Most generative AI never makes it out of the demo room.
MIT’s Project NANDA put a number on it last year. Its report The GenAI Divide: State of AI in Business 2025 found that 95% of organisations are getting zero measurable return from generative AI. Only around 5% of integrated pilots are extracting real value, and the rest are stuck with no measurable P&L impact. The researchers blamed neither the models nor regulation, but the approach: brittle workflows, systems that do not learn, and tools that do not fit how work gets done.
Now apply that to regulated industries. A chatbot that is wrong now and then is an annoyance. A compliance engine that is wrong now and then, and cannot explain why, is a liability. In RegTech, ClimateTech and compliance tech, “it works in the pilot” has never been the bar. The bar is: can an auditor follow it, can a regulator question it, and can it keep up when the rules change next quarter?
That is the gap The Improbability Company and Savvycom have agreed to work on together.
What was signed
On 17 September 2026, at Savvycom’s headquarters in Skypark Tower B, Hanoi, Tony Nguyen (Nguyễn Quang Hưng), Chief Growth Officer of Savvycom, and Oscar Omegna, Managing Director of The Improbability Company, signed a strategic partnership agreement. Sam Conroy, President of AusCham Vietnam, witnessed the signing. Both companies are AusCham members.
The focus is plain: AI for RegTech, ClimateTech and compliance tech across Australia, Vietnam and Chile.
To be clear: the agreement sets the direction. As AusCham put it, “the specific research agenda and the first areas of joint work will be agreed between the two companies in the months ahead.” There is no product launch and no joint R&D studio yet. There is a commitment between two engineering companies to decide together what gets built, and to build it for regulated environments from the start.
Why regulated AI stalls in pilot
Three things keep regulated AI in pilot purgatory, and none of them are about model quality.
1. Auditability gets added late. Teams prove the AI can do the task, then try to add logging, explanations and controls on top. In regulated work that order is backwards. If you cannot show why a decision was made, the decision is unusable.
2. Rules move faster than roadmaps. Compliance obligations are not fixed. Australia’s AML/CTF regime widened to lawyers, accountants and real estate agents from 1 July 2026, which we covered in AUSTRAC Tranche 2 Is Live. Chile’s CMF has moved from registering fintechs to cancelling and refusing them, as we wrote in Chile’s CMF Fintech Registry. A system that takes a quarter to absorb a rule change is already behind.
3. Every market is a different rulebook. Building for one jurisdiction and “localising later” creates a rebuild, not an adaptation.
AusCham’s write-up describes the partnership’s answer the same way: in these sectors, “AI systems must be fully auditable, transparent, and adaptable to evolving compliance standards”, and working across three countries means “making multi-market compliance a foundational design principle rather than a later adjustment.”
Why Australia, Vietnam and Chile
The three markets look like an odd set until you look at the work.
Australia has some of the most demanding compliance and energy regulation in the region, and buyers who expect software to stand up to it.
Vietnam has deep engineering talent and a technology sector that is moving, in AusCham’s words, “from service provision toward a role in deciding what gets built.” That shift is the subject of our recent post From MoU to Mekong, and it sits on top of the wider Australia–Vietnam tech corridor.
Chile is a live test of both compliance and climate: a fintech regime that now enforces, and an energy market where distributed solar and storage are scaling fast.
A system designed for all three from day one beats one patched to fit each in turn.
What each side brings
Savvycom was founded in 2009 and has served more than 100 enterprise clients, bringing AI consulting and software development for large organisations with “demanding requirements for security, compliance and scalability.” The Improbability Company brings the engineering behind complex systems for workforce compliance, electrification and climate technology.
Tony Nguyen described the shared intent:
“Our partnership with The Improbability Company brings together our technology expertise and shared commitment to responsible innovation, enabling us to deliver AI-powered solutions across RegTech, ClimateTech, and Compliance Tech. Together, we aim to help businesses navigate complex regulatory environments, strengthen transparency and governance, and turn their ESG commitments into measurable business outcomes.”
Oscar Omegna explained why Savvycom:
“The reason Savvycom caught our attention so much is because of the integrity and transparency which they have behaved with in the last 16 years… As for what we build together, it’s whatever we can create that has the biggest splash.”
And Sam Conroy summed up the model:
“It’s not just a buy and sell. It’s a true collaboration.”
Co-building, not outsourcing
That last line matters most. The usual Australia–Vietnam story is a buyer and a vendor: specs go one way, invoices go the other. That model is fine for clearly scoped delivery. It is poor at the thing regulated AI needs most, which is shared judgement about what to build, how to make it auditable, and when it is ready to leave the pilot.
The partnership is set up the other way round. Both companies set research directions and product roadmaps together, and both carry the outcome. For regulated industries that is the difference between a demo and a system you can put in front of an auditor.
What comes next
The research agenda and first joint work will be agreed in the months ahead. When there is something concrete, we will write about it here.
Until then the aim is simple. Most AI pilots die in the gap between “it works” and “we can stand behind it”. We intend to close that gap.
Sources: AusCham Vietnam, “Savvycom and The Improbability Company Sign Partnership Agreement to Accelerate AI Adoption” (24 Sep 2026), https://auschamvn.org/news/savvycom-and-improbability-company-sign-partnership-agreement-accelerate-ai-adoption · Savvycom, “Savvycom và The Improbability ký kết hợp tác thúc đẩy AI”, https://savvycom.vn/news/savvycom-improbability-ky-ket-hop-tac-ai/ · Savvycom LinkedIn announcement (23 Sep 2026) · MIT Project NANDA, The GenAI Divide: State of AI in Business 2025.