Before you fix it, find the twist.
The expensive mistakes are made in the first fortnight, when nobody has looked properly yet. I look properly. Then I build a working thing to prove whether the answer holds.
- $220M+Program recovered
- 4+Applications shipped
- DaysTo a working prototype
- 1Peer-reviewed method
One method, pointed at two questions almost nobody answers with evidence.
Both get asked constantly in meetings and answered from memory. I answer them from data, then build something to test the answer.
Where does the effort actually go?
Not the process map. The real distribution of human hours across your organisation, measured rather than remembered — then each task scored on how well a machine could carry it, what trying would cost, and what breaks if it fails.
- Effort baseline, measured
- Task-by-task automation scoring
- Ranked candidates with reasoning
- Runs on local models — nothing leaves the building
Where do the dependencies actually go?
Trace your inputs to their true origin — vendors, components, cloud regions, model providers, the single supplier three tiers down that four of your products quietly share. Most exposure registers stop at tier one, which is where the interesting risk starts.
- Dependency trace beyond tier one
- Concentration and single-source flags
- Jurisdiction and control exposure
- Substitution options, costed
Nobody could picture it. So I built it.
A conversation with a luxury retailer turned to what a digital sales tool might look like for one of their products. Nobody in the room could picture it, and the honest answer to "what would it look like?" was that none of us knew.
The usual next step is a proposal — a deck, a mood board, a statement of work, and a decision made months later on the strength of an argument. Instead I built the demonstrator: rendered in the client's own design language, driven by a calculation engine validated against published reference data before a single screen was designed.
The meeting stopped being about whether to build something and became about the thing itself.
It has since become a small family of demonstrators, each built around a different product, plus one released independently to prove the underlying engine in public. Each exists because describing it would have taken longer than making it.
This is the method at personal scale. Map what you actually know, notice the gap, then build the smallest working thing that closes it. It scales to a $220M program the same way it scales to a weekend.
- What it is
- Installable web applications, offline-capable, retail-grade finish.
- Built by
- Specification and direction, not a development team.
- Time to first working version
- Days, not quarters.
- Engine
- Calculations validated against published reference data before any interface work.
- Outcome
- A product conversation opened by the artifact rather than a pitch.
Things that shipped, and what they changed.
A federated ontology for capability sustainment
A structured way to describe how long-lived systems change over time, so upgrade decisions across separate organisations can be reasoned about rather than renegotiated every cycle. Peer-reviewed and published. It is also the backbone of how I trace dependency across organisational boundaries.
DOI 10.1177/15485129261453170
Agentic Automation Discovery
The instrument behind question one. Maps where human effort genuinely goes, classifies each task by automation potential, and produces an evidence pack an investment committee can act on. Designed to run offline on local models.
investment case
A business unit, from proposition to governance
Set the strategic direction for a new business unit and built the governance model it would run under. Financial models and forecasts underneath a five-year, $5M investment case projecting annual revenue past $6M by year six — and the technology choices made at governance level rather than left to whoever moved first.
Strategy and governance model
A $220M program, turned around
Taken on mid-flight and returned to a defensible position: scope re-baselined against what the evidence supported, reporting rebuilt so the numbers meant something, and the delivery team restructured around the work rather than the org chart.
Sea trials, safety and national shipbuilding
Trials on amphibious and frigate platforms, a fleet-wide safety fix, and a systems-engineering approach for a national shipbuilding program recognised with its parent company's highest internal award. Long-lead items, obsolescence and sovereign supply were the daily material.
Three passes, in order. The order is the point.
Skipping to the third is the most common and most expensive mistake in both disciplines.
Map
Where the effort goes, or where the dependencies go. Not the diagram on the wall — the real one. I instrument and interview until the picture is defensible, and I say so if the answer is that nothing needs building.
Classify
Everything scored on the same three axes: what it would cost to change, what changing it would return, and what breaks if it goes wrong. The output is a ranked list with the reasoning attached, never a vendor shortlist.
Prove
Build the top candidate as a working thing, measured against the baseline from pass one. If it doesn't beat the baseline you have spent a fortnight instead of a financial year.
Three rungs. Most clients only need the first.
Each stands alone. Take the next one only if the previous one earned it.
Discovery
One of the two questions, answered with evidence. Fixed scope, fixed fee, and a written finding you own outright.
You get: baseline, ranked candidates, investment case.
Prototype
The top candidate built as a working thing and measured against the baseline. Specified and directed here, so it arrives while the finding is still current.
You get: a running artifact and a measured result.
Delivery assurance
Your team builds it. I hold the specification, the measurement and the hard questions — the role that gets a program back on its baseline rather than staffing it.
You get: independent oversight and honest reporting.
Shavi Nachman
Twenty years directing complex capability programs — maritime platforms, national shipbuilding, rail signalling — most recently across a multi-program portfolio at the two-billion-dollar scale, alongside setting the strategy, governance model and investment case for a new business unit. The through-line has always been the same: work out what is actually true before committing anyone's money to it.
TW Innovations exists because that discipline turned out to transfer. The same instinct that recovers a stalled program will tell you which half of an automation business case is imaginary, and where a supply chain is quietly one supplier deep.
Engagements are directed personally. Specialists are brought in by name when the work calls for them.
- MBA — Carnegie Mellon, Tepper
- Fellow, Engineers Australia (FIEAust)
- Senior Member, IEEE (SMIEEE)
- Asia-Pacific Engineer (APD)
- Published — J. Defense Modeling & Simulation, 2025
- Chairman's Award, 2021
When not to call me.
Saying this up front saves both of us a fortnight.
The decision is already made
If the platform is chosen and you need a document supporting it, you want someone else. Discovery that can only reach one conclusion isn't discovery.
You need bodies
I don't staff delivery teams or place contractors. If the gap is capacity rather than clarity, a resourcing partner will serve you better and cost less.
The answer must be reassuring
Findings go where the evidence goes. Sometimes that means the program is fine and you have spent money confirming it. More often it means something uncomfortable.
Tell me what isn't working.
The first conversation is about your problem, not my method. If discovery isn't what you need, I'll say so and point you somewhere better.
admin@twinnovationsaustralia.com- Based
- Melbourne, Australia. Engagements run remote by default.
- Sectors
- Industrial, transport, manufacturing, luxury retail.
- Typical first step
- A two to four week discovery ending in an evidence pack you own.