Open to senior AI strategy & product roles · Sydney

Annabel
Nguyen

I turn AI ambition into funded, working product.

Most organisations have the ambition. What they need is someone to work out which bet is worth making, build the case for it, and stay until the PoC ships. At Laing O’Rourke, I did that across four AI projects in six weeks. At Nous Group, it became a $1M ARR product. I don’t hand over a strategy deck. I stay until something ships.

Schedule a call LinkedIn Test the methodology
$1M+ ARR: Labour & Skills analytics platform, now sold standalone
$700K Saved annually: five process automations, all live across 7 projects
4 / 5 AI bets assessed, business-cased and funded in six weeks

How I decide which AI bets to fund

Most AI initiatives fail before they start, not because the technology is wrong, but because no one does the work of deciding which problem is actually worth solving. Before any roadmap or build decision, I run this assessment. At Laing O'Rourke, it took five candidate use cases down to four funded bets in six weeks.

The matrix forces an honest conversation. Every dimension gets a score: economic upside, technical feasibility, data readiness, adoption risk, speed to value. Only opportunities that score above threshold get funded and resourced. The rest get a clear 'not yet' with the reasoning documented.

Data readiness is the dimension most organisations skip. Approving a high-upside AI use case before confirming the data exists is how promising projects stall six months into delivery. I weight it heavily because I've seen what happens when you don't.

Broad AI adoption is not the goal. Finding the three use cases where the economics are compelling, and funding those properly, is.

Score your AI opportunity

Rate each dimension 1–3. The total tells you whether this is worth funding, needs de-risking first, or should be parked.

Economic upside:
will this make or save money at scale? ($500K–$5M+ annually)
Feasibility:
do we have the tech maturity and internal skills?
Data readiness:
do we have clean, labelled training data?
Adoption risk:
will teams actually use this?
Speed to value:
how fast until we see measurable results?

AI Investment Readiness Matrix

Tap a score per row
DimensionLowMidHigh
Ready to fund?
— Select scores above

Selected work

Three engagements. Each shows a different part of the job: strategy, business case, shipped product.

Laing O'Rourke · Construction & Infrastructure

Pioneering an AI vision for project delivery: 5 use cases, 4 funded in 6 weeks

Tier-1 Australian contractor. 17 stakeholders across executive, GM, director and project-lead levels. Six-week mandate.

OutcomePoC live and in use. Production roadmap funded, targeting 2027 deployment.

4 / 5use cases funded

Problem

The organisation had real AI ambition and a blank slate. No one had worked out which problems were worth solving or how to make the case for funding them. My job was to build an investment thesis that got money, not a deck that got nodded at.

Approach & tradeoffs

Ran discovery across 17 stakeholders. Mapped six project-delivery phases. Surfaced 27 business problems, narrowed to five AI candidates, then applied the investment-readiness framework. Wrote the business cases. Four were approved and funded in six weeks. The fifth wasn't ready; that decision was documented, not buried.

PoC development

Built working PoCs for the two highest-priority use cases: AI-assisted process playbooks and HSE planning, using Microsoft AI Foundry and Databricks. Stakeholders could interact with something real, not evaluate a concept on a slide.

Lesson learned

In project-delivery environments, the barrier to AI adoption is rarely technical. It's trust. A working PoC does more in one meeting than a business case does in three, because it answers the question executives are actually asking: does this work in our context?

AI StrategyStakeholder AlignmentBusiness Case DevelopmentMicrosoft AI FoundryDatabricksPoC DevelopmentConstruction & Infrastructure

Laing O'Rourke · Construction & Infrastructure

$700K a year in savings: automating five business processes end-to-end

A parallel workstream. Mapped processes across three business areas, identified highest-return automation candidates, led a team of two to build and ship them.

OutcomeRunning live across 7 projects, generating savings on each.

$700Kannual savings

Problem

Teams were spending real hours on manual, rules-based work. Everyone knew automation was possible; no one had worked out which processes were worth the effort, or built anything the teams would actually use.

Approach & tradeoffs

Ran workshops across three business areas to map processes and find where automation had the strongest payoff. Narrowed from many candidates to five, the same disciplined filtering I apply to AI use cases. Then built them: Python automation with a React front end, designed around how teams actually worked.

