting: large language models make support friendlier and more accurate

“Fast, accurate responses, and preparing early for major promotions, Black Friday and Christmas.”

Drone faults are varied and complex: any part can have problems, from accessories, GPS and the drone itself to the battery, APP connection and onboard camera.

In the GPT-3.5 era, even with broad knowledge, AI could only handle fairly simple troubleshooting. The previous generation of large language models struggled with more complex messages, such as “the battery is fine, but the other problems are still there.”

Solvea uses the GPT-4o large language model, fine-tuned with its deep understanding of large language models and extensive hands-on experience. As a result, AI keeps repeat questions to a minimum based on what the customer has already provided and answers the remaining questions from the knowledge base.

“Because Solvea AI took over ticket handling for consumer engagement, we could focus our people on organizing the knowledge base. Today, AI automatically runs the initial troubleshooting diagnosis within tickets. As major promotions, Black Friday and Christmas approach, fault reports will grow exponentially, and responding to customers promptly is bound to help bring down the return rate. We’re training and preparing AI in advance to handle these complaint surges better,” said Holy Stone’s customer service manager.

image

Purchase inquiries and negative review follow-up: human agents & AI working together automatically

On every e-commerce platform, consumers are free to voice their needs and opinions, from positive feedback to complaints. Every product has strengths and room for improvement, so consumers often leave reviews based on their actual experience.

On Amazon, negative reviews are routed automatically to the seller, and Holy Stone is no exception. If a negative review isn’t handled promptly, product sales take a direct hit.

For this scenario, Solvea designed a combined AI and human approach for Holy Stone, with flexible handling to meet consumers’ actual needs. It keeps AI from overpromising to consumers; if that happened more often, the cost of handling negative reviews would rise sharply.

Human agents use professional, flexible solutions to reassure consumers quickly and resolve negative reviews. AI acts as a customer care concierge, automatically following up on tickets until the consumer accepts the solution and replies to the email. However busy human agents are, AI makes sure every consumer’s request gets attention.

Customer service goes well beyond after-sales.

Not every message customer service receives is about after-sales; a good share are pre-sales purchase questions. With Solvea AI’s pre-sales solution, AI identifies the products or accessories a consumer wants to buy, and an automated process pulls stock levels and purchase links and sends them back quickly, protecting sales opportunities 24/7.

The AI customer service system has greatly improved Holy Stone’s efficiency and customer experience, making complex service scenarios simpler and more precise and letting staff spend more time improving knowledge and the customer experience. Going forward, Solvea and Holy Stone will keep exploring more, and deeper, customer experience scenarios, using AI to support Holy Stone’s steady global expansion.

Holy Stone: Project milestones illustration

Put AI Customer Service Agents
to work in your real service workflows

Share your industry, channels and most common questions. Our industry experts draw on work with brands in your category to map out a rollout plan and expected impact.