British gas Centrica

British Gas agent tool

3 months
London & remote
As part of a 12 week engagement with Centrica, we were tasked with improving the overall usability of their internal (Salesforce) tool used by service agents, with the aim of reducing average call times and repeat calls, while simultaneously increasing the first-call solution rate. As UX lead on this project, my role was to assist in the research process, derive key insights and then drive the high level design of a better tool, to then be handed over to Centrica’s SF consultants to be produced and shipped.
User research
UX design
Interaction design
Taxonomy
Service design
Project overview
Aims
A better facilitated customer support journey for Centrica agents.
Reduced repeat contacts and contact duration.
Increased end customer satisfaction.
Pains
Agents currently use multiple disparate tools, which creates high levels of inertia, slows down contacts and creates frustration for both agents and end customers.
Some tasks especially (e.g. identification and verification) are unnecessarily long.
Limiting factors
Limited functionality in Salesforce, not best practice.
Tools do not talk to each other.
Agents reportedly resistant to behaviour change.
Deliverables
Insights and recommendations for new functionality based on research.
Roadmap for Salesforce implementation of new features.
High level designs leveraging Salesforce Sales Cloud functionality.
Final
Designs
Big
Outcomes
What am I most proud of?
The
Process
Chapter 1
Opening scope
3 areas of focus
As part of this wider aim, our workstream was set up to specifically target the portion of the wider customer experience that did involve direct engagement with a service agent, by targeting the agent-side tool as a necessary component driving this experience. We scoped the following 3 agent journeys to focus on:
"
Our goal is to drastically simplify the customer support journey for Centrica agents. By identifying the best tool for the job at every step of the way and focusing intensely on how to make that tool work for the agent, we can free the agents to focus on what matters most – the customer.

When agents have a toolset that provides consistency, ease of use, and surfaces relevant information at the right time, they can serve the customer with empathy, ensuring Centrica customers are delivered the level of support they deserve.
"
The
Timeline
Week 1
Learn and empathise
Paragraph text
Gain deep understanding of the users and the challenges they face.
Identify specific journeys to focus on.
Create a shared understanding of success.
Map end-to-end process.
Identify enablers of and blockers to accelerate from here.
Week 3
Map user journeys
Paragraph text
Identify pain points to target on each journey.
Detail the current breakdown of tasks between ENSEK & SF, including overlap.
Identify the to-be breakdown of tasks.
Identify what information needs to be surfaced, when and in the best way.
Define success for each journey.
Week 10
Create prototypes
Paragraph text
Work with SF to create detailed wireframes that are testable and feasible within the SF construct.
Document the best way to implement in a manner that allows for component extensibility across SF for a consistent experience.
Week 10
Test and iterate
Paragraph text
Validate the to-be behaviour with real agents.
Identify further improvements, iterate, and re-test.
Week 12
Deliver designs and playbook
Paragraph text
Furnish a playbook to the business detailing how best to proceed should they decide to do so.
Foster a shared understanding of the benefits to be gained, considerations to keep in mind for implementation, and risks or challenges to be mitigated.
Chapter 2
Initial Research
6
Hours of silent observations
12
Moderated interviews
4
As-is journey maps created
Journey maps
We first mapped out the journeys for the key focus areas through a combination of user interviews, SME interviews and exploration of the Salesforce and ENSEK stools used by agents.
Pain point
Findings
There were multiple aspects of the current agent tool that created inertia and overcomplicated the process for agents, that extended beyond simply the journeys of billing or metering issues.
Recommended
Improvements
Our researched surfaced several key insights, but most significantly: within the scope of our 3 key journeys were crucial improvements that would have a far greater impact on the agents’ experience.

