AI-Driven Chat Terminal for Customer Support/Sales Agents

Role

Lead UX Designer

User Research, Prototyping, User testing, Interaction design

Background

24/7.ai wanted to use the huge amount of chat transcripts data and help customer support/sales agents resolve issues or close a sale by understanding the intent/needs of the customers and using the historic data and AI models.

This product was built ground up, the products that chats agents were using were outdated and were just a mere chat terminals.

The problem we wanted to solve was reduce the time taken for issue resolution and help chat agent be successful at closing a sale. The goal is to provide a chat application that is easier, efficient and helped agents and leads be successful and thereby CSAT.

I joined 24/7.ai as the lead designer for this product and 1 of the 2 designers in a company of over 80 engineers, 8 product managers & 10 data scientists.

Some of my key achievements -

• Facilitated design thinking workshops involving product managers, chat agents, leads, developers. This helped bring in multiple perspectives/ideas, motivations, fears and creativity.

• Hired and mentored user research interns.

• Improved usability by aligning product with user needs.

• Established a design library/system.

The final design radically changed the way chat terminals worked on work floors. We re-imagined IA and blended the chaotic work floor experiences with with chat, social feeds and gamification.

Process

Understanding the problem

We knew that using history data and AI models we can predict the intent of the customer or use historic data to resolve an issue faster. The key over here making sure the chat platform is efficient in delivering this information to the chat agent while making sure that the main job of the agent is conversing with the customer to gather information needed to serve the customer and improve CSAT which is the success parameter.

My research encompassed:

• Understanding the user goals & needs

• Uncover pain points with the existing user journey

• Determining success of the tasks measured

Gathering Insights

After collecting the recordings from the user interviews, identifying behavioral patterns by shadowing the chat agents and leads and cultural probes, I worked with the team to synthesize the pains & behavior. We grouped the problems and ideated on possible ways to solve them.

I relied on a data-driven approach known as the severity framework to inform my process and list of features in order of priority. This is calculated based on the following variables:

Task Criticality x Impact x Frequency of use

Wireframing the solution

Based on the above problems identified, I worked towards addressing these pains by coming up with potential solutions.

• Easy view of concurrent chats

• Easy access to pre-chat information (Any context of user intent, historic information)

• Easy access to intelligent suggestions to the chat agent.

• Efficient chat tools to improve speed letting them handle multiple customers at once.

Multiple versions of solutions were prototypes on axure and tested with users to validate and some of the other concepts that were tested with the users and didnt make it are linked below.

Concept 1

Concept 2

Concept 3

Delivering designs

I created my high fidelity mockups with detailed documentation, I worked closely with front end team to spec out any missing interaction. I was reviewed frontend tickets that were implemented to ensure they aligned with the designs.

Results & Takeaways

Since the product launched we have seen 45% increased sales conversions by empowering chat agents with great insights on who to target, when to engage and what to recommend. Customer support saw improved Average response time and CSAT as much as 73%, the Average Handling time reduced by 35% and reduced the resolution time by 83%.

Some takeaways form this project are :

• User testing doesn't have to end after development. Design is a constant iteration of improving the experience for the end user. Always find ways to collect and listen to your user's.

• Involve engineering upfront. This helps reduce rework, understanding technical feasibility and and its applicability can help your design perform better.

• Involve stakeholders in the design process where possible. This helps especially to get buy in & move faster, convincing is easier when they are involved in the process.