Airgentic for Local Government
Not one chatbot. A team of council specialists, each of which knows where its job ends.
Residents ask in their own words. A frontline agent answers from your own pages, or hands the question to the specialist that owns it — bin night for their street, the zone and overlays on their block, a rates concession, a hall to hire, what's on in April. Accurate, cited answers at 10pm on a Sunday, and a continuous record for your team of what the community actually needs.
Airgentic in under 60 seconds
Residents are already asking AI about your services. It is already answering.
Ask an AI assistant which day your bin goes out, or whether a front fence needs approval, and it will answer with total confidence — from a page you retired two years ago, an aggregator, or the neighbouring council. Your council carries the consequence, and nobody in the organisation can see it happening.
Put your own numbers on the rest: monthly contact volumes, the share that are simply "when, where, how much" questions, and your cost per contact by channel. This is not a headcount argument — councils do not cut frontline staff. It is a satisfaction argument, and satisfaction is measured, published and politically consequential.
Why this is a live decision, not a next-term project
- Residents arrive expecting to ask, not to navigate — The A to Z service list is a 2010 interaction model, and it is the first thing a resident gives up on.
- A CMS replatform no longer has to gate this — We deploy with a script tag, a Tag Manager container, or an integration for Squiz Matrix, Datacom Datascape or WordPress — beside your web project, not behind it.
“Airgentic has helped us deepen the connection between our community and the services they rely on. The assistant is already improving how residents access information—quickly, accurately, and in their own time. For a region as vast and unique as the Flinders Ranges, this technology is helping us bring people closer to their council than ever before.”
Eight specialists, one conversation
Each agent owns one part of council business, with its own instructions, sources and limits. Residents never have to choose one: Frontline reads the question and passes it to the agent that owns it.
Frontline
Answers general questions from your indexed pages and PDFs, with citations. When a question belongs to a specialist, it hands over inside the same conversation, so the resident never starts again.
Bin Collection
Give it a street address and it returns the collection day for general waste, recycling and organics, and the next date for each. It never guesses a day from the suburb alone.
Planning
Looks up the zone, overlays and bushfire status for an address, and the applications on your register or on public notice. Explains how to apply or object. It will not decide whether a permit is required.
Council Services
How to get something done: what goes in which bin, missed bins, pets, parking, fines, forms, grants, and building work that is not a planning permit.
Rates
Explains the notice, how rates are worked out, valuations, payment options, concessions, hardship and payment plans — with empathy first when a resident is worried. It cannot see an individual account, and says so.
Places
Finds parks, pools, playgrounds, halls, gardens and sports venues, with the address published on your page, and narrows long lists using the fields in your index.
Events Concierge
Answers what's on, when and where, including what is coming up in a particular month or period: "I'm away in August — is anything on before then?"
Search Guide
Steps in when a search returns more than is comfortable to read. Picks out the few results that fit best, and links through to the full filtered results.
What a specialist does that a chatbot cannot
Three everyday questions, and how the agent that owns each one answers it.
Bin night
- The resident"When are my bins collected at 12 Smith Street?"
- Who answersBin Collection
- What they getThe day for general waste, recycling and organics, and the next date for each — looked up for that address against your published collection zones, with a link to the bin night page.
- Then"Can a pizza box go in the yellow lid?" passes to Council Services, which answers from your waste and recycling pages.
Planning at an address
- The resident"Can I put a second storey on my house at 8 Hill Street?"
- Who answersPlanning
- What they getThe zone, every overlay, whether the block is bushfire prone, any applications already lodged there, and how to apply. It says plainly that a zone is not a decision, and points to your permit guidance.
- Then"Do I need a building permit as well?" passes to Council Services.
A worried ratepayer
- The resident"My rates went up again and I'm on a pension. Can I get help?"
- Who answersRates
- What they getThe concern acknowledged first, then the pensioner concession, hardship support and a payment plan told apart, with a link to the one that fits. No promised outcome, and no figure it has not just retrieved from your pages.
- ThenQuestions about their own balance go to customer service, because the agent cannot see an account and does not pretend to.
