The Moats That Matter in AI
Animation by Grok
Read our take on what sets ANZ and Southeast Asia AI companies apart below. You can read our other article with a deep dive on built-world AI, “Cranes, Code and Co-Pilots: Why AI is Finally Cracking Construction and Property,” here. March 2026.
A new breed of AI-first software firm is emerging from Australia, New Zealand and Southeast Asia. Their competitive edge comes from one of two strategies: building AI agents that can execute enterprise workflows autonomously or focusing on defensive moats and industry-specific workflow depth. North Ridge Partners’ analysis of regional AI firms highlighted this, with most companies choosing to compete in regulated, operationally complex and data-intensive corners of the economy that generalist AI platforms struggle to reach. The chosen sectors, often stickier and with higher switching costs, include healthcare, financial compliance, legal operations and physical infrastructure. The result is a regional AI ecosystem that is beginning to look less like a curiosity and more like a cohort with genuine global relevance.
Capital markets are starting to treat the region’s AI-native companies with the same seriousness once reserved for their American counterparts. Harrison.ai’s US$112 million Series C in 2025 and the US$355 million valuation achieved by legal-tech startup Ivo in January 2026 suggest the old assumption that breakthrough AI companies must be born in Silicon Valley no longer holds true.
The timing now matters. A cohort of AI-native software companies that raised meaningful growth capital between 2023 and early 2026 is entering a decisive phase. Firms such as Heidi Health, Relevance AI, Affinda and Lorikeet deliver their products through SaaS platforms, but AI sits at the core of their functionality rather than as an add-on. The next 18–24 months will likely determine which of these companies evolve from promising regional startups into globally relevant AI platforms.
AI down under finds its edge in the hard bits of the economy
For a brief moment in the early 2020s, it seemed as if artificial intelligence would flatten geography. Models trained in San Francisco or London could, in theory, be deployed anywhere, reducing the importance of local ecosystems and national technology clusters. Yet the emerging generation of AI companies in Australia, New Zealand and Singapore suggests something rather different. Far from competing head-to-head with Silicon Valley’s frontier labs, many of the region’s most promising ventures are burrowing deep into the awkward, regulated and infrastructure-heavy corners of the global economy. In those niches, distance from California may be less of a handicap than once feared.
Our analysis of 41 AI-native and AI-enabled business software firms across Australia, New Zealand and Singapore illustrates the shift. Together, they have raised over A$2 billion and span sectors from healthcare diagnostics and financial-crime detection to construction analytics and energy-grid modelling. Their diversity reflects a broader change in how investors think about AI businesses. Rather than betting solely on raw model capability, investors increasingly favour companies that own the workflows, data and regulatory entanglements that make enterprises reluctant to switch providers.
That logic has sharpened as software valuations wobble. In early 2026 the multiples attached to software companies compressed substantially. The cause was not collapsing earnings but a growing fear that “agentic AI,” systems able to automate complex tasks, will hollow out much of the traditional software stack, ending “per seat” pricing and replacing it with a results-driven model. Investors talk increasingly of a “thin middle”: generic user-interface software squeezed from above by intelligent agents and from below by systems of record that store essential data.
The companies most likely to survive that squeeze tend to share a particular set of traits. They are deeply embedded in industries where regulation, data ownership or operational complexity create barriers to entry. Healthcare offers a vivid example. Sydney-based Harrison.ai, valued at more than A$475 million at an earlier funding round, develops AI-powered diagnostic support and workflow tools for radiology and pathology. Three “breakthrough device” designations from America’s Food and Drug Administration have strengthened its regulatory credentials and made it a plausible target for global medical-technology groups.
When the company announced a raise in February last year, chief executive and co-founder Dr Aengus Tran said the firm was tapping into the growing demand for equitable and effective healthcare, which calls for advanced systems like AI to enhance human diagnostics and address disparities in access to care.
“Harrison.ai meets this need by developing clinical-grade AI models designed to improve capacity,” he said.
Others focus less on diagnosis than on the administrative burdens of modern healthcare. Heidi Health, based in Melbourne, uses AI to listen to consultations and generate clinical notes automatically, easing one of the profession’s most tedious chores. NexusMD.ai, another Australian startup, processes clinical content and medical data using autonomous AI agents that handle referrals, prescriptions, and administrative coordination between GPs, specialists, and patients. In such areas, AI is being deployed as a highly practical piece of operational plumbing.
Regulation provides similar opportunities in finance and compliance. Singapore-based Silent Eight uses AI to automate the investigation of suspicious financial transactions, helping banks meet stringent anti-money-laundering requirements. Auror, founded in New Zealand, tackles organised retail crime by sharing intelligence across retailers and identifying repeat offenders using computer vision and data analytics. These products thrive precisely because they sit inside compliance regimes that organisations cannot easily abandon.
Legal departments offer another sticky domain. LawVu, a Kiwi platform now used internationally, acts as an operating system for in-house legal teams, managing contracts, workflows and external legal spending. The more documents and records accumulate inside the platform, the harder it becomes for companies to migrate away. The company shows how an AI-enabled platform can deepen an existing moat by layering AI onto an already sticky workflow. In enterprise software, the duller the workflow, the stronger the lock-in.
