Lance Tan

People have
more in them.

I build AI products that help them do their best work.

Lance Tan smiling, wearing glasses and a blue shirt.

I’m Lance, an educator turned AI product manager. I ask what people and products could become, then focus on the changes that create the most value for the time, effort and resources we invest.

Let’s connect ↗
01 / SELECTED WORK

From Problem to Product

DATA ASSISTANT · SLS

Less time sorting.
More time understanding.

Helps teachers find themes and misconceptions in student responses, so they can decide what to teach next.

My role · User research, workflow design and delivery
The classroom challenge
THE PRODUCT CHALLENGE

The lesson doesn’t pause
for analysis.

Teachers need to understand student responses and decide what to teach next, while keeping the class on track.

Three demands. One live lesson.

Keep the class
on track.

Questions, behaviour, engagement

Understand
the responses.

Themes, gaps, misconceptions

Decide what
to teach next.

Revisit, probe, move on

Data Assistant helps surface the patterns. The teacher reviews them and decides what to do next.

LEARNING ASSISTANT · SLS

A thinking partner,
within the lesson.

An AI assistant that helps students work through questions, with guidance and reference materials configured by their teachers.

My role · Product direction, stakeholder alignment and rollout In use · 60,000 monthly conversations
02 / HOW I WORK

How I work

PRODUCT JUDGMENT

Make the trade-offs
visible.

I weigh quality, speed, cost and safety against the problem we’re trying to solve. Testing helps make those choices clear, including when to change direction or deprioritise an idea.

DECISIONS IN CONTEXT
When evidence changes the plan.Recommended deprioritising always-on facilitation.

AI facilitation for small-group student discussions.

THE PROMISE

Teachers were keen to try it.

WHAT TESTING SHOWED

Learning gains weren’t clear.

MY RECOMMENDATION

Deprioritise always-on facilitation.

The AI had to speak periodically, even when a discussion was going well. Student testing and independent chatlog review found inconsistent learning benefit, interruptions and frustration.

I recommended deprioritising always-on facilitation to the Product Owner and exploring less intrusive support. On-demand help was a direction to test next.

Read the full case study ↗
When safety has to fit the classroom.Protect students without overwhelming teachers.

Safety validation and teacher oversight for student-facing AI.

THE STANDARD

Generic safeguards weren’t enough.

THE TRADE-OFF

False alarms burden teachers.

THE APPROACH

Validate models. Escalate proportionately.

I built a validation pipeline using expert-annotated, de-identified conversations, with a second pass to filter likely false positives.

I designed teacher review and escalation around severity and repeated signals, balancing student protection with the risk of alert fatigue.

At the time of the case study, model benchmarking and operational rollout were still in progress.

Read the full case study ↗
Conference reflections · Palermo, 2025
RESEARCH & LEARNING

Bring research
into practice.

Research helps me question assumptions and make better product decisions. I attended AIED 2025 in Palermo, where I explored learning outcomes, trust in AI and how research translates into our own context.

DEVELOPING PEOPLE

Capability is
part of the product.

I help people build confidence through hands-on workshops, mentoring and opportunities to learn together. Understanding what people need shapes both the tools I build and how I help them use them.

Lance in a small-group discussion, seated with a laptop while another participant shares an idea.
Better tools start with understanding people.
03 / EXPLORATIONS

From the workbench

04 / ABOUT

Learning with others.

Some of my most fulfilling moments come from meeting people and being part of a community. I love how a conversation can spark an idea, and how that idea grows when others bring their perspectives to it. Events and shared experiences give me space for both: new possibilities and meaningful connections.

Lance with another attendee in front of the AI Engineer Singapore 2026 event backdrop.
AI Engineer Singapore, 2026.
Lance holding a microphone during a discussion at AIED 2025 in Palermo, Italy.
AIED 2025 · Palermo, Italy.
Lance with collaborators at the Gemini Hackathon with SystemEarth in 2026.
Gemini Hackathon with SystemEarth, 2026

Interested in the work
or have an idea to share?

Let’s connect