Live answers for the resident's own address
Many council questions are really about one property: which bin goes out tonight, which overlay covers the block, what has been lodged next door. No web page can answer that. Airgentic's look-up agents query data your council and the state already publish, at the moment the resident asks.
- Bin Collection — Matches the address to your waste collection zones and returns the day and next date for each bin.
- Planning — Reads the planning zones, overlays and bushfire prone areas for the property, and the applications on your register or on public notice.
- Configured, not procured — Each look-up is a small function connected to an existing data source, so adding the next one is a configuration change rather than an IT project.
Results are quoted, never extrapolated. If an address falls outside a collection zone, the agent says so and links the authoritative page instead of inventing a day.
Every specialist knows where its job ends
Councils carry the consequence of a confident wrong answer. So each agent is defined by what it must not do, as well as what it does.
No permit decisions
Planning describes the zone and overlays and links the scheme. It does not tell a resident a permit is not required unless your own page states that rule.
No account guesswork
Rates cannot see a balance or a valuation, and will not calculate a bill or promise a concession. Account questions go to the right team.
No invented dates
Bin Collection quotes only the dates the look-up returned, and says once that the published zones are a guide.
Clear lanes
Each agent's role lists what it handles and what it hands on. Bin day goes to Bin Collection; what goes in the bin goes to Council Services. Two agents never give two answers.
No unnecessary personal details
The agents answer without asking residents for a name, phone number or email address.
Tuned one at a time
Each agent can be tested privately, switched on, adjusted or switched off on its own, without touching the rest of the team.
Three questions that separate real answers from a search box
Every vendor demonstrates well on "what day is my bin collected". These three expose the difference, because council content is messier than almost anyone else's. Take them to your own evaluations and make each vendor type them live, on your content.
The four-constraint question
- The question"I need a hall for 80 people on a Sunday, with a kitchen and step-free access."
- Most platformsReturn the Venue Hire landing page and a PDF list of halls, leaving the resident to check four things by hand, one venue at a time.
- AirgenticThe Places agent reads capacity as a numeric range, the day as availability, and kitchen and step-free access as facility attributes — applies all four as hard filters, and tells the resident which ones it used.
The real calendar
- The question"What's on for kids during the school holidays, and again over the Easter long weekend?"
- Most platformsOffer a single from-and-to date box, and treat a weekly library program that skips a fortnight as one long event running for months.
- AirgenticThe Events Concierge indexes every occurrence with its own dates, including series that skip weeks, and matches both time windows from the one question.
The cross-service task
- The question"I want to build a deck — what do I need from council?"
- Most platformsReturn the Planning landing page and stop, because the real answer is spread across planning, building, fees, forms and neighbour notification.
- AirgenticPlanning checks the zone and overlays at the resident's own address and explains the permit process; the building permit side passes to Council Services in the same conversation. Every answer cites its source and links the forms and fees that apply.
Then it does the thing the resident actually came to do
Every specialist states what it applied and links through to the same filtered results a person would reach by hand. When the list runs long — forty playgrounds, sixty school holiday activities — the Search Guide picks out the few that fit best and links to the rest, instead of reading out a wall of links. Then the conversation finishes the job: "Log it for me." A pothole, dumped rubbish, a missed collection — captured with location and context and routed to the right team. That is a completed transaction rather than a deflected call, which is a far better thing to be measured on.
Precision is an architecture, not a prompt
Most AI search asks a language model to be careful. Airgentic does not rely on care. The agent never writes a URL or a keyword string — it emits a structured plan, and the platform validates every attribute and value against your live index before the search runs.
- Capacity, days and features are real fields — "Seats 80 or more" is a numeric range, "Sunday" is availability and "step-free access" is a facility attribute — not words to pattern-match.
- Values come from your data, not the model — An unsupported value is dropped rather than guessed, so the assistant cannot invent a fee, a facility, a collection day or a permit you do not issue.
- Clarifying questions come from the same place — Too broad a request earns a question about an attribute that actually exists in your index, so every answer narrows the list.
The guarantee: precision is enforced by the platform, not hoped for in a prompt.
Your community tells you what it needs every day. Now you can read it.
Annual satisfaction surveys are sampled and retrospective. Consultation portals are self-selecting, skewed towards the already engaged and the already aggrieved. Meanwhile every question typed into your website is unprompted consultation: continuous, in residents' own words, at the moment of need.