We approached the client with this conundrum. Ultimately, it wasn’t about simply finding clicks to reduce - the problems (and solutions) were far more entrenched. The findings of our researched were presented and they were well received, along with the following feature recommendations:
Chapter 3
Streamlined customer ID
69%
Decrease in time to completion
100%
Of agents preferred the new method
83%
Fewer clicks
We totally revamped the process of identifying and validating customers, which was the very first task that agents would do in every call. After hours of agent interviews and observations, we designed a whole new ID process based around the following significant features:
  • Salesforce would automatically detect and pull in any available customer information generated before the interaction began (e.g. entered through chat or in the pre-queue).
  • The agent could validate the customer through any number of fields, rather than just the Account Number which had to be obtained manually from ENSEK Ignition.
  • A completely reimagined customer details page that included previous cases and recommendations for next actions.
  • An issue overview widget that pulled in contextual information using Salesforce's new AI capabilities.
  1. Case opens in Salesforce.
  2. Switch to Ignition.
  3. Ask for customer details (either email address, postal address or account number).
  4. Enter details in Ignition and search.
  5. Locate consumer account number.
  6. Switch back to SF and enter account number.
  7. Assign customer to case.
  8. Add case details and continue.
  1. Contact tab opens in SF - prepopulated with any incoming information.
  2. Agent enters one or more pieces of customer information and clicks search.
  3. [optional] SF returns multiple results to choose from. Agent selects one.
  4. Confirm details with customer, view recommended actions.
  5. Decide on next action (e.g. open new case)
Key
Benefits
Chapter 4
Redesigning Cases
01
Simplified case structure
Over a period of 6 weeks (alongside designing other features) we formulated a brand new case and object structure. This feature formed potentially the largest overall revision to the tool.
Case structure
Before
Paragraph text
New cases would be opened automatically with every inbound customer contact.
Agents had to search for a customer in a separate platform (ENSEK Ignition), copy their details back into SF and assign the customer to the new case.
Cases sat at the upper tab level of a 2-tier tab system, with customer details (or sometimes even other cases) sitting below them on the lower level.
Non case-related customer information is found primarily under the case itself, duplicating information on customer details pages.
Tab structure
Home hub
Cases due today
Knowledge hub
Single case
ID
Cases due today
Knowledge hub
Single case
Customer details
Consumer account
Tabs in the tool sat across two levels high a high amount of repetition. Cases for example could exist as a top level tab with several other objects below it, or beneath other cases in a strange sort of infinite loop. This often led to confusing and highly inconsistent tab structures with no clear object hierarchy.
New cases would be opened automatically with every inbound customer contact.
Agents had to search for a customer in a separate platform (ENSEK Ignition), copy their details back into SF and assign the customer to the new case.
Cases sat at the upper tab level of a 2-tier tab system, with customer details (or sometimes even other cases) sitting below them on the lower level.
Non case-related customer information is found primarily under the case itself, duplicating information on customer details pages.
Case structure
After
Paragraph text
Inbound customer contacts first require the customer ID&V, after which the agent has the option to open a new case OR a recommended existing case.
Cases always sit in the lower level of the 2-tier tab system, under either a customer account or contact at the upper level.
Exiting a case tab or closing a case automatically prompts the agent to complete notes.
Cases are always opened against an action that would require one (even if it just taking a note), otherwise no case is created.
Tab restructure
Home hub
Cases due today
Knowledge hub
Incoming contact / Consumer wrapper
Consumer account / ID
Billing account(s)
Case
We updated cases to follow a new tab model. Cases no longer existed at the top level, now having to be tied to a specific consumer account. The new top tab was reimagined as the Incoming contact/Consumer account tab (the former for during a contact and the latter for outside of a call or message). Beneath this, there could only exist 3 possible objects: A consumer/customer, a service address (billing account) or a case.
02
New case creation
The creation of new cases was a huge overhead for agents, who had to complete 30+ fields in a single page form when opening any new case. This was problematic for a number of reasons:
Paragraph text
Numerous duplicated and redundant fields.
Large number of mandatory fields front-loaded in journey.
No logical chunking of form completion.
No smart pre-population.
Redesign v.1
So we redesigned the case creation journey to better guide agents through the process, removing duplicated fields and better grouping them into multiple pages after an in-depth, SME-driven analysis of the existing form.

The result was a much more manageable process that brought in existing information, either from the ID&V process or customer utterances, to pre-populate as many fields as possible.
Redesign v.2
On testing our new case creation journey gained some significant accolades. However, it became evident that it did not go far enough to cut down the overheads required of the agent at the most customer-centric part of the interaction.