Operational technology
In Sydney, Neara, which just completed a Series D at a A$1.1 billion valuation, builds digital twins of electricity networks, allowing utilities to simulate how power grids will respond to extreme weather, vegetation growth or changing demand. Gridsight, another Australian startup, helps operators manage the surge of rooftop solar panels and home batteries feeding electricity into local networks. In these cases, AI does not merely optimise business processes; it models physical systems whose complexity grows with the global energy transition.
Construction sites provide another unlikely laboratory. Singapore’s Ailytics analyses video feeds from ordinary CCTV cameras to generate three-dimensional spatial intelligence about site safety and productivity. Sitemate, an Australian firm, digitises inspections and reports across construction projects, turning what were once scattered spreadsheets into structured datasets that can be queried and analysed. The value of these companies lies less in the novelty of their algorithms than in the data they accumulate about how work actually unfolds in the field.
Hardware firms appear in the AI ecosystem, too. Arkeus is an early-stage Australian venture in defence AI, one of the fastest-growing government procurement categories in the country following the AUKUS submarine agreement and the national Defence Strategic Review. It develops autonomous optical systems and AI-powered computer vision for defence applications. Morse Micro designs specialised Wi-Fi chips that allow low-power sensors to transmit data across kilometres, enabling AI systems to monitor farms, mines or construction sites. Such businesses sit in a different technological universe from software startups, but they share the same strategic logic: their products are anchored in physical realities that cloud platforms cannot easily replicate.
A second path to defensibility
Not every successful AI company, however, needs a regulatory moat. A parallel group of firms from the region is proving that agentic AI platforms can build defensibility through speed, proprietary intelligence layers and network effects, sometimes even in unregulated markets. These horizontal players compete not by locking customers into compliance workflows but by becoming indispensable to how their clients operate and make decisions.
Australian-led Mutinex is attempting to do for marketing budgets what Bloomberg terminals did for financial markets: replace instinct and spreadsheets with a continuously learning machine that tells executives where their money actually delivers. Mutinex is not the marketing software being displaced but the intelligent agent doing the displacing.
It’s little wonder that Mutinex co-founder Henry Innis now finds himself in New York, from where he is overseeing exceptional growth as the firm expands across North America. The company, founded in 2018 during what might be considered AI’s “traditional machine learning era,” has doubled its revenues since its A$17.5 million capital raise 17 months ago, and its US business is growing at 300 per cent.
Innis tells North Ridge Partners that agentic AI has turned the company’s MMM platform, called MAITE, into an always-on marketing analyst that’s now orchestrating eight agents behind the scenes to answer questions, interrogate data and surface insights on demand.
“We are seeing a huge increase in the volumes of queries on the platform, at rates we weren’t even close to getting before we agentified the platform.”
But that success was not without discomfort. His engineers were so far out ahead of the agentic pack that they needed to develop many of their own evaluation and observability tools, because offerings in the market at the time were so immature. “You wouldn’t need to do that today.”
Sydney-founded Relevance AI takes a similar approach in a different domain. It offers a no-code platform that allows companies to build and orchestrate their own AI agents without deep machine-learning expertise. The idea is to let businesses automate complex processes using teams of digital workers. Investors including Bessemer Venture Partners believe the approach could position the company as an orchestration layer within the enterprise AI stack.
Lorikeet, another Australian AI-native company backed by venture firms including US-based QED Investors, deploys AI agents for complex, high-stakes customer support interactions. Rather than offering scripted chatbot responses, Lorikeet’s agents integrate deeply into backend systems to resolve issues autonomously. Like Mutinex and Relevance AI, its moat is operational: the more workflows it handles, the more institutional knowledge it accumulates, and the harder it becomes to displace. On top of that, Lorikeet focuses on complex, highly regulated industries such as financial services, healthcare, and insurance, further embedding its offering with customers.
The capital behind the code
The venture ecosystem supporting these companies is relatively concentrated. A handful of Australian funds, particularly Blackbird Ventures, Square Peg Capital and AirTree Ventures, appear repeatedly across the cap tables of the region’s most prominent AI startups. International investors such as Insight Partners and Bessemer Venture Partners provide additional validation when local firms begin to scale globally.
The broader lesson of this emerging AI landscape is that there are two distinct paths to defensibility. The first one runs through the stubborn realities of regulated industries: the paperwork of hospitals, the compliance rules of banks, the wiring of power grids and the concrete of construction sites. Companies on this path build moats from regulatory embeddedness, proprietary data, vertical complexity and integration of AI with physical infrastructures. The second path belongs to agentic AI platforms that win by becoming indispensable intelligence layers; not because regulation compels their customers to stay, but because the speed, accuracy and institutional knowledge they accumulate increase their stickiness. In both cases, these businesses remain competitive rather than becoming features inside generic LLM offerings because their defensibility comes not from model access, but from control of workflow, proprietary operating data, integration depth, regulatory fit and domain-specific trust.
In a world increasingly obsessed with artificial intelligence, the winners from Australia, New Zealand and Southeast Asia may be those who understand that the real competition is not about who builds the smartest model, but about who entangles themselves most deeply with the economies they serve.
For enquiries, please contact:
Gerry Gimenez, Director of B2B Software & AI at North Ridge Partners, gg@northridgepartners.com
Like this article? Join the thousands of tech founders, board members and investors who subscribe to our free monthly newsletter, Tech Round-Up. Sign up below!