The Best AI Carbon Accounting Software for Australian Government Agencies, Councils, and GBEs (2026)
A machinery-of-government change restates your baseline. A grey fleet claim gives you dollars and kilometres but no fuel. A panel contract worth $400 million arrives as a spend extract with no quantities on it. Here is how the carbon platform options actually stack up against the public sector data shape, and the procurement gate most of them have to clear first.
The hard part of public sector carbon reporting is not the electricity. Electricity is a meter, a bill and a state factor, and once the bills are flowing it stays solved. The hard part is the two billion dollars of procurement sitting in Scope 3 Category 1, where the only artefact you have for most of it is a general ledger extract with a supplier name, an ABN and a dollar figure.
That is the number that dominates a department's disclosure. It is also the number that cannot be traced, cannot be improved, and cannot be defended in front of an auditor who asks a simple question: what actually changed between last year and this year.
Commonwealth entities have been publishing climate-related financial disclosures since FY2024-25 under Department of Finance guidance, against the Australian Public Service Net Zero by 2030 commitment for non-Defence operations. State agencies work to their own targets (Victoria's whole-of-government emissions target, the New South Wales net zero by 2050 commitment, Queensland's 75 per cent by 2035). Councils report to their communities, their ratepayers and increasingly their lenders. Government Business Enterprises get the full AASB S2 treatment on the Group 2 timeline, first reporting years starting 1 July 2026.
Different mandates, one shared problem. So this is an honest look at which platforms fit the public sector data shape. We build one of them (Carbonly), we say so up front, and we are equally direct about where we fall short.
Where public sector reporting actually breaks
Start with Scope 3, because that is where the failure modes live. Everything else in a government inventory is tractable by comparison.
Material ambiguity. A department's largest procurement lines are services: managed ICT, program delivery, labour hire, consulting, facilities management. A road authority's largest lines are asphalt, aggregate, concrete and steel, ordered against panel arrangements where the contract line reads "supply and lay" and the mix design sits in an attachment nobody sent to the sustainability team. You cannot factor what you cannot identify, and "professional services, $18.4 million" identifies nothing.
Unit chaos. Asphalt is ordered in tonnes, priced by the square metre laid, and factored per tonne of binder. Concrete is ordered in cubic metres against factors published per tonne, so somebody has to apply a density before the arithmetic works. Get the conversion wrong on a council's road maintenance program and the error can easily exceed a typical materiality threshold. Our note on unit conversion errors walks through the ones that recur.
Double counting. This is endemic in the public sector because the same tonne passes through several reporting entities. A council weighs waste at the gate, the contractor reports the same tonnage in its own inventory, and the regional waste group reports it again at the aggregate level. A GBE in a joint venture with a private partner can be counted under operational control by the operator and under equity share by both partners. The Scope 3 double counting problem is the same shape in civil works delivered by an agency as it is in construction.
Spend-based estimation destroys the signal. This one deserves more attention than it gets in agency reporting. Convert dollars to emissions with an environmentally-extended input-output factor and your disclosed number tracks your budget, not your carbon. Switch a panel supplier to lower-carbon concrete at the same delivered price and your reported emissions do not move by a single tonne. The department did the work and the report cannot see it. Worse, a budget increase looks like an emissions increase, which is an awkward line to explain at Senate Estimates. The spend-based accounting piece sets out where the method is defensible and where it is not.
Suppliers send evidence, not data. An agency that writes climate clauses into a major contract gets compliance, and compliance arrives as PDFs. Environmental Product Declarations for materials, scanned waste transfer notes, a fuel summary as a spreadsheet screenshot in an email. None of that is a data feed. Somebody has to turn it into ledger records, and at agency scale that somebody is either a graduate or a piece of software.
Traceability for assurance and for the public. Every other sector worries about one auditor. An agency worries about the auditor, the Australian National Audit Office, a Freedom of Information request, a question on notice, and a journalist comparing this year's table to last year's. A restatement that is routine in a listed company becomes a story when it is a government number. Data lineage is the whole job here, not a feature.
Sitting on top of all that are three structural facts about the public sector that, in our view, platforms built for private companies rarely put first.
