Alessio Scotto´s interview: Stikets success story
How did you come up with the idea of implementing Oct8ne with chat GPT in your online store, and what is the need you covered?
We are always very attentive to everything that is the customer experience, we know that the first response, the famous KPI FRT has a strong impact, so what we did was to identify the communication channels where we wanted to put the focus and see where we could put AI. So we decided to put AI on WhatsApp, Instagram/Facebook Messenger, and web chatbot so that all customers go through this AI layer to solve their doubts, of course, offering in case the customer needs it, the possibility to be transferred to an agent. We have decided that in some cases, the customer will always be transferred to the agent so the case will be handled by a human. Why did we make this decision? To offer a 24/7 service with a response time of almost zero, or seconds, the time that Stiky, our chatbot needs to read and answer. Also keep in mind that agents work for example from 9 to 5 from Monday to Friday, but our customers are online also after 5 o’clock and knowing that there is someone behind to answer your questions, gives confidence to the customers.
How was the bot configuration process?
Very well, it has been almost a plug & play, the Oct8ne team has provided us with the easiest way to launch the project, and it has been nothing more than collecting all the information, which we already had both in the bot we had previously, and in our help center – a tool that customers do not want to use because they do not want to enter to read everything – and create the prompt, but then all the design of the bot has been done by you from Oct8ne. We had the initial idea but then all the elaboration of this text so that an artificial intelligence can absorb it and then be able to answer has been thanks to your team.
Was the configuration complex considering that Stikets has a lot of items in its store?
We had to think like an open AI playground, i.e. set just the right amount of time for an artificial intelligence to read it and answer. The bigger the prompt, the longer the response time of the bot. So with the help of your technicians, we have cut the information that goes in the prompt, a bit like SEO Guru’s do, when they create copy for SEO, that when you read it the information makes sense, but there is a bit of friction. Basically, it is a matter of creating a text, so that the AI can have as much information as possible without repeating itself and have the information it needs to nourish itself and give the answers.
What is the feedback from customers interacting with the bot?
We have a lot of feedback, but I would say that the top feedback is when the customer ends the conversation by saying “thanks to you” which is the top KPI, because it is very likely that when the customer ends the conversation like that, he did not realize that he has spoken with a bot. So, if this happens, it means that Stikets has done well, Oct8ne has done well, and the customer is satisfied and is very likely to have bought, so for the moment, because of course, we are talking about new things since we have just entered this fantastic world of chatbots with artificial intelligence, the feedback is good. It’s good because the customer always has an answer and if he doesn’t have the answer, he can talk to the agent.
Before implementing the bot, how did you handle all the queries?
In 2022 we have shipped more than 400 thousand orders, invoiced 10 million euros, and spoken to the customer 10 thousand times. We have spoken to 100,000 customers and 70% of those numbers happened in 2 months between August and September which is the peak we have for back to school. This means that we have to be very attentive and very prepared in terms of production, sales, and customer service in those two months. Before, we had another bot, but all these contacts that we received from customers came in through various channels: calls, Instagram, and all the channels that we have, so it was all 100% human to manage those 100,000 contacts. Now it is different, as I said at the beginning, it is important to identify channels, instead of adding customer service channels, because if you add channels, it is like the traffic on a highway, more lanes and more traffic you will have, so you don’t solve the problem. In our case it has been a bit the same, in the first months, we had all the channels, plus an AI chatbot, but the impact of the bot was not noticeable, then we did a test, which is the one we still have in place today, which is based on removing all channels of human attention and leave as a first layer, as a filter, the AI, then on the website of Stikets.es you can do nothing but call -which we know is a very high effort for the customer- or send a message in the chat or a WhatsApp and the AI starts to answer you and is there 24 hours a day, 7 days out of 7 and then, the case is evaluated and passed to an agent if necessary. Now the impact on Stikets.es is starting to be noticed because of the customers that have contacted in October, they go through there.
Can you solve all the doubts through the bot?
We know that the chatbot we have set up with Oct8ne to date solves more than 60% of the queries, this means that 60% of customers who talk to the chatbot, do not need human interaction, and I am sure that these percentages will continue to improve. But currently, it answers 60% because we have decided that in specific cases it passes the query to the agent, but it solves more than 60%. I don’t have the exact figure, but it could be 70/80% of queries.
Before you had 1 or 2 agents assigned to customer service?
Stikets management decided that AI would not replace staff resources, so what we decided was to allocate those human resources to more effective and useful tasks such as spending more time on a complex case, and working on customer loyalty now that they have more time because the bot takes away the repetitive work of where is my order, etc. Now our agents can dedicate their time to answering reviews for example, which is something else that has a strong impact and has to be taken care of, instead of spending their time answering basic information that can be automated.
What is the main reason why potential customers contact Stiky?
Two main groups are pre- and post-sales, pre-sales responds to questions related to product, or how can I customize a product, use of the product etc and post-sales are questions related to order status. In summer we sent more than 8000 daily orders, and in those 8000 daily orders there are surely customers who contact you with questions about delivery, which are simple questions that can manage the bot without problem, offering an immediate response. The bot provides a link where they can see the tracking and solve the doubts.
You have to keep pace with the evolution of the customer, who is becoming more and more demanding and who compares you more and more to companies that nowadays deliver in seconds, that have huge and personalized catalogs, etc.
How was the process of creating Stiky? How did you think of its name, its design, and the tone with which it responds to potential customers?
Sticky already existed because it was another bot, and the name has been chosen by the marketing team, since we are Stikets, Stiky made sense, and the tone was more something we decided with you. We decided to put emoticons, we did a lot of tests to make it answer in a friendly way, etc.
Stiky is available both on Web and Messaging, which is the channel through which you receive more queries? Do the questions tend to be similar in each of the channels or do they differ a lot from each other?
Web has a stronger impact, 70% web and 30% messaging.
Do you want to share any funny questions you have received that Stiky has been able to answer successfully thanks to Oct8ne integration with Chat GPT?
Every day we monitor the answers he gives, how he interacts with customers etc, and we read the questions and answers, and depending on what we read we make changes to the prompt because sometimes customers ask questions that we hadn’t thought of.
But there are very funny cases in which the client hyper-personalizes the query and says for example “Hi, my daughter Valeria starts school tomorrow and I need stickers for her clothes, but I have no clear ideas, can you recommend me?” and the bot answers “I’m so excited that your daughter starts school!” and these phrases are not in the prompt, and these cases fascinate me a lot. Another case was with a customer from Argentina who wanted to open a store in Argentina with products from our website and Stiky replied that he was pleased with his idea, but for this kind of proposal, he suggested contacting our marketing team.