Experts in Conversational AI: Interview with ICS.AI's Data Analyst
Updated: Jul 29, 2022
Shubhangi Goyal is ICS.AI's resident Data Analyst. She works across the business on ensuring data accuracy and providing insights both internally and to customers on the effectiveness of our conversational AI solutions.
Hi Shubhangi, can you give us a brief explanation of your role at ICS.AI?
My role at ICS.AI involves collecting, organising and studying data from our AI assistant solutions to provide insight into business benefits, and identify areas for improvement. I work with various tools and techniques for data analysis, visualizations, present data to the customers and define new product opportunities.
What kind of benefits are you delivering to customers?
With the data and insights generated on that, the customers can understand how our chatbot solutions have been performing. We provide them with various information based on our analysis which can in turn be used to improve their operational efficiencies. This also presents an opportunity for continuous improvement of the product and innovation.
Who else within the business do you work with? What kind of collaborative work do you do?
I work with members of development team, infrastructure and language modelling based on the type of task that needs to be done. From creating dashboards and setting up environments, I work with infrastructure team. To work with improvising the language modelling, I collaborate with AI training team. If there is something I want to understand about the product and its data logging, I collaborate with the development team. In addition to this, I also work with Sales Team to provide them with any necessary information they need about our existing product performance.
What kind of technology do you use and what do you use them for?
I use various technology based on the type of task; Power BI for dashboards, SQL Server is used for data logging, R and Python for extensive data analysis, ETL frameworks for data extraction and various Office tools like Excel and PowerPoint for presenting data.
What do you like best about your role? What do you like about working with AI?
I enjoy working with data and that is the best part of my role. With data, we can be creative and generate useful insights, visualize the data and make it easier to make decisions based on it. In addition, I like working on different tools as they help me to explore different things from our existing data sets. The best part about working in AI specifically is getting to work on innovative solutions that help to save time, speed up repetitive processes, recognising patterns and the ability to make decisions based on that.
How do we remove personally identifiable information (PII) from our data to ensure compliancy for our public sector customers?
We use a combination of different tools; ETL frameworks, PowerShell and SQL Server to remove PII information from the data and ensure compliancy. After the removal of PII, we log the data into our database systems.
What data do you find is most important to our customers?
The most important data to our customers is
the performance of bots
what kind of questions are users asking to the chatbot solutions
what questions are not answered
how many questions are asked out of hours
what are the different categories of the questions
For our SMART Live Chat product the metrics differ to average waiting time, how many chats were answered, how many chats were abandoned etc.
Can you briefly describe how we quantify our language models?
To quantify our language models, we make sure that the efficiency of the model does not go below our benchmark. Our AI training team works constantly to train the model with the latest updates and questions. This helps us to keep them in line with the expected efficiency.
What do we mean by intents and utterances?
Intents and utterances are a part of our language modelling. An intent is basically the unique question that can be asked to the assistant. Under the intent, we have various utterances which means potential ways the same question can be asked.
Can you briefly describe our SMART Mesh product? How do we collate our customer data and process it?
SMART Mesh is a product where one AI assistant learns from another on the same MESH. We collect data from various sources and combine it together using different tools. After performing data pre-processing, we generate it as one, and also find out different anomalies in the data. Based on that, we update and retrain the base MESH model with the learned content.
How do we share the data capture with customers? Who usually asks for this data? What value do they get from it?
We share our Power BI dashboards with our customers and explain them the about data present on it. In addition, we do monthly meetings with them to discuss the performance of the existing chatbot solution and also identify areas of improvements. I work with customer success managers and customers to explain them about reports and answer any questions they have. Through this our customers are constantly up-to-date with the product performance and they know the different areas where we can help them to improvise the existing product.