Real-Time Intelligence Across Australian Government

The Challenge

Australia makes decisions that move markets and reputations at three levels at once: more than five hundred local councils, every state and territory parliament, and the Commonwealth. For the public-affairs and government-relations firms whose entire job is to see those decisions coming, the underlying data is a nightmare of fragmentation. Every council publishes its agendas differently. Every parliament runs its own Hansard, its own search engine, and its own terminology for the same procedural act. Tracking a single issue, say a planning reform or a short-stay-accommodation crackdown, means watching dozens of incompatible websites by hand, and still missing the early signal.

The intelligence already exists in the public record. It is simply unusable, scattered across hundreds of sources, unstructured, and impossible to query as one picture. So a firm advising a national client cannot answer the questions that actually matter. Where is this issue surfacing first? Who is driving it? How is it framed differently from one jurisdiction to the next?

Stromy built two products to close that gap, one for the grassroots and one for the parliaments, and made both queryable through a single AI-native interface.

Our Approach

Council monitoring, intelligence from the ground up. Most political risk is visible at the local level long before it reaches a state or federal agenda. Our council-monitoring engine scans local-government agendas and meeting papers for the topics that matter to a national player, starting in Victoria and built to scale across every council in the country. Point it at an issue, say every mention of short-stay accommodation across council chambers, and it surfaces where and when that conversation is happening at the earliest tier of decision-making, before it climbs. It runs on a resilient crawling stack designed for the messy reality of hundreds of independent council websites, so coverage holds where ad-hoc alerts and manual clipping break down.

The unified parliamentary database, every Australian parliament in one model. This is the heart of the platform. Nine parliaments (the Commonwealth, all six states and both territories) publish their records in nine different ways: structured feeds here, scanned documents there, archives that actively resist machine reading. We built extraction machinery for every one of them and normalise it all into a single canonical model of parliamentary business: debates, divisions, questions on and without notice, petitions, tabled papers, and bills across every reading stage. One vocabulary, one relational schema, across all nine jurisdictions. The corpus spans hundreds of sitting days of captured debate, grows with every new one, and a five-year historical backfill is under way. Questions that were effectively impossible before, such as how a party frames the same issue in New South Wales versus Victoria, or how often a topic has divided a chamber anywhere in the country, collapse into a single query.

An owned asset, not a wrapper. Most government-data tools are a thin layer over somebody else’s API, and live or die by it. This platform is different by design: Stromy extracts the record at the source, builds the database, and hosts it ourselves. The extraction layer is hardened against the real behaviour of government publishing: portals that block, formats that shift, archives that go quiet. A source failure is loud and fixable, never a silent gap in the record. Owning the asset changes what is possible: query depth is not limited to what an official search box allows, history does not disappear when an upstream website redesigns, and the schema is engineered for analysis rather than publication. Every sitting day adds to a corpus we control outright, which means the product compounds instead of depending.

Built for the way agencies actually work. Both products are exposed through MCP, so a government-relations team can interrogate them in plain language inside their own AI workspace, against a clean relational source of truth: no scraping, no stitching together exports, no spreadsheets quietly going stale. And on top of the database, the analytical playbook we proved on our Dutch government-intelligence engine is now being formalised for Australia: cross-jurisdiction issue tracking, party-framing comparison, deep historical research, forward-looking sitting-calendar watch, all packaged as repeatable intelligence workflows that client teams run themselves, from ad-hoc questions to standing reports.

The Outcome

A firm working with these products sees all Australian decision-making as one continuous surface, from a shire council’s Tuesday-night agenda to a Senate division, and can follow an issue across every level of government from a single place. The grunt work of collection disappears, and what remains is the analysis and the advice that clients actually pay for.

Because the data is structured, self-hosted and MCP-native, the advantage compounds twice over: the corpus deepens with every sitting day, and every new question a team asks can be captured as a repeatable workflow. The intelligence operation gets sharper with use while competitors are still working the websites by hand. The firm stops reacting to the news and starts seeing it form.