CDEK: 86% of all text-based customer requests are now bot-processed
Our cooperation with the delivery company CDEK dates as far back as 2017. The original goal was to reduce the costs of outgoing notifications to customers along with call center maintenance costs. Having achieved the desired result, CDEK decided to automate the other side of the process as well — that is, the processing of incoming requests.
Outcome
Customer
The delivery company CDEK has over 1,000,000 active customers, both individuals and businesses. They send 300,000 packages every day. The company has customers in 67,000 locations worldwide.
Context
With our help, CDEK has already automated outgoing requests: we moved them to messenger apps, enabled cascade mailing, and reduced the call center operator workload. Still, incoming requests continued to be processed by the hotline. At peak hours, customers had to endure long waits for their turn, which left them dissatisfied. At the same time, the company incurred significant costs for the call center maintenance.
Goal
- Reduce the call center maintenance costs
- Improve customer service quality: reduce response time, add convenient communication channels
Solution
Omnichannel chat platform and chatbots to automate communication
Until our cooperation, CDEK had no text-based channels that could be used to contact the company.
- We connected an online chat widget on the website and built a chatbot into the mobile app. Apart from that, we connected messenger apps: customers can enter them through the website widget.
- We introduced our Dialogs service to process all text-based inquiries. It collects all messages from the website chat and mobile app, as well as from Viber, Telegram, VKontakte, WhatsApp, and Apple Messages.
- Initially, in 2017, CDEK used a button bot only. During the conversation, customers would choose from response options like Order Details, Select Delivery Address/Time, etc. At first, the bot capacity was enough, but as the company grew, requests increased in number. The company needed to find a solution for automated processing of the maximum number of requests.
- In 2020, we expanded the button bot capacity with the Smart Bot. The Smart Bot is an AI-based bot. It allows customers to ask a question right away in a free format — the bot will then recognize the wording and offer the right solution to the problem.
- If necessary, the chatbot will automatically hand off the conversation to an operator. In this case, the employee using the Dialogs service will see the entire history of the customer's conversation with the company. Thanks to our integration with CDEK's own CRM system, all customer data is also displayed in Dialogs.
The project was implemented on a turnkey basis by Flomni specialists.
Button bot connected
We developed a button bot and connected it to the website chat and messengers:
- Together with CDEK specialists, we analyzed the most frequent topics of customer requests — it was necessary to automate work with them in order to minimize the load on the operators.
- For the most frequent topics of requests, we outlined the Q&A logic in conversations with a call center operator.
- This logic was transferred to the chatbot. Thanks to this, the bot can ask for all the information necessary and issue an appropriate response.
For example, on the most frequent topic — order tracking — the bot asks for the order number in the online store or in the CDEK system. Then, it verifies the information received from the customer with the database. If no order is found under this number, the bot will ask the customer to check the number they provided. The bot answers instantly.
After the chatbot implementation, operators were no longer needed to process most requests. This significantly reduced the call center workload.
Gradually, CDEK expanded the list of messengers the bot could work with. Scaling went smoothly as Flomni makes it easy to move the bot logic to new communication channels. You don't need to create a bot from scratch.
AI solutions added
In 2020, we added a free-form text recognition bot to the button bot. It was aptly named Smart Bot.
Implemented Smart Bots increase service automation by 30—40% on average compared to button chatbots. According to our calculations, a bot "takes on" about 40,000 conversations a month from call center operators.
CDEK Commercial Service
Although the contact center operations had already been automated thanks to the button bot, the operators still occasionally received routine requests. For example, requests to change the delivery address, since customers tended to ignore the corresponding button in the chatbot. The Smart Bot automated this part of the job.
To train the AI, the Smart Bot used all clients' wordings from CDEK's entire chat history. The wording versions were divided by topic, so that the bot could match them with specific customer requests.
Shortly after launch, the AI bot identified and processed over 103,000 requests in a month for the most common topic of text requests: order tracking.
Not only did we significantly reduced the number of requests processed by operators, but we also managed to keep advice quality at the high level. After the chat, customers can rate the bot if they want to. This year, the average rating was 4.67 out of 5.
CDEK Commercial Service
After the Smart Bot launch, Flomni's project manager analyzed cases when the neural network bot failed to process customer requests. The request pool was then sent to our experts for further bot training. Thanks to this, the Smart Bot is getting even smarter: next time it receives a similar request, it will be able to respond without assistance from an operator.
How requests are processed now
Answers to frequently asked questions are collected in a widget on the CDEK website
If they cannot find an answer at the first stage, they can move on to a chat with a bot in a messenger app of their choice. They can choose a menu item or ask their question in a free form. The bot will suggest the right solution in any case.
An operator can join the conversation if necessary. Statistically, it happens in 15% to 25% of all cases. All other customer needs are covered by bots.
Solutions that work, no extra effort required
Like many of our customers, CDEK ordered a turnkey project. Our experts worked on technical issues and provided integrations and setup. After the project launch, we keep monitoring the system status and, as agreed with the customer, fine-tuning it to achieve maximum efficiency.
Get free advice
Our experts will answer all your questions
and help you find the right solution