We decided to make a huge pivot; to a barebones journey that would only require the agent to enter the bare minimum details to move forward, and allow them to complete the remaining fields later on, potentially once contact with the customer had ended.
03
A new place for case details
We moved most of things to do with a case away from a generic and repeating feature in the customer details page and into it's own kind of tab. Within this tab, we surfaced all the key functions agents needed to action any case.
Enablers:
  • View all the customer's cases (open and closed) in one place
  • See cases that are recommended to action
  • Open a new case
Enablers:
  • Update any details of a current case (automatically created via SF AI on creation)
  • Close case and add notes
Enablers:
  • View all notes created against this single case
  • Add a case note
Enablers:
  • See and sort all files attached to a single case
  • Upload a file to a case
The
Result
Chapter 5
Billing and metering
The task was to integrate British Gas's billing and metering tool - ENSEK Ignition, into Salesforce, using APIs to pull in key data and displaying each data point in the best possible way, in the right place, at the right time. Based on our research we developed the following 3 screens:
Enablers:
  • View the customers balance at a glance.
  • Check what product the customer is currently subscribed to.
  • Quickly take a payment.
  • Key task: create a new bill
Enablers:
  • View all of the customer's latest incoming and outgoing transactions.
  • View all payment and bill settings at a glance and update these easily.
  • Quickly review if the customer is able to afford their payments (direct debit review).
  • Key task: create a new bill
Enablers:
  • Allow the agent to see the customer's latest readings and all the most important details associated with each.
  • Show all of the customer's meters and the most important information for each one.
  • Key task: submit a meter reading
Leveraging
Quick cards
After dozens of hours of research to architect exactly what information agents needed, when and where - we decided on the use of a SF Lighting design system component known as 'quick cards.' These were highly flexible card-type components that enabled us to elegantly reorganise, categorise and display key data for agents.
Chapter 6
smart actions
Unpacking
The flow
We used the following three forms of research as sources of knowledge to create a ‘Next Best Actions map’ for the billing issues and meter readings journeys:
User driven
Data input - Utterance
What is the user saying coming into the call that we can use for early or even pre-call triage?
Gain deep understanding of the users and the challenges they face.
Identify specific journeys to focus on.
Create a shared understanding of success.
Map end-to-end process.
Identify enablers of and blockers to accelerate from here.
System driven
Data input - Agent
What is the user saying coming into the call that we can use for early triage?
Gain deep understanding of the users and the challenges they face.
Identify specific journeys to focus on.
Create a shared understanding of success.
Map end-to-end process.
Identify enablers of and blockers to accelerate from here.
Paragraph text
Customer details
Previous/recent cases
Latest bill reading
Current balance
Agent driven
Triage
What gaps does the agent have to close to refine the potential solution?
Gain deep understanding of the users and the challenges they face.
Identify specific journeys to focus on.
Create a shared understanding of success.
Map end-to-end process.
Identify enablers of and blockers to accelerate from here.
System driven
Identify problem
What problem can we then ascertain from the combination of?
Gain deep understanding of the users and the challenges they face.
Identify specific journeys to focus on.
Create a shared understanding of success.
Map end-to-end process.
Identify enablers of and blockers to accelerate from here.
System driven
Propose
What is the solution directly derived (usually 1:1) from the problem?
Gain deep understanding of the users and the challenges they face.
Identify specific journeys to focus on.
Create a shared understanding of success.
Map end-to-end process.
Identify enablers of and blockers to accelerate from here.
Agent driven
Solve
What now needs to be done to action this on the agent’s side?
Gain deep understanding of the users and the challenges they face.
Identify specific journeys to focus on.
Create a shared understanding of success.
Map end-to-end process.
Identify enablers of and blockers to accelerate from here.
Magic
Moment
To better understand the Triage step of this [map], we mapped out a full decision diagram to represent the manual ways agents would reach a solution to a customer’s problem, based on every input they might receive. This exercise had never been done in British Gas before and took over 20 hours of interviews with agents and SMEs, but proved invaluable.
A Next Best Action for
Everything
The result of all this was the blueprint for a much improved NBA sidebar that would go on to enrich the overall agent experience, not only for the billing and metering journeys, but throughout all tasks.Other examples of NBAs we introduced included:
Chapter 7
Case closure notes
Workflow
Pains
A major pain point we encountered in our research that was not foreseen in our brief surrounded the case workflow. Cases would occasionally be opened by one agent but picked up by another, who would have little context besides the notes they were given.
Design
Changes
Based on this we updated the design of the opening and closing case journey to work in the following ways:
Chapter 8
final testing
The
Structure
After our design phase, we went on to conduct 2 modes of research & testing:
SME reviews with agent team leads
Moderated testing with agents
SME reviews
We embarked on a series of review sessions with the aim of finessing the information structure of our key components, cards and pages to ensure we were displaying the correct data points where agents needed them most. This involved over 30 hours of review sessions, presenting visual reference points either in a 1-on-1 or focus group settings.
Moderated testing
We also conducted 6 x 40 min testing sessions, each asking agents to complete 1-2 tasks in a mid-fidelity prototype. The aim was to test the true impact of our design on those who will be using them, agents, with the aim of getting accurate and actionable feedback and thus make improvements.
The result was a refinement of:
Paragraph text
A new case creation journey
Core information architecture
Chapter 9
Impact and feedback
69%
Decrease in time to completion
100%
Of agents preferred the new method
83%
Fewer clicks
"
Andrew’s contributions were critical to the success of an important initiative with Centrica. His expertise in UX and User Research ensured our designs met the needs of the end user and client goals. The project had significant challenges as we had to balance the constraints of Salesforce, with reported desires of call centre staff, needs of the end customer and the business goals of our key stakeholders.

Andrew demonstrated strong leadership skills, leading client presentations and building their trust in our ability to provide impactful solutions. He also provided mentorship and guidance to more junior designers, creating a supportive team environment.

His proficiency in wireframing and prototyping enhanced our project deliverables, ensuring they were both feasible and functional. Andrew will approach any project at CI&T with positivity and a solution led attitude.

Andrew's collaborative nature and commitment to delivering great work make him a valuable team member. I look forward to working with him again the future.
"