For the first time in the history of work, expertise scales like software. The judgement of a senior accountant, a strata manager, a commercial broker — the decisions that until now lived only in the heads of experienced operators — can be captured, structured, and deployed at near-zero marginal cost.
That unlock is the thesis behind Dragonfly. We believe the largest economic transition of our lifetimes is already underway, and that the firms who own it will not be the AI labs, not the SaaS incumbents, and not the advisory houses. They will be a new kind of operating company — one that owns operating services businesses and rebuilds them from the inside, AI-native from day one.
Services scaled with people. Technology scaled with capital. AI collapses the divide.
The old constraint
Every service business eventually hits the same wall: the cost of its best people. You can grow revenue, but only by hiring. Quality degrades as expertise gets stretched across more clients. Margins compress. The best operators burn out, the median regresses to the mean, and the business becomes a people-management problem rather than a customer problem.
This isn't a failure of leadership. It's the structural limit of a model where the unit economics are bounded by human judgement — judgement that takes years to develop, cannot be transferred cleanly, and leaves the business when the person does.
SaaS tried to solve this by selling better tools. It didn't work — or rather, it worked only at the edges. Annual labour spend on services dwarfs software spend by more than an order of magnitude. Tools helped operators document their work; they never replaced the work itself.
What actually changes
AI is the first technology that can do the work, not just organise it. It can read the contract, draft the response, run the calculation, triage the incident, classify the invoice, resolve the exception — with the same kind of context-sensitive judgement an experienced human would bring.
But “doing the work” isn't a feature you buy. It's a whole new operating model. And that's the gap between what the frontier makes possible and what most companies are actually delivering.
An AI-native services business looks nothing like a services business with AI bolted on. Four differences show up immediately:
- 01Knowledge is the product, not a feature
The moat is the structured, auditable, continuously updated model of how decisions actually get made in the business. Agents are a surface; the knowledge substrate is what compounds.
- 02Every action is inspectable and reversible
Production AI in regulated services can't be a black box. Every decision carries its reasoning chain, confidence score, and the rules it applied — so humans can audit, and systems can improve.
- 03Generation and validation are separate
AI proposes; knowledge validates. The same engine that drafts the response checks it against policy, precedent, and constraint — before it ever reaches a customer.
- 04Autonomy is graduated, not binary
Start in Assist mode with a human in the loop. Earn your way up to Autonomous as the model proves reliability on a given task class. The path to scale is trust, not bravado.
Why ownership is the moat
You cannot do this work from outside the business. Advisors write decks. Software vendors ship features. Neither sees the decisions get made, owns the margin when they go right, or pays the cost when they go wrong. The compounding loop — extract judgement, structure it, deploy it, improve it — runs only where capital, technology, and operating talent sit under one roof.
That's why we acquire rather than advise. A firm that owns the P&L can invest in a five-year knowledge graph because it captures every basis point of the upside. A consultant cannot. A vendor cannot. The architectural patterns that take the first acquisition two years to build take the second one six months — because the platform, not the people, is what compounds.
You cannot scale judgement from the outside. You have to own the business where the decisions get made.
What we're building
Dragonfly is an AI-native operating company. We acquire established services businesses in Australia and New Zealand — cashflow-positive, people-intensive, knowledge-rich — and rebuild their operating models around intelligent systems. Not tools. The work itself.
Our platform is a knowledge engine built around four layers: a data substrate that unifies documents, databases, voice, and operational exhaust; a cascade resolution model that layers regulatory, domain, company, and entity-specific rules; a constraint engine that separates AI generation from knowledge validation; and an agent architecture with graduated autonomy, verifiable traces, and production-grade reliability.
The same engine ports across verticals. Strata management, aged care, accounting, customs & trade, home services, IT managed services — industries where judgement is the cost base, rules are entity-specific, and the data is rich but unstructured. Each acquisition extends the platform; the platform accelerates each acquisition.
The bigger claim
We think the dominant operating model of the next era is the AI-native services firm: a company that owns the delivery, captures the knowledge, and compounds across decades rather than quarters. Not a consultancy. Not a SaaS platform. Not a traditional holdco. A new category of firm, defined by active ownership, vertical intelligence, patient capital, and full-stack execution.
Software could organise work. It could not do the work. Until now.