Machinery of government changes restate your baseline whether you like it or not. Functions move between departments. A program transfers on an administrative arrangements order and takes its staff, its tenancies, its fleet and its supplier contracts with it. The receiving department inherits emissions it never measured; the losing department has to restate its own prior year so the comparative means something. This happens mid-year, on short notice, and it is not an edge case. If your carbon system cannot re-cut a boundary retrospectively and keep the original disclosed figures intact alongside the restated ones, you will be rebuilding the model every time the ministry changes shape.
Grey fleet is a distance claim, not a fuel purchase. A pool car gives you a fuel card and litres. An employee driving their own vehicle for work gives you a reimbursement line: a dollar value and a kilometre count, at the ATO cents-per-kilometre rate. No fuel quantity. No vehicle. No fuel type. You either apply a distance-based factor by assumed vehicle class and record the assumption where an auditor can find it, or you spend-base it and lose the signal entirely. Most agencies do the second thing without writing down that they did it.
The procurement and security gate comes before the product comparison. This is the difference that catches vendors out. The Protective Security Policy Framework and the Information Security Manual set the baseline for Commonwealth entities, states run their own equivalents, and the Hosting Certification Framework applies to certain workloads. Some agencies require an IRAP assessment before a SaaS product can be approached at all. Add the Commonwealth Procurement Rules value-for-money obligation, plus state panels like BuyNSW and the Victorian eServices Register, and a good part of the market can be out of contention before anyone has looked at a screenshot.
Sort your own shortlist by that gate first. It is cheaper than discovering at contract stage that the data sits offshore.
The five options worth evaluating
The order is not a quality ranking. It reflects our view, as the vendor of one of these products, weighted on Australian regulatory fit, evidence traceability under assurance, and cost that can be scoped to a defined piece of work. An agency that weights security certification above everything else will order this list differently, and should. Competitor descriptions summarise each vendor's public materials as at September 2026. We have not run their platforms, so treat them as a starting point for your own diligence.
1. Carbonly
Australian-built. The reason to look at it for a public sector footprint is what it does with documents rather than what it does with spend.
Start with the panel contract problem. Where a supplier sends an invoice, a delivery docket or an EPD, the document engine reads it and pulls the physical quantity: tonnes of asphalt, cubic metres of concrete, litres of diesel, kilowatt-hours of consumption billed on behalf of the agency. Those records land on the ledger as activity-based Scope 3 with the source document attached. Where only a spend extract exists, the record is spend-based and is floored to low data quality, so the disclosure can show how much of Category 1 is estimated and how much is measured. That split is a number an auditor is likely to be interested in.
For machinery of government changes, period locking and restatement reasons mean the original period stays locked and any restatement is recorded with its reason. Joint venture consolidation runs under operational control, financial control and equity share, which is what a council with a shared waste facility or a GBE with co-owned infrastructure needs in order to answer the double counting question rather than argue about it.
Pool fleet comes in through the document engine: a fuel card export or a pool vehicle report lands as a source document. For grey fleet reimbursement lines that carry only dollars and kilometres, decide the vehicle class assumption deliberately and record it where an auditor can find it.
Activity data is matched to NGA factors, and electricity takes the state factor for its location: a Canberra tenancy uses the NSW and ACT factor of 0.64 kg CO2-e per kWh, a Melbourne office Victoria at 0.78, and a Darwin depot the NT Darwin-Katherine figure of 0.56. Each record stores the emission factor version applied, so when the workbook updates the prior periods keep the version they were calculated on. Emissions are calculated on AR5 global warming potentials, the set NGER prescribes in regulation 2.02 of the NGER Regulations, where methane carries a single value of 28 rather than a split fossil and biogenic pair. The current state electricity factor table is published free on this site if you want to check a number without installing anything.
NGER facility and corporate-group threshold status is computed from the ledger rather than a spreadsheet somebody maintains: 50,000 tonnes CO2-e at the corporate group level, 25,000 tonnes at the facility level. Relevant for a GBE, and relevant for any council operating a landfill of consequence. The corporate group aggregation rules are where most public sector entities get the boundary wrong.
Data quality checks are deterministic, not a model guessing. Threshold breaches, z-score outliers against a site's own history, rolling-baseline trend breaks, missing periods, duplicate documents and emission factor drift. Every flag has a rule behind it that a reviewer can read, which is the kind of control that can be explained in an FOI response.
