Pain № 01
Information asymmetry.
Coaches struggled with roles where they lacked domain expertise. They couldn't provide the context clients needed for unfamiliar industries or occupations.
Enterprise · GovTech · 2024
CC4 Rework.
A 12-to-16-month rework of Workforce Singapore's career coaching platform, co-created with coaches, powered by AI, and validated through quarterly usability cycles.
The Brief
01A platform for the coaches.
CC4.0, a collaboration between WSG and GovTech, serves as a platform for career coaches, offering tailored career matching services for job seekers at different career stages. It needed a major rework to meet increasing needs.
“We launched a major rework that included 2 new core features: the Role Informer Tool and the 360 Report, built through co-creation with coaches and validated through quarterly usability cycles.
70%
Case completion improvement
3.3→4.4
Ease of case creation rating
60+
Variables in Employability Index
Q3
Usability testing every 3 months
The Problem
02Coaches were drowning in admin.
Career coaches struggled with tedious workflows, information asymmetry, and tools that didn't reflect how they actually worked. The platform needed to evolve from a data-entry system into a genuine coaching assistant.
How might we rework the CC4 platform so that career coaches can spend less time on administrative tasks and more time helping job seekers find meaningful employment?
Pain № 02
Tedious data entry.
Manual data entry consumed time that should have been spent coaching. Multiple systems required duplicate entries with no data extraction between them.
Pain № 03
No task management.
Coaches managed multiple clients simultaneously with no overview of their workload. Ensuring follow-through on the case management process was a constant challenge.
Pain № 04
No actionable insights.
The platform tracked cases but didn't help coaches. There were no data-driven recommendations, no predictive models, no way to diagnose a client's employability.
Feature 01 · Role Informer Tool
Research any role in seconds.
Coaches were often struggling with jobs where they lack domain expertise and lacked the context the clients needed. We solved this by creating a tool to combine SERP API and leveraged the capabilities of an LLM to help coaches reduce the efforts required to research roles they are unfamiliar with.
- ALLM + SERP API tool chain. One of the ways we were able to use LLMs to provide up-to-date, real-time job searches is by combining it with SERP API (Google's search API). This allowed us to leverage both the language capabilities of LLMs without the risks and issues that commonly come with it.
- BRole context. Salary ranges, required skills, industry trends, and adjacent opportunities, all surfaced in one view for coaches to brief themselves before a session.
- CWhite-box explanations. Transparent, auditable recommendations that coaches can trust and share with clients. Every insight is traceable to its data source.
Impact Through conversations with career coaches from NTU and SGEnable, leads were generated to onboard the Role Informer Tool. Coaches showed enthusiasm for the AI enhancements.
Feature 02 · Task Management
Never lose track of a follow-up.
Coaches often shared their challenge of ensuring they are able to follow through the case management process as they have to deal with multiple clients at once. To help them with their work, we introduced two key features: Upcoming Tasks and an updated Banner System.
- AUpcoming Tasks. Allowed coaches to have an overview of all their tasks the moment they log into the CC4 platform. We also created a direct entry point for them to access the page they need to complete their task.
- BBanner system. Banners were employed to aid coaches in task clarity and data input without undue mental effort. Their efficacy was assessed through A/B testing, followed by iterative design improvements informed by feedback. Subsequent unmoderated usability testing on Maze further refined banner effectiveness.
- CWorkflow integrations. We integrated external systems such as FormSG and Appointment Booking System (ABS) which allowed clients to self-serve and self-assess, and created ways to extract data from other sources to reduce data entry work.
Impact We saw an overall improvement in the completion of cases, with one organisation going from a 30% completion to 70%. The ease of case creation received a rating boost from 3.3 to 4.4 by Career Coaches.
Feature 03 · Employability Index
Predict employment in 90 days.
A significant outcome of the 360 Report was the development of the Employability Index, a data science initiative. Our team constructed a predictive model capable of determining the likelihood of a job seeker securing employment within 90 days, while also dissecting the contributing factors behind this outcome.
- A60+ variables. Leveraging data provided by Career Coaches, the model incorporates over 60 variables to generate insights, facilitating more accurate client diagnoses.
- BContributing factors. The model does not just predict outcomes. It dissects what drives them, giving coaches a clear view of which factors are within their client's control.
