Back

Back

Sky’s AI search integration

Integrating Gen AI tool with agent search was designed to reduce average call handling times for call centre executives, improving efficiency and lowering operational costs.

Integrating Gen AI tool with agent search was designed to reduce average call handling times for call centre executives, improving efficiency and lowering operational costs.

Role

Role

Product Designer

Approach

Approach

Design sprint

Tools

Tools

Figma, Miro

THE WHAT

KOSMO is an AI-powered search tool designed to assist call centre executives in efficiently addressing customer queries and reducing average handle times. The primary goal was to lower operational costs by improving response efficiency.

THE HOW

To ensure smooth progress, I built trust with key stakeholders by conducting design sprints and testing sessions with call center experts. I also organized playback sessions for executives, allowing for transparent feedback and alignment. This collaborative approach helped set clear deliverables and meet the timeline without sacrificing quality.

THE CHALLENGE

One major challenge I faced was during the development of KOSMO, an AI-powered tool integrated with the call center search bar. The goal was to reduce the average handle time of customer calls, ultimately lowering costs. The project had high visibility, as it was overseen by the CTO, which placed additional pressure on meeting tight deadlines.

KEY TAKEAWAY

Simplicity is crucial when managing complex projects under tight deadlines. Quick iterations allowed us to gather valuable user feedback early, enabling us to refine and improve the solution within the time constraints.

THE IMPACT

$1.4 Million annual cost savings

We were able to measure 14 sec overall Average Handle Time (AHT) savings from KOSMO, this equates to $1.3Million annual cost savings

Gen AI research (Workshop)

The deadlines were extremely crucial for the project, so it was important to set the expectations for the project and also involve end-users from the beginning for a better output

Gen AI design sprint workshop

With stringent deadlines, I borrowed a page from the Design sprint and facilitated a 2-day workshop for my Gen AI project.

View case study

Early wireframe ideations

I created low-fidelity wireframes to quickly visualise and communicate my ideas for the AI project. This approach allowed me to gather feedback from stakeholders more efficiently, given the tight deadlines, It helped me align the team early on to key design concepts and project direction. By sharing the wireframes, I was able to prioritise features and streamine discussions for faster decision making

KOSMO search results - Maximum 2

KOSMO search results - Maximum 3

Loading Animation

Error results

KOSMO - Search results

The project involved redesigning an existing landing page where Sky Broadband customers land after searching for product-related issues on Google. The main objective was to enhance the page’s aesthetics and improve user engagement by creating a more intuitive and visually appealing layout.

How to use prompt

This section offers clear guidance for call centre executives on interacting with the AI and effectively framing prompts. The design priorities simplicity, ensuring that even those with limited technical experience can easily utilise the tool while assisting customers in real-time.

Generated answers

The decision was made to limit the number of visible responses on the screen to minimise decision fatigue for users. The top answer is prioritised and prominently highlighted, while other relevant answers are organised into tabs. Additional related articles, identified by KOSMO’s algorithm, are displayed as links within a closed accordion, which users can expand as needed.

The articles visible only to experts are marked with a headphone icon, indicating to users which article they are referring to.

The decision was made to limit the number of visible responses on the screen to minimise decision fatigue for users. The top answer is prioritised and prominently highlighted, while other relevant answers are organised into tabs. Additional related articles, identified by KOSMO’s algorithm, are displayed as links within a closed accordion, which users can expand as needed.

The articles visible only to experts are marked with a headphone icon, indicating to users which article they are referring to.

Other relevant article section

We intentionally limited the number of relevant articles to four. Through observations of call centre executives, we recognised that minimising customer handling time was essential. Considering their typical experience, executives could quickly identify articles relevant to the queries they were addressing. This led us to restrict the visible articles to four, with the option to access more if needed, presented in an expandable accordion.

We intentionally limited the number of relevant articles to four. Through observations of call centre executives, we recognised that minimising customer handling time was essential. Considering their typical experience, executives could quickly identify articles relevant to the queries they were addressing. This led us to restrict the visible articles to four, with the option to access more if needed, presented in an expandable accordion.

Open customer view

In the existing experience, the executives could open the customer view for the help articles and guide them effectively as to where they need to focus. This created a re-assurance with customers that, the help they were looking for was available to them.

Final Screens

  • I created low-fidelity wireframes to quickly visualise and communicate my ideas for the AI project.

  • I created low-fidelity wireframes to quickly visualise and communicate my ideas for the AI project.

  • This approach allowed me to gather feedback from stakeholders more efficiently, given the tight deadlines.

  • It helped me align the team early on key design concepts and project direction.

  • By sharing the wireframes, I was able to prioritise features and streamline discussions for faster decision-making.`

  • This approach allowed me to gather feedback from stakeholders more efficiently, given the tight deadlines.`

  • It helped me align the team early on key design concepts and project direction.

  • By sharing the wireframes, I was able to prioritise features and streamline discussions for faster decision-making.`

KOSMO - How to use prompt

Search bar integrated with KOSMO

Multiple steps articles

Feedback and validation