Contractors can submit their own data through the Supplier Portal, scoped per project. Records are retained for seven years under internal policy, ledger writes carry typed audit events, and an evidence pack is produced for assurance.
Where it fits. Agencies, councils and GBEs whose Scope 3 procurement tail is the disclosure risk, and who expect to have their numbers read line by line in public. It is strongest where source documents exist but nobody has the hours to process them.
Honest gaps. No IRAP assessment completed as at September 2026, which rules us out of some agency procurements outright, and we would rather you know that now than at contract stage. No one-click Commonwealth Climate Disclosure generator: we hold the ledger, the disclosure is drafted from it. No direct submission to the Global Covenant of Mayors portal for councils. No Climate Active submission integration. No PCAF financed-emissions engine for GBEs with investment portfolios. We also have no paying customers and no production pilots as at September 2026, and the document engine has been tested on Australian supplier document formats, not against a live agency portfolio.
Pricing. Per project, with a $100 per month workspace minimum. The reason that shape suits an agency is that the unit of cost is a defined piece of work you can name in an approach to market and document against the value-for-money test, rather than a seat count that moves every time the department does.
2. IBM Envizi
Worth putting second for a public sector reader rather than fourth, because the property-heavy shape of an agency or council portfolio is close to what this product grew out of. On public reporting, Envizi began as a utility bill and portfolio energy data business before becoming part of IBM Sustainability Software, and that lineage shows in how it handles meters, sites and bill data at volume.
Where it fits. A department or large council whose dominant workload is hundreds of buildings and thousands of bills, particularly one that already holds an IBM relationship and can procure through an existing arrangement. Procurement path matters in government, and an existing panel arrangement is a real advantage.
Honest note. Ask where the data sits and under which jurisdiction, and ask whether any Australian Government hosting certification is held. Then ask the Australia-specific questions: NGA factor pinning by year, the NGER category structure, and how a machinery-of-government restatement is recorded. As at September 2026 we could not find public documentation of native support for those, which does not mean it is absent, only that it is not published. Put the questions in the approach to market.
3. Watershed and Persefoni
Two international enterprise platforms, both credible at large footprints, both with public positioning built around heavy procurement Scope 3 programs. That is a genuine fit for the agency problem described above, and for a GBE with a multinational structure or an existing global sustainability function it may be the pragmatic answer. Neither publishes a public price list. Publicly reported figures suggest Watershed runs somewhere in the range of $37,000 to $264,000 per year depending on scope; confirm current pricing with the vendor. Our alternative to Watershed and Persefoni analysis goes further into the fit test.
Honest note. For a Commonwealth entity, residency and certification are the first gate and the rest of the evaluation is academic until they are answered in writing. After that, ask how supplier-provided activity data and spend-based estimates coexist on the same record, and whether the disclosure output flags the method per contract. Auditors ask. So do committees.
4. NetNada
Australian-built, with public materials describing integrations into the accounting stack (Xero, MYOB, QuickBooks Online) and a financial-data pathway where coded transactions flow into the emission engine. Where it fits. Smaller councils, statutory authorities and government-adjacent not-for-profits where the operational cost base is coded consistently in a small business accounting system and the site count is modest. For that shape it is a fast start, and a fast start has real value when the first disclosure is due.
Honest note. The pathway depends on the general ledger carrying usable detail. In a department running SAP or TechnologyOne, or a GBE on S/4HANA, the posting is frequently a bulk-coded journal with no meter number, no litres and no tonnes on it. That is the same signal-loss problem described earlier, arriving through a different door. As at September 2026 we could not find public documentation of quantity-level extraction from supplier PDFs in NetNada's public materials, so ask the team directly how physical quantities are captured on the large panel contracts.
5. A consultant-built model
A common answer in the sector, and for a first disclosure it is often the right one. A consultant is engaged for a reporting year, builds the model, delivers the report, and either closes out or continues annually.
Where it fits. A first cycle where the reporting perimeter is genuinely unsettled, or a small statutory body with fewer than ten sites. Also a sensible bridge while a platform decision goes through a proper value-for-money process under the Commonwealth Procurement Rules.
