designing ai tools for care agents supporting customers
Understanding user behaviours and pain points to inform how the latest AI technology could support business and customer needs
Research context - What was the problem we were trying to solve
OVO, an energy supplier with over 4 million customers, had a huge and increasing demand for call agents to support customers with their needs.
Due to increases in the cost of electricity, customers reaching out for help had increased 40% year on year, leading to increased agent work loads and poor customer satisfaction.
OVO wanted to explore how they could use artificial intelligence tools (AI) to save agents time, support more customers and improve their satisfaction.
Key challenges where research could add the most value
A solution-first approach - A decision was made to use an AI solution without fully understanding the needs and paint points of the agents involved.
An agency brought on board with little context - without much knowledge of the challenges agents faced, there was a risk they would building a solution that didn’t support agent or business needs.
A tight delivery timeline - Project timelines were fast and furious!
My role
User research lead on this project
Responsible for identifying how user research could inform OVO’s approach
Responsible for articulating research goals and research questions to answer
Responsible for creating a research plan, conducting research and synthesising key insights back to the team
My approach
User research as a team sport - involving the whole team in user research so there was a shared understanding of agents’ needs and behaviours.
Support key decision making - By creating easy to understand artifacts explaining user needs and pain points.
Test and learn - Given the tight timelines it was important to be able to test design direction quickly and effectively.
Methodology
Ran workshops with teams to identify research questions and riskiest assumptions to explore
Organised shadowing sessions with 20 care agents, remote and in person
Highlighted key needs of agents through reports and playback sessions
Created research artifacts such as an empathy map and a ‘jobs to be done’ map
Conducted usability testing sessions to test out tool prototype with agents
Research goals identified
Understand how care agents support our customers
Understand how they find the information they need to support customers
Understand how well their knowledge repository supports them
Identify key pain points
Outputs and artefacts
Playback report identifying key insights - to articulate what tasks agents need the most support with
Empathy map - to give the whole team a clearer understanding of agents’ behaviours, the tools they used and their biggest pain points
‘Jobs To Be Done’ map - to articulate to the team the key activities agents performed and which of them were being well served or poorly served
Methodology
Ran workshops with teams to identify research questions and riskiest assumptions to explore
Organised shadowing sessions with 20 care agents, remote and in person
Highlighted key needs of agents through reports and playback sessions
Created research artifacts such as an empathy map and a ‘jobs to be done’ map
Conducted usability testing sessions to test out tool prototype with agents
Key insights
Agents were unclear how and where to pass customers onto specialist teams for additional support
There is no single source of information agents can rely on
It’s hard to find correct and accurate information in knowledge repository
Impact
The team used insights to inform how the AI tool would support agents, prioritising allocating cases correctly and highlighting information most valuable to customers’ enquiries.
This led to a 20% uplift in correctly allocated customer cases, a huge win for the business and customers.
Shared knowledge of key agent pain points led to significantly influencing team’s future priorities.
The results
The team used insights to inform how the AI tool would support agents, prioritising allocating cases correctly and highlighting information most valuable to customers’ enquiries.
This led to a 20% uplift in correctly allocated customer cases, a huge win for the business and customers.
Shared knowledge of key agent pain points led to significantly influencing team’s future priorities.