- CWhite-box recommendations. By kick-starting these data projects, we are able to provide a white-box explanation to coaches on the insights and recommendations that are generated from the data.
Insight Having a tool set that we can use to act as the "brain" to handle cognitive tasks, analytics and data transformation allowed us to produce more confident recommendations. We can then use that data and feed it into LLMs as few-shot prompts to generate recommendations.
Feature 04 · Industry & Occupation Insights
Data-driven coaching.
The data we use to create the Employability Index, while useful, is skewed as it is limited to the pool of clients that comes to WSG and E2i for coaching. We explored how we can use data sets from data.gov and mom.stat to discover insights on the industry to better inform coaches.
- AEmployment trends. We were able to look at data such as employment change from public data sources, surfacing industry-level insights coaches could not access before.
- BLLM + analytics fusion. Combining LLM capabilities with data analytics to generate coaching recommendations. Data handles cognitive tasks, analytics and data transformation; LLMs generate natural-language recommendations from those signals.
- CTransparent recommendations. Every suggestion is traceable to data. Coaches can explain the "why" behind every recommendation to their clients.
Insight The Employability Index data is limited to WSG and E2i clients. Public data sources from data.gov and mom.stat add crucial industry-level context that balances this skew.
The Process
04Three phases of a rework.
We started with the coaches, not with screens. Co-creation workshops, atomic research, and a consolidated IA gave us the blueprint for a 12-to-16-month execution roadmap.
Co-creation: designing with coaches, not for them
To gain valuable insights, we engaged in co-creation. This involved conducting a co-design workshop where we actively involved Career Coaches in discussions. Through various sessions, we delved into their pain points, fostering collaboration to explore potential ideas together.
I was able to derive invaluable insights by having the coaches design what they think they want their own dashboard to look like. We broke them into groups, printed out wireframes, drew and made notes and then had the groups share their ideas with each other. We then let them vote for the different designs and were able to see what issues resonated with the majority of the career coaches.
Atomic research: a repository that lasts
We used the atomic research framework to store our learnings in our research repository. By coding our research, we were able to construct a research dashboard for us to access our learnings easily. Through this process of co-creation and documentation, we were able to store what we learned and access those insights easily. This was vital as some of these improvements ended up taking about 12 to 16 months for us to implement.
From sketch to north star
While the ideas that coaches had were useful, it was difficult to fit all of it into our UI. Instead, we extracted core themes in their "artwork" and used that to inform our decision to create a consolidated IA. This showed us what information was commonly used by coaches. The consolidated IA created a framework for us to think about what needed changing.
Based on these recommendations, I created a new design that matched what the coaches wanted. This redesign helped to act as the north star to the rework that would span the next 12 to 16 months. I presented these concepts to the group director of WSG and we were able to get buy-in and renew our project for another 2 years.
From concept → execution
Three pillars, one vision.
01
Intelligence · AI + Data
Solve the information asymmetry.
Combine SERP API and LLM capabilities to help coaches research unfamiliar roles in seconds, providing up-to-date, real-time job insights without the risks of raw LLM output.
02
Workflow · Integration
Streamline the workflow.
Integrate external systems (FormSG, ABS) for client self-service. Extract data from existing sources to eliminate duplicate entry. Reduce cognitive load at every touchpoint.
03
Insights · Analytics
Deliver actionable insights.
Build the Employability Index: a predictive model using 60+ variables to determine likelihood of employment within 90 days. Turn data into coaching recommendations.
The Impact
05A year of research, validated.
Over the span of a year, extensive research and execution processes were undertaken. Usability tests were conducted every three months, and numerous sessions were held with Centre Management, coaches, directors, and other stakeholders to ensure alignment with the product's direction.
30%→70%
Case completion improvement
3.3→4.4
Ease of case creation rating
60+
Variables in Employability Index
Q3
Usability testing every 3 months
The final tested version exhibited notable improvements, with the ease of case creation receiving a rating boost from 3.3 to 4.4 by Career Coaches. Coaches showed enthusiasm for the AI and data enhancements.
Through conversations with career coaches from NTU and SGEnable, leads were generated to onboard the Role Informer Tool. The ongoing project aims to enhance support for coaches by utilising data and AI capabilities. The next significant development is the introduction of the Occupation and Industry Insights feature, designed to offer coaches valuable perspectives on various occupations and industries.