Honest gap. A spreadsheet model may not retain the source PDFs in a form that lets someone trace a disclosed number back to a bill two years later, which an auditor or an FOI applicant may want. And it does not absorb a machinery-of-government change without a rebuild. The important point about this option is that it is not a competing category: consultants are the people who run most public sector climate engagements, and a platform is the equipment they work with. A practice running six agency engagements still owns the materiality judgement, the scenario work and the board conversation. See our consultant cost versus software comparison for the economics both ways.
The questions that separate a fit from a mismatch
Five, and they are all answerable in a procurement process.
Where does the data physically sit, who are the sub-processors, and what certifications are held? Get it in writing before the demo, not after. If IRAP assessment is a hard requirement, say so in the approach to market and save everyone the cycle.
What happens on a machinery-of-government change? Ask the vendor to demonstrate moving a program between reporting entities mid-year, restating the prior comparative, and producing both the originally disclosed figure and the restated figure with the reason recorded. If the answer involves re-importing data, that is your answer.
How much of Scope 3 Category 1 will be activity-based, and can the disclosure show the split? For a $2 billion procurement spend this is the single most important question on the list. A platform that cannot tell you what proportion of your largest emissions line is measured rather than estimated cannot help you improve it.
How is grey fleet handled? Ask specifically what the system does with a reimbursement line that carries dollars and kilometres. The right answer names the factor, the vehicle class assumption and where that assumption is recorded.
Can one data set be cut by financial year, by program, by portfolio and by voluntary framework without re-ingestion? Agencies report the same operations several ways: annual report, Portfolio Budget Statement programs, ministerial commitments, and for councils a voluntary framework on top. One ledger, many cuts, or you will be reconciling forever.
How to run the evaluation
Take five sites that represent the real spread. For a department: head office, two service centres in states with very different grid factors so the per-site selection is actually tested, a leased regional office, and one specialist facility. For a council: town hall, aquatic centre, depot, library and a sportsground with its own meter.
Load a quarter of bills and fleet data. Then, and this is the test that matters, load one panel contract's worth of supplier documents. Not the spend extract. The actual invoices and dockets, with their attachments, in whatever state they arrive in. Ask the vendor to show you how many of them produced a physical quantity, how many fell back to spend, and how many needed a human.
That ratio is the product. Everything else on a demo screen is presentation.
FAQ
Where is Carbonly data hosted, and does that meet Australian Government requirements? We can walk through the hosting arrangement and sub-processor list under NDA. No IRAP assessment has been completed as at September 2026. If your procurement requires one, we are not a candidate yet and we would rather tell you early.
Can Carbonly produce the Commonwealth Climate Disclosure output Finance requires? Not as a one-click generator. The ledger, the category tagging and the evidence trail sit in the platform, and the disclosure is drafted from that. The narrative involves policy judgement that belongs with the department, not with software.
How does Carbonly handle Scope 3 for an agency with a $2 billion procurement spend? Two paths that coexist on the same ledger. Supplier documents carrying physical quantities become activity-based records with the document attached. Everything else runs spend-based on published input-output factors and is flagged as such, so the proportion of Category 1 that is estimated is visible rather than buried. Over time the mix should shift as climate clauses in major contracts start producing real data.
Does it support Global Covenant of Mayors or other council reporting portals? The underlying data supports what those frameworks ask for. There is no direct submission integration. The workflow is to manage the operational ledger in Carbonly, export the cuts required, and submit in the relevant portal.
How does pricing work for an agency, council or GBE? Per project, with a $100 per month workspace minimum. There are no per-seat fees. Actual pricing depends on how the work is split into projects, which in turn depends on site count, fleet size and the Scope 3 scope you take on. Email hello@carbonly.ai and we will scope it against your actual portfolio.
Related reading
- Carbon accounting for Australian local government and councils
- Scope 3 hub: the 15 categories, and the ones agencies get wrong
- Spend-based emissions accounting: where it works and where it fails
- Emissions data lineage and record traceability
- NGER corporate group aggregation rules
- NGER reporting hub
- NGA 2025 state electricity factors, free reference table
Related articles
- The Best AI Carbon Accounting Software for Australian Retail Chains (2026)
- The Best AI Carbon Accounting Software for Australian Utilities and Energy Retailers (2026)
- The Best AI Carbon Accounting Software for Australian Logistics and Freight Companies (2026)
- Australian Alternative to Watershed and Persefoni for AASB S2 and NGER