Airgentic clusters questions into themes automatically, then classifies them against categories you define — by service area, by ward or suburb, by intent such as report, apply, pay or find out — each with a named owner.
A worked example
- A cluster forms around changed collection days after a waste route revision
- Volume concentrates in three suburbs within days, with clearly negative sentiment
- Targeted comms go to those suburbs, instead of the problem arriving as a spike in calls and a councillor's inbox
Over a quarter the same data becomes evidence for which service redesigns are worth funding. Confidence scoring, review queues, a named owner per category and drill-through to the conversations behind any number mean the report survives its first council meeting.
When it matters most, change the answer in minutes
Bushfire, flood and storm. An unplanned pool or depot closure. A waste route change, rates notices landing, the caretaker period. These are the moments when a stale answer becomes a real problem — and the moments when a website change request still takes three days.
- Pin, override or replace a specific answer, and publish emergency messaging, without a developer and without a release
- Point residents to the authoritative emergency source rather than letting an assistant improvise
- Narrow what agents may discuss during a live incident or a caretaker period
- Keep a record of what was published, when, and by whom
Outcome: your communications team controls the answer in minutes, in the channel residents are already using.
What procurement will ask, answered
Airgentic answers only from your approved content — pages, PDFs and connected sources — and cites the source of every answer. Grounding is the control that prevents misinformation, not a feature bullet.
- WCAG 2.1 AA conformant, so the assistant does not become an accessibility exception
- AU and NZ data residency options
- A Privacy and Retention Manager, so conversation data can be held and disposed of in line with your records obligations
- Role-based access, audit trails, and content scoped to authenticated users — including the same conversational search behind your staff login, for procedures, delegations and depot processes
And on deployment: one script tag, a Google Tag Manager container, or a purpose-built integration for Squiz Matrix, Datacom Datascape or WordPress, scoped to the URL patterns you choose. Live in days rather than a budget cycle, which matters when the digital team is two people.
Already delivering for local government
Hutt City Council (New Zealand)
A metropolitan New Zealand council. Hutt City launched Awa, an Airgentic agent on the Too Good to Waste website, where residents look up collection days for their own address and get accurate answers about what goes in which bin. Routine waste enquiries are resolved online, so customer service can concentrate on the conversations that need a person.
Ask every vendor to demonstrate these live
Every item below can be performed in front of you, on your own content. We are confident enough in our answers to hand you the questions.
- Take a real street address in your area and return its bin day, and the next date for each bin, live.
- Take a real address and return its zone, overlays and bushfire status — then ask whether a permit is needed, and see whether the assistant decides for you.
- Ask about a specific rates account, and show the assistant declines rather than guesses.
- Move from bin day, to what goes in the bin, to a rates concession in one conversation, without the resident choosing an agent.
- Handle four constraints in one facility question — capacity as a number, a day, a feature and an accessibility requirement.
- Show that its clarifying questions come from your indexed facet values, not from a hand-written script.
- Handle an "or" across two different fields without quietly turning it into an "and".
- State which filters were applied, and link to the same filtered results a resident would see.
- Demonstrate that the assistant cannot invent a fee, facility, collection day or permit that does not exist in your data.
- Search events with recurring and skipped occurrences, across two separate time windows in one question.
- Answer a cross-service task from multiple sources, with a citation for each.
- Lodge a service request, with location and context, from inside the conversation.
- Publish emergency messaging or override a specific answer in under five minutes, with no software release.
- Show both automatically discovered themes and governed categories mapped to service areas and wards, each with a named owner.
- Produce evidence on retention and disposal, audit trails, data residency and WCAG 2.1 AA conformance.
Bring your hardest questions. We will type them live.
Most demonstrations are built around the questions a platform can answer. We would rather you brought the ones your residents actually ask — a bin day for a real street, the overlays on a real block, a hall with a capacity, a day and an accessibility requirement, or four service areas tangled together.
No obligation and no procurement process — just a working session on your own content.
Still have questions?
Ask our AI — get cited answers about the council agent team, address look-ups, facility and event search, and community intelligence for councils.