Government
Public-service modernisation - the next generation of automation and experience.
Government departments, agencies, and service partners have the challenge of delivering customer (citizen) services that meet modern expectation, with the added need to architect for scale.
Every corner of the government tech stack has undergone rapid evolution over the past 5 years. Bigger, faster, smarter, and more reliable. Adopting these advancements is often a complex program of change, but the real opportunity lies in measurable interventions that ship value in months, build evidence, and earn the right to do more. This is the work we do in government.
Our proof in government is two production engagements, a state insurance regulator and a state government department, both anonymised at the client's request. The discipline underneath them, making the rules that govern public services visible, auditable, and owned by the people responsible for them, transfers to any department where rules matter and audit trails are not negotiable. Newer work in accessibility and citizen service is emerging, and we mark clearly what is proven and what is still possibility.
The landscape
Four problems departments, agencies, and regulators often recognise
We have worked with and spoken to a number of government departments and their extended network of supporting partners over the years. Here are some of the key challenges we often observe.
The rules are invisible to the people responsible for them
The logic deciding who gets which service, who pays what, who qualifies, lives in legislation, policy documents, and the heads of long-serving staff. It is also buried in systems built fifteen years ago and modified hundreds of times. When policy changes, finding where the rule actually lives is detective work.
Policy changes. The system doesn't.
When the policy team needs a rule changed, the answer is a release cycle measured in quarters, queued behind every other IT priority. Policy intent and system behaviour drift apart. And because the rules are opaque, as much as you may test, some gaps emerge late in the program, or even after go-live.
You can't always show your work
When a decision is challenged, on review, appeal, FOI, or audit, you have to reproduce exactly which rule applied, in which version, against which data, at that moment. Most systems can tell you what they decided today. Far fewer can tell you what they would have decided two years ago, and that is the question that matters.
The real work is in the exceptions, and no one can see it
Most systems are built for the happy path. The edge cases and ambiguous submissions fall into a catch-all queue with a generic error and a reviewer eyeballing everything. The exception path becomes the actual workload, invisible to everyone judging whether the system works.
What good looks like
A department that has solved these
Four shifts that follow when the four problems above are addressed structurally. None of them require a five-year program. Each one is achievable in increments, with evidence built at each step.
The rules are visible, and they belong to the business
The logic governing a service is written where the policy team can see it, test it, and change it. No archaeology. No dependency on the one person who remembers why.
Policy moves at the speed of policy
A rule change is modelled, tested against the live ruleset, and shipped in days. The team sees the impact before it commits, not after.
Every decision can be reproduced
Versioned rules and immutable logs mean any decision can be replayed exactly, against the rules in force when it was made. Audit becomes verification, not recovery.
The exception path is designed, not endured
Routine cases clear automatically. The hard ones route to the right person with the context already gathered. The exception queue stops being the workload.
Two state-level engagements anchor what we do in government.
Both anonymised. Both running in production. Together they cover the two patterns most government technology programs care about: making the rules governing public services visible to the policy team that owns them, and making high-volume decisioning auditable at scale.
Use case 01
State government department
Contract management. We made the business logic governing public funding visible, auditable, and manageable by the policy team. Rule changes moved from quarters to weeks. The policy team gained the ability to model changes before recommending them.
Use case 02
State insurance regulator
Lodgement validation at scale. High-volume batch processing with complete audit trails, protecting millions of citizens behind the scheme. Decision logic externalised, owned by the regulator, ready for regulatory and product change.
The role of AI
Where AI fits
AI earns its place in a public-service decision when it stays in its lane. It reads and classifies the messy input, documents, forms, free text, and turns it into something clean. Deterministic rules make the decision, so it stays consistent, explainable, and auditable. A person holds the exceptions. The audit trail covers both. That split is also how you stay on the right side of the assurance and accountability expectations governments are now setting for AI.
How we have helped
Two engagements, two patterns
Two production engagements anchor our government work. Both anonymised at the client's request. Together they cover the two patterns most government programs care about: making the rules that govern public services visible to the team that owns them, and making high-volume decisioning auditable at scale.
State government department
Contract management - making policy rules visible and shippable
An Australian state government department running a high-volume contract program. We made the business logic governing public funding visible, auditable, and manageable by the policy team. Rule changes moved from quarters to weeks. The policy team gained the ability to model changes before recommending them.
State insurance regulator
Lodgement validation at scale, with full audit
An Australian state insurance regulator running high-volume lodgement validation. The platform validates filings in batch processing with complete audit trails, protecting millions of citizens behind the scheme. Decision logic externalised, owned by the regulator, ready for the regulatory and product change cycle the regulator operates against.
How we engage
Pilot first. Then more.
The government buyer is risk-averse for good reason. Five-year programs with fixed scope and immovable end dates are how regrets are born. We do not work that way. We deliver in increments that ship, get used, and get measured.
Pilot
A pilot is bounded - small scope, defined success criteria, evidence-based outcome. Three to six months. Targeted at one high-value pattern in your context. Cost low enough to fund from operational budget if needed.
Evidence
The pilot produces measured outcomes - not promises. Time saved, errors reduced, citizens served, decisions consistent. The evidence sits with the department, not the vendor.
Extension
If the evidence justifies it, the pilot extends. New pattern, adjacent context, deeper integration. Each extension a new commitment with its own scope, its own success criteria, its own evidence. Not a march toward a megaproject.
This is not the only way to engage with us. For departments that prefer a different model, we will work the way that suits the procurement and risk posture. The pilot-first pattern is what we recommend by default because it is what we have seen work in government contexts.
Where this is heading
Beyond the decisioning work, three capabilities are moving from production elsewhere toward public-service use in ANZ. We mark them as emerging because that is what they are.
Citizen services that speak sign language
In production today in Italian Sign Language, with the architecture extensible toward Auslan, so deaf and hard-of-hearing citizens get equal access without an interpreter in the loop.
See the capability →Conversational AI for routine citizen queries
Answers routine questions in the citizen's own language and hands off to a person for anything that needs judgement. The same logic as the exceptions problem: clear the routine, free people for the hard cases. Proven at scale in a healthcare setting handling 100,000-plus conversations.
See the capability →AI plus immersive kiosks
Service and information points for physical locations, regional offices and service centres, combining conversational AI with an interactive interface. In production in European public and cultural settings.
See the capability →Solutions that apply
Where to start
Two Solutions cover most of the work above. Process & Decision Automation is the primary entry point - both our state-level engagements anchor here. Agentic Operations covers the citizen-facing dimension.
Process & Decision Automation
Public-service decisioning, rule externalisation, program integrity, lodgement validation. State-level proof anchors here.
Agentic Operations
Citizen-facing services and self-service for routine matters, with accessibility designed in from the start so more people can self-serve without an intermediary.
Related reading
Go deeper
Three pieces on public-service decisioning, the boundary between automation and judgement, and where AI helps in regulated contexts.