AUTOMATING CUSTOMER TICKETS
Mission Statement
Our client uses a CRM system which has a customer ticketing tool to handle customer service requests and to track, manage and resolve problems and issues.
This involves a lot of manual work and they approached us to assist with streamlining their operations with the ticketing tool.
Our objective is two-fold.
1. When receiving a message (ticket) from a customer, determine if the customer ticket needs further action or not.
When a client message is received, the ticket is automatically reopened. Some messages do not require any action to be taken and staff need to read the message and close the ticket accordingly. This prolongs the time the ticket is open (skewing the company’s KPIs) and takes up staff time.
2. Analyse the sentiment of the message.
Our client would like to analyse the sentiment of the incoming messages to understand whether their customer is positive or negative. And then further analyse whether their customers sentiment is improving or declining. They would also like to automate whether a ticket should be escalated to management, should that be necessary.
Impact
- 80% cost savings
- Staff time on ticketing reduced
Tools Used
Alteryx
- Alteryx Server
- Alteryx Designer
- Chat GPT connectors
Detailed Solution
Our client was investigating using an AI-based system to handle the above requirements. However this solution was very expensive and required a model to be created for each type of email and the system required a large amount of emails in order to learn, taking a long time.
We presented a solution to our client that allowed them to implement a new solution tailored closely to their needs with a much cost effective solution.
Firstly we connected ChatGPT to their data via Alteryx, with Continuum’s exciting new ChatGPT Connector - which we’ve made available for free in the Alteryx public gallery.
Firstly we connected Alteryx to their shared Outlook Email Database to read in emails and parsing them, doing some prep and blending along the way. We then engineered prompts in Alteryx to send the information to ChatGPT, via our Alteryx Connectors.
ChatGPT was instructed to rate the email in terms of sentiment on a grade of 1 to 10, with 1 being the most negative and 10 being the most positive and also output a list of key phrases (comma separated) to clarify why the rating was selected. This is then logged in a data store to monitor over time to track the sentiment of their customers.
Another prompt was also created to determine if the message required further action. The ticket was then closed if it was determined no further action was required.
The added benefit of using ChatGPT in this instance is that the tool is multi-lingual, whereas the client’s alternative was only in English.
For the purposes of clarity, the client did need to pay OpenAI for tokens in order to utilise the ChatGPT service via an API, but this was a minimal cost compared to the client’s previous proposed solution.