Commercial generalistAI-native operator
MEL [AU]
Useful before the role has a clean name.

Put me somewhere unfamiliar. I’ll find the signal.

Heidi taught me to follow usage before opinion. Lembra taught me to turn a question into something people can touch. I’m still early. The useful part is how much more capable I can become between one unfamiliar room and the next.

Read the chapters How I workTalk shop
01 / Heidi Health · 2022—25
Intern to business development; saw activation, drop-off and retention inform the week’s priorities.
02 / Lembra · 2025—now
Customer discovery and AI-assisted prototypes; a shorter path from question to working first version.
03 / Working range
Product signals, commercial research and technical translation across unfamiliar terrain.
Working thesis/00

The stack changes. The habit is to notice, learn and become useful while everyone else is still discussing titles.

I’m a commercially minded generalist from Samoa, now in Melbourne. My range covers early GTM, product-led growth, customer discovery, commercial research and enough AI-native building to get a credible first version out of my head.

I don’t confuse range with mastery. The point is not to impersonate every specialist. It is to take unfamiliar terrain from fog to a sensible first move, then recognise when the work needs someone deeper.

Two chapters/01—02

Where the learning curve became the work.

Not a victory lap. A record of what changed in how I operate.

01HEIDI HEALTH · 2022—25

Inside an early PLG motion

Learn the product by watching people use it.

I joined Heidi’s GTM team as an intern and moved into business development. I did not build or own its product-led growth system. I did get a close look at how an early team used it.

Week by week, product behaviour mattered: where users arrived, where they stalled, whether they returned and what those patterns meant for the next priority. PostHog was not a monthly ornament. It was part of the operating conversation.

“Data can locate the bruise. Customers usually tell you how it happened.”

What stayed with me

Observe
Activation, drop-off, retention and repeated use.
Translate
Move between product behaviour and customer language.
Prioritise
Let real behaviour interrupt a comfortable internal story.
02LEMBRA · 2025—NOW

AI prototypes, zero to one

Then the distance from question to prototype collapsed.

By Lembra, agentic tools had changed the economics of being a generalist. I was not suddenly an engineer. I could learn enough to turn an idea into a testable product, deploy it, inspect what broke and bring it into a customer conversation.

That learning curve became part of the work: Claude Code, Codex, Cursor and Grok for building and debugging; Vercel for shipping; Braintrust and Sentry for inspection; Linear for keeping the next move legible.

“The receipt is not a perfect product. It is that I can now build the first version myself.”

What changed

Before
A good product question still needed a long queue.
Now
I can make the question tangible before the meeting.
Still true
Tools do not replace customer judgment or product taste.
Operating loop/03

How I get useful when the map is bad.

“Fast learner” is cheap copy. The sequence below is the thing that has to be repeatable.

  1. 01

    Map the terrain

    Start with the product, the public record and the people living with the problem. Separate fact, assumption and blank space before forming a view.

  2. 02

    Find the live signal

    Look at behaviour and listen to customers. A dashboard can show the bruise. A conversation usually explains how it happened.

  3. 03

    Build the smallest proof

    Turn the question into something testable: a prototype, a message, a workflow or a tightly bounded commercial experiment.

  4. 04

    Change the belief

    Decide what the evidence earns. Keep the useful idea, kill the decorative one and write down what the second attempt should do better.

Working toolkit/04

What’s on the bench right now.

Working fluency, attached to a job. No badge wall. No claim that installing the software made me wise.

Build & debug

Claude CodeCodexCursorGrok

From product question to working first version, with enough visibility to understand what the agents changed and why.

Ship & coordinate

VercelLinear

Put work somewhere real, keep decisions legible and make the next bet smaller than the last meeting suggested.

Observe & improve

PostHogSentryBraintrust

See what people do, what the product breaks and where an AI system is producing confidence instead of truth.

The stack will change again. The durable part is knowing what to ask, which signal deserves attention and when the machine is bluffing.

Independent field notes/05

I learn by making the question smaller.

Outside-in studies built from public evidence. No client access. No borrowed outcomes. That distinction matters.

A

Sovereign AI adoption

Mapped a public product and buyer surface, then reduced a broad commercial ambition to one workflow, one buyer and a 90-day pass-or-stop test.

RESEARCH / ADOPTION / EXPERIMENT DESIGN
B

Document AI commercial system

Separated category noise from defensible product proof, then built a practical ICP, evidence hierarchy and sequence of market tests.

POSITIONING / ICP / PROOF ARCHITECTURE
C

Ambient AI growth reconstruction

Rebuilt the path from product-led adoption to larger deployment using public evidence, with sourced facts, assumptions and unknowns kept visibly apart.

PLG / MARKET EVIDENCE / GTM SYSTEMS
Operating range/06

A route, not a ladder.

Samoa → Auckland → Melbourne. Five stops, each changing the next.

  1. Origin

    Samoa

    The first operating environment. Still the place I measure distance from.

  2. 2021—23

    BleFos

    Founder and operator of an eyewear business. A practical education in demand, margins and doing the unglamorous work.

  3. 2022—25

    Heidi Health

    Intern to business development inside an early healthcare AI GTM team.

  4. 2025

    Achazt

    Acquisition search and diligence. Learning to inspect a business before falling in love with the story.

  5. 2025—now

    Lembra

    AI prototypes, customer discovery and a sharp increase in what I can build for myself.

A useful boundary/07

One line I won’t blur.

Range without theatre.

I’m not a career enterprise closer, and I haven’t carried a complex institutional deal from first hello through procurement and signature. That is a real gap, not a footnote.

What I do bring is adjacent and useful: early GTM, product-led growth instincts, customer discovery, commercial research, technical translation and an AI-native habit of building the first system myself. I would rather be precise about the gap than decorate it.

01

Unfamiliar market → sourced point of view

02

Product and customer signal → next experiment

03

Technical reality → clear commercial narrative

04

New tool → working first system

The next room

Past the whiteboard. Before everything has a department.

I’m drawn to AI-native companies with a real product, unfinished commercial systems and work that sits close to founders, product, data and customers. If that sounds familiar, let’s compare notes.

Start a conversation on LinkedIn
Melbourne, AustraliaOriginally from Samoa© 2026 Lucky Westerlund