The Latest News

By Agent AI 19 Dec 2017

What types of customer questions can AI really answer?

Originally Posted on TOPBOTS
December 19th, 2017
By Fred Hsu, CEO Agent.ai

Tis the season to be jolly…and to hire nearly 700,000 seasonal workers to help with the holiday rush of shopping and travel activities. As customers flock to shop online, retailers like Macy’s, Target, and Amazon plan to add between 80,000 to 120,000 temporary workers to handle offline and online customer support.

Can AI help alleviate the stress of providing top notch service when customer anxieties are running at their highest? While AI may be able to help address common customer questions about order status, refunds, or flight status, not all inquiries are created equal. Informational requests such as “what is your refund policy?” tap into static information and are generally easier to automate, since AI can query a structured database or series of FAQs to respond to the customer. Transactional inquiries, such as “where is my order?” look at historical or open customer transactions and require tapping into nonstatic data and one or more custom systems. In general, information queries are much easier for AI to address at scale for global companies.

AI’s ability to automate tasks differs by case and by company. Tophatter is a forward-thinking e-commerce and live bidding site which sees over $1M in daily sales. My company Agent.ai recently conducted data analysis of Tophatter’s customer inquiries to get a better sense of what kinds of questions they have to field on a daily basis. Here’s the breakdown we found:

Tophatter Agent.ai Customer Service Questions

Refund Questions and Issues: 38% of Inquiries

Refunds/Return Policies (Informational)

Refund/Return Issues (Transactional)

Product Issues/Questions (Informational + Transactional)

Shipping, Ordering, Purchasing: 37% of Inquiries

Shipping cost (Informational)

Bidding & Selling Basics (Informational)

Order Status (Transactional)

Order Details (Transactional)

Advanced Sales/Advertising Training (Informational + Transactional)

Payment Issues: 12% of Inquiries

Billing/Payments Issues (Transactional)

Purchase Issues (Transactional)

Account Questions: 10% of Inquiries

Receipts/History Requests (Transactional)

Account Details (Informational)

Reset Password (Informational)

Store Credit Details (Informational + Transactional)

Other Issues: 3% of Inquiries

Downtime and Outages (Informational + Transactional)

Feedback, complaints, praise (Informational + Transactional)

For an e-commerce site like Tophatter, about 50 percent of questions by volume could be automated today with zero human intervention. While these questions are simple and repetitive, they are also further from the point of sale. To truly understand how AI can be combined with human capabilities to maximize customer support, we want to make sure we align agents with the right type of activity.

Customer Service Artificial Intelligence Align Value

Humans are still needed for the highest value activities, such as dealing with sensitive billing and product issues. AI makes existing workers more efficient by halving the number of rote tickets they need to address, enabling them to concentrate on more strategic and valuable efforts. Assistant technology can also help seasonal service workers on-board more quickly by providing suggested responses in real-time, allowing them to focus more of their time dealing with complex and challenging issue resolution.

As customer service automation improves, AI will help not just seasonal spikes, but also support variability that comes from unexpected traffic, outages that cause backups, or product releases that require massive on-boarding of new customers.

But, for the time being, those 700,000 seasonal jobs should be sticking around.