Delivery & outcome

Led a team of two. Shipped five automations. $700K in annual savings, running live. This project and the AI vision work are two sides of the same capability: I can set the strategy and I can ship the product.

Lesson learned

The build is the easy part. The value is in correctly identifying which five processes deserve it, and saying no to the other ten. Scope discipline with a small team is what makes the savings real.

Process MappingAutomationPythonReactTeam LeadershipStakeholder Workshops

Nous Group · Government & Education

From internal tool to $1M ARR product: labour & skills analytics for universities

Labour & Skills analytics dashboard integrating university enrolment data, HEIMS, and Burning Glass labour-market data, with AI to surface insights. Started as a consulting deliverable. Now a standalone product.

Outcome3 paying university clients at launch. Still growing as a standalone product.

$1M+ARR achieved

Problem

Universities had the data: enrolment systems, HEIMS, Burning Glass labour-market signals, but none of it in a form that supported actual decisions. Strategy conversations were happening without a common factual basis.

Approach & tradeoffs

Managed design and delivery of the dashboard in R Shiny, integrating all three data sources. Added AI to surface the patterns that mattered. Built executive decks generatable directly from the product, reducing the gap between data and decision to near zero.

Product outcome

What started as a client deliverable became a product. Nous Group now sells it as a standalone offering layered on top of consulting services. $1M+ ARR across three university clients.

Lesson learned

A data product is only as valuable as how well it fits into a real decision-making workflow, not how sophisticated the technology is. And consulting IP, if it's good, can be productised. This one was.

Product ManagementLabour Market AnalyticsR ShinyAI IntegrationHigher EducationCommercialisation

Working principle

Every organisation I work with starts by asking where they can use AI. That's the wrong question. Start with the most expensive problem you have. Find the best solution. If AI is the answer, fund it properly. If it isn't, don't use it.

I work best with organisations that have real AI ambition and are willing to make hard prioritisation calls. I communicate frequently, especially when the news is a deprioritisation. My measure of success is what ships and what lands, not how many people agreed with the slide.

Background

Period Role & Organisation
2023–
Present
Laing O'Rourke Current
Senior Analyst, Digital Development & Insights
2021–
2023
Nous Group
Manager, Business & Digital Strategy Consulting
2021
Deloitte
Analyst, HR Technology & Transformation Consulting

Point of view

Three things learned the hard way, across construction, higher education, and enterprise consulting.

01

Problem first, technology second. Every time.

Most organisations start by asking where they can use AI. That's the wrong question. The right one is: what's the most expensive problem we have, and is AI actually the best way to solve it? The Nous Group product reached $1M ARR because it solved a real decision-making problem. The technology was R Shiny and Burning Glass, nothing exotic. AI earns its place when it's the highest-return answer. Chosen for its own sake, it produces expensive pilots that never ship.

02

Concentration beats coverage.

Organisations that try to apply AI everywhere get mediocre results everywhere. The ones that pick three high-value use cases and fund them properly get compounding returns. At Laing O'Rourke, five candidates went through the framework. Four were funded. The fifth was documented as not-yet-ready. Saying no to the weak cases is what made the strong ones succeed.

03

The technology is rarely the hard part.

In construction, in higher education, in large enterprises: the people closest to the work have the most to gain from AI and the most invested in how things currently work. Change management determines adoption outcomes more than product quality does. The sequencing of who hears what, and when, matters as much as what you build. A working product with a bad rollout fails. A good rollout with a mediocre product often succeeds.

The framework, in two working tools

Most consultants describe how they work. These let you test it. Run the diagnostic against your own initiative — or interrogate the methodology directly. Either way, you'll know whether it's worth a call.

The Scope Diagnostic

Is your AI initiative worth funding?

Four questions against the five-dimension framework. A full investment readiness report — dimension scores, a prioritised fix list, and an unambiguous verdict: Fund it. Pause. Redirect.

Run the diagnostic

The Interrogation Room

Put the methodology under pressure.

Ask anything about the framework, specific case studies, how the funding decisions were made — or what got killed and why. AI-powered, grounded strictly in what's documented.

Open the Interrogation Room

Have an AI opportunity worth funding?

15 minutes. We'll work out whether the opportunity is real, what it would take to fund it, and what the right next step is.

Schedule a call LinkedIn