People Skills

At a glance

Why now

Skills are the new currency

Organizations are shifting from job-based to skill-based talent models. A job title says what someone was hired to do; skills say what they can actually contribute, including strengths their title hides. That shift is what made People Skills worth building, and it set the bar for the design: skills had to show up as an ingredient in the tools people already use, not as another destination to remember.

Who this is for

Three user roles, fundamentally different needs

Ecosystem touchpoints

How skills flow across Microsoft

Design process

Three challenges at the intersection of AI, trust, and enterprise

Designing AI-inferred skills across Microsoft's ecosystem required solving challenges at the intersection of AI, user trust, and enterprise compliance.

Here is how those challenges played out in practice, on one surface, from brainstorm to ship.

Deep dive

Skills in the People Agent

Discovery

What makes a skills response worth trusting?

An AI answer about a person's skills is only useful if people believe it and can act on it. I ran a structured brainstorm across three intent areas (my skills, insights, and collaboration), then co-designed response requirements with cross-functional partners. Three requirements emerged. Responses must be engaging, using personalized data and natural language. They must be accurate, matched to real intent and rendered with the right UI elements. And they must be actionable, always mobilizing the user to a clear next step. The feedback also got specific in ways that directly shaped the design: keep suggestions to a short, tailored list so advancement feels achievable; tie every recommended skill back to business impact; split guidance by career track and by short, mid, and long-term horizons; and use multi-turn conversation to clarify intent before recommending.

"I'd like a short list of skills tailored around my areas of interest or strengths. This makes advancement seem achievable."

Exploration

Finding the hero prompt

I explored several iterations of prompts and response layouts, testing which entry point made skills feel immediately useful rather than abstract. The design converged on a single hero prompt: 'Who can help me with [skill/task]?' It won because it starts from a real moment of need, produces an answer with obvious value, and quietly teaches people that skills data exists and works.

The Shift

From standalone agent to ingredient

Midway through, the direction changed: rather than shipping Skills as its own standalone agent, the organization consolidated capabilities into the People Agent. The reasoning was sound, since a separate agent competed for attention with every other surface and asked people to remember it existed. But it meant rethinking the design, not just relocating it. My job was making skills feel native to its new home rather than bolted on. I carried the hero prompt, the trust patterns, and the skills snapshot into the People Agent so the intelligence stayed intact while discovery improved, because the answer now appears in a flow users were already in. It also hardened a principle the whole system follows: meet users where they already are instead of building a new place for them to visit.

Detailing

Getting down to the details

With the hero prompt defined, I designed how else skills communicate inside the People Agent, providing skill-based insights and collaboration recommendations. After iterating on multiple approaches, the final design prominently features the hero prompt, adds discovery prompts for insights and collaboration, and surfaces a streamlined snapshot of the user's own skills profile, with confirm and delete controls one tap away and the full profile card one click deeper.

Systematizing

Built to scale past MVP

Once the direction was aligned, I componentized the design against Microsoft's Fluent tokens: the prompt field states, the discovery prompt cards, and the skills snapshot. That way, post-MVP changes would be design-system updates rather than redesigns. In parallel, I compressed a complex web of admin and user scenarios into simple, plain-language controls.

Craft

Accessible, one focus point at a time

Accessibility was not a final pass. Every state was annotated with focus order, keeping the experience fully navigable by keyboard and legible to screen readers, from the search field through discovery prompts to each skill chip's confirm and delete actions.

Shipping It

What it took to ship

Getting this out the door meant less time in the canvas and more time creating the conditions for six product teams to ship one coherent experience. For the surfaces I did not personally design, I set the quality bar the work had to meet, from component states to focus-order annotations, so partner teams could build to spec without guessing.

Solution

The deep dive followed one surface. This is the full set that shipped across the ecosystem.

Key deliverables

A comprehensive design framework and component library that enables consistent, privacy-forward skills experiences across all Microsoft products.

Decisions and tradeoffs

What we chose, and what we protected

Every deliverable above carries a judgment call about intelligence and trust. These were the two that mattered most.

Impact

What shipped and where it landed

A consistent, trusted, and scalable skills experience now lives across Microsoft's productivity ecosystem. Below is what shipped and where it landed.

Specific adoption percentages, profile-completion rates, and tenant counts available on request under NDA.

Looking ahead

What I'd do next