Talent Builder
Product strategy and GenAI
for global workforce planning
TL;DR
How I owned product strategy, identified AI opportunities through research, and shipped the first GenAI modules into BCG's flagship SaaS platform for talent management, used by the United Nations and global enterprises.
THE PRODUCT
A platform built for the complexity of global talent management
Talent Builder is BCG's premium SaaS product for talent strategy and workforce planning, relied on by global enterprises. It helps HR and business leaders forecast, plan, and optimise talent at scale: covering supply-demand analysis, skill gap ratios, job readiness assessments, and workforce optimisation across large, distributed organisations.
Used by major organisations including the United Nations and leading health-tech companies, the platform sits at the intersection of workforce data, compliance requirements, and strategic planning. It needed to be industry and country agnostic, working equally well for a UN HR compliance lead in Geneva and a workforce analyst inside a multinational health system.
When I joined the team, the product was functionally solid but analytically slow. Clients had the data; they didn't have the insight. Getting from raw workforce inputs to an actionable decision required significant manual effort and, frequently, a BCG consultant to interpret the output. That gap was what the team set out to close.
BCG set a goal to elevate Talent Builder from a consultant-driven tool into a self-reliant, AI-powered platform clients could use entirely independently.
MY ROLE
Product strategy, AI design, and research ownership
I owned product strategy end-to-end: synthesising user research, usage logs, market signals, and AI model metrics to shape the roadmap, and then working across a cross-functional team of BCG consultants, product leads, AI engineers, and a vendor dev team to deliver it. My background spanning UX, cognitive psychology, and product gave me the range to move between research, strategy, and detailed design decisions without losing thread.
This was consultancy-paced delivery: high stakes, multiple senior stakeholders, short cycles, and decisions that needed to be evidence-led to stick. I ran sprint planning, facilitated cross-functional critiques, and presented directly to executive stakeholders, while also doing the hands-on research and prototyping that grounded those decisions in reality.
THE PROBLEM
What the research surfaced
The problems were not obvious from the outside. Talent Builder looked capable on a demo. It was only when we dug into usage logs, ran cognitive walkthroughs with HR leads, and mapped the actual workflow end-to-end that the friction points became clear.
APPROACH
How we shaped the roadmap and built toward it
The strategic direction came out of a research and synthesis phase the team ran together. We looked at user feedback, session logs, market trends in HRIS and AI-native SaaS, and model performance metrics side by side. What emerged shaped a clear set of priorities that we then committed to the roadmap and delivered in focused cycles.
01
Market research and competitive benchmarking
Ran consultancy-grade competitive analysis across HRIS platforms, workforce planning tools, and AI-native SaaS products. Combined with the synthesis of user feedback, usage logs, and AI model metrics to map where Talent Builder had real differentiation opportunities and where it was falling behind. This became the evidence base for our roadmap prioritisation conversations with product leads and senior stakeholders.
02
GenAI UX playbook and AI opportunity framework
Before any feature work, I built a playbook to help the team reason about AI properly: start from a real user problem, assess whether AI is the right lever, and define what success looks like before building. Using intent modelling and cognitive walkthroughs with HR leads, I identified 7 specific LLM opportunities embedded in genuine workflow friction, covering areas like workforce simulation, summarisation of complex forecasts, and skill-gap pattern recognition. These became the foundation of our GenAI strategy and shaped what went into the roadmap.
03
Defining requirements and acceptance criteria for AI modules
Worked directly with AI engineering to translate strategy into buildable specs. Defined requirements and acceptance criteria for three core AI modules: workforce scenario simulation, AI-powered summarisation of forecast outputs, and skill-gap analysis with smart recommendations. The criteria covered model behaviour, output explainability, edge cases, and the thresholds at which AI suggestions should surface versus stay hidden.
04
Dashboarding, insight mapping, and smarter recommendations
Redesigned the core forecasting and analytics surfaces to reduce manual effort and get clients to decisions faster. This meant rethinking data hierarchy across supply-demand views, skill gap dashboards, and job readiness panels: surfacing the most decision-relevant signal upfront, using progressive disclosure for depth, and embedding AI-generated recommendations at the right point in the workflow rather than as a separate "AI tab." The result was faster analysis with less cognitive load.
05
A/B testing, usability, and analytics validation framework
Built a repeatable validation process combining A/B testing, moderated usability sessions, cognitive walkthroughs, and product analytics. This was not a one-time test plan: it became the standard process by which all AI feature decisions were evaluated and iterated. The framework drove a 21% improvement in mobile task success rate and gave the team a shared language for what constituted a validated improvement versus an assumption.
06
Sprint planning, exec presentations, and cross-functional critiques
Ran sprint planning with the vendor dev team and facilitated weekly cross-functional critiques across BCG consultants, product leads, and AI engineers. Presented strategy and design decisions to executive stakeholders, structuring each presentation around evidence rather than preference. Scaling insight adoption by 40% came in large part from how decisions were framed and socialised across a complex, multi-stakeholder organisation.
07
Scalable design system for global, context-agnostic delivery
Built a cohesive component library and design system that worked across client industries and regulatory contexts without customisation debt. Every pattern had to function for the UN's compliance-heavy skill taxonomy and a health-tech enterprise's workforce optimisation workflow using the same core language. This cut QA cycles and design rework significantly, and made onboarding new enterprise clients a repeatable process rather than a ground-up effort each time.
IMPACT
What shipped and what it moved
The strategic direction came out of a research and synthesis phase the team ran together. We looked at user feedback, session logs, market trends in HRIS and AI-native SaaS, and model performance metrics side by side. What emerged shaped a clear set of priorities that we then committed to the roadmap and delivered in focused cycles.
Shipped the first GenAI-powered modules within Talent Builder, including AI-driven workforce simulation, forecast summarisation, and skill-gap recommendations. Established a repeatable AI validation model the team continues to use across new product lines.
LEARNINGS
What this project reinforced
-
A principled AI framework is itself a product deliverable. The GenAI playbook shaped how the whole team evaluated ideas. Getting alignment on the right questions to ask before building was more valuable than any individual feature we shipped.
-
In consultancy delivery, evidence is currency. Working across BCG consultants, vendor engineers, and UN stakeholders meant product decisions had to be communicated differently for each audience. Structuring critiques and exec presentations as evidence-led, not opinion-led, was what made them land and stay on the roadmap.
-
Validation frameworks compound over time. Building a repeatable A/B and usability process meant every subsequent feature had a clear standard for success. The 21% mobile task improvement was not a one-off: it was the first proof point of a system the team kept running.
-
Global agnosticism is a design constraint, not a disclaimer. Designing for clients across industries and countries forced cleaner abstraction at every level: components, content patterns, AI output framing. That rigour made the platform genuinely more robust, not just more portable.
Problem -> Ideation -> A number of scrappy wireframes
