Introduction

Peecho operates as the intermediary between websites and the best print facilities. Using a global cloud print network, they make printing fast, simple, and cost-effective worldwide. They have 15k registered merchants, for whom they're fulfilling millions of orders.

With the number and variety of orders they receive, the number of customer support cases is relatively large. Peecho takes pride in taking all of these very seriously. However, even though the company wants to ensure the best possible experience for its customers, this can become a challenge if you are flooded with tickets and crunched for time.

Peecho found the answer in the AI feature delivery platform of Unless, with an AI trained specifically on their business and the ability to leverage order information through their API. And the results are great. The AI-powered virtual agent now manages to reduce the number of tickets for certain questions on order status and lead times by over 90%, answering many specific questions that previously needed human intervention.

The challenges of customer support

If you work in customer support, you can probably list the top questions you always get asked. Chances are you won’t even need pen and paper! FAQs (frequently asked questions) exist for this reason. By listing them in advance in a prominent location on your website or in your emails, you can counteract some questions.

In the case of Peecho however, they get a lot of questions from people about their orders. Where’s my order? Has it been shipped? When was it shipped? When will it arrive? Can I change the address? etc. So much so that these questions make up 70% of all tickets they get!

At the same time, this is a question that requires some digging and a custom answer each time. You cannot simply have an FAQ for it. It is also possible that a single customer asks this question multiple times, in many different ways, and with a different answer over time.

Normally, this would require a support agent to check the order number in their system and reply to the customer. The customer sees the reply, has a follow-up question, and emails again. The support agent replying to other tickets in the meantime has to re-acquaint themselves with the ticket and reply to the customer and so on.

This interaction can take from 10-15 minutes (if during working hours) to about a day or two (if the question was asked outside work hours and/or there’s a significant time difference). If you add the productivity impact from constant task switching between tickets while waiting for a potential reply, the costs cannot be overlooked.

Then, there is the added complication of seasonality as people place more orders around the holidays. The number of customer support inquiries can skyrocket to 5x their normal levels in a given week in December, putting a strain on the customer support department.

So, to minimize the pressure on the customer service department, Peecho teamed up with Unless to deploy an AI-driven solution. The question was this: can we eliminate a significant amount of tickets by letting an AI answer most questions that customers may have, before creating support tickets that a human needs to look at? This is not as simple as it sounds, because the AI should not just be a trained product expert, but also have the ability to access individual order information for each customer.

Training the AI

An AI is like a new employee - you need to train it to become an expert. Luckily, it’s easy to train your own Unless AI in the blink of an eye. After the training has been completed, you can simply add the AI to your website or web application using one of the many component templates we provide. The Peecho case was slightly more complex, though.

The first part is simple. The AI needed to be trained on the Peecho knowledge base, requiring an integration with Zendesk. This was easy since we can already integrate with Zendesk, Salesforce, and others. What took the Peecho use case to the next level was the ability of the AI to look into the individual case details.

To do that, we created a custom AI plugin that allows the AI to answer a question by accessing data from the Peecho order API and combining that with information from the knowledge base hosted in Zendesk. This makes it possible for the AI to answer any question about the individual orders of a customer.

Furthermore, the AI is configured in such a way that if a question cannot be answered, the customer can leave their email address. On behalf of the customer, a support ticket will be created in Zendesk.

The solution at work

Let’s start by looking at the setup Peecho had before. When an order was placed, they would send out a confirmation email with an order number and delivery details. Often, people would reply to this email inquiring about their order, immediately creating a support ticket. To avoid this, the new solution had to be implemented at this stage: within the email, before a reply was sent, and a support ticket was created.

So, to trigger the AI in time, we placed a prompt in this email in the form of a hyperlink that says: Check order updates. Additionally, some context was given in the body of the email, stating: Get real-time updates on your order status from Peecho’s online assistant. It’s available 24/7. In the screenshot below you can see what such an email looks like.

Once the customer clicks the link, they are taken to the Peecho website with the AI assistant in full view as an overlay component. Here they are able to ask any and all questions they might have and the AI will answer using the documentation it was trained with, as well as the individual order information for that particular customer.

The screenshot above shows what the order overview can look like in practice. The response includes the order ID, customer name and email address, order status, destination address, estimated delivery time, and tracking number. However, depending on the order, the answers may vary.

Most importantly, people can ask follow-up questions. Whether it be about their order specifically or the different print solutions and products Peecho offers, the AI assistant is able to answer their questions. This is made possible by the integration with Zendesk and the connection to the Peecho order API.

This way a customer can ask as many questions as they like and receive answers in an instant without waiting for a reply! Below is a screenshot of what it looks like when the Peecho team monitors these conversations in Unless. In this case, a customer inquiring about the delivery date of their order. The team uses this to check for accuracy, observe trends, and improve their customer support documentation.

Sometimes, the training documentation does not provide a sufficient answer. In these cases, the user will be requested to leave their email address. Once provided, using this information, a new case will be created in Zendesk for the customer support team to pick up.

Results

In just a short amount of time, they’ve seen great results. From the 1st of September until the 30th of November, customers checked their order updates close to 3k times. On more than half of these occasions, the customer asked a follow-up question, initiating a conversation with the AI. Imagine the level of information provided, without creating a single ticket.

There were 223 conversations with a significant number of messages back and forth, leading the Peecho team to believe a support ticket would have been created, if not for the AI. They estimate that this has already saved their support team +14 hours.

Peecho’s monthly overview shows a noticeable decline in the number of support tickets created in Zendesk, with a clear uptick in conversations had with the AI. Their goal now is to continue this trend with fewer tickets that need to be handled by the team and more that can be answered by the AI.

“When I heard about the conversational UI components of UNLESS, I knew we had to use it at Peecho. Combine that with AI plugins that can help customers decide on and place orders, it’s an absolute game changer for our sales and customer support teams!” - Eric Kickert, CEO at Peecho