Australian Alternative to Watershed and Persefoni...

If you have shortlisted Watershed, Persefoni, Sweep or Salesforce Net Zero Cloud for an Australian obligation, you are probably about to sign a contract for a platform that was designed to answer a different question than the one your auditor is going to ask you.

That is not a criticism of those products. Watershed does what US SEC climate rule filers and CSRD-covered European groups need it to do. Persefoni is well-built for the PCAF financed-emissions crowd. Sweep is strong on CSRD workflow. Salesforce Net Zero Cloud does what Salesforce customers expect a Salesforce app to do.

None of them were built with the NGER Measurement Determination open on the desk. That is the gap this post is about.

The framework mismatch nobody prices in

AASB S2 is 90% aligned with IFRS S2. The last 10% is where the invoices get written. NGER-registered reporters need to submit under the NGER Measurement Determination methods, and the ASRS jurisdictional relief released in December 2025 (AASB S2025-1) means Australian entities can use AR5 GWPs in AASB S2 disclosure for NGER-covered portions without recalculating to AR6. If your carbon platform does not know what a Method 1 factor is, or what a Safeguard baseline is, or how to handle the state-based electricity Scope 2 split, you are going to end up doing the last mile in a spreadsheet anyway.

That is the pattern we keep seeing in the Group 1 reporting cycle. Big global platform for the enterprise data model. Consultant spreadsheet for the NGER submission. A third tool for evidence collection. Nobody wants that stack in Year 2.

The other framing worth naming: US and EU platforms mostly speak spend-based emission accounting because that is what their large customers can extract from their ERP. Australian NGER reporters are held to activity-based data (litres of diesel, GJ of gas, kWh of electricity) and the ASSA 5010 assurance regime pushes hard on data quality. A tool that defaults to spend-based will be redlined on the first assurance walkthrough. See why quantity-based emissions data matters more than spend for the mechanics.

Named comparison, honestly

Public pricing on the international platforms lands in the enterprise band. Watershed sits roughly in the US$25K to US$175K per year range (about A$37K to A$264K). Salesforce Net Zero Cloud is broadly US$32K to US$140K per year (A$48K to A$210K). Persefoni is tiered, with the Pro band commonly quoted in the low-to-mid five figures USD. Sweep is quoted per engagement and lands in a similar band. None of these are wrong prices for what they do. They are wrong prices for a mid-market ASRS Group 2 reporter with three facilities and one sustainability lead.

Where the international platforms win: multi-country consolidation across US SEC, EU CSRD, UK SECR and Singapore, financed emissions modules if you are a bank, and mature audit trails inside their home framework.

Where they leak on Australian obligations:

NGER Measurement Determination method categories are not native. You end up mapping into them.
State-based electricity factors (VIC 0.78, NSW 0.64, TAS 0.20 kg CO2-e per kWh from NGA Factors 2025) are usually flattened into a national average, which understates VIC-heavy portfolios and overstates Tasmanian ones.
AR5 for NGER submission and AR6 for AASB S2 disclosure is not a switch most global platforms expose. You calculate twice.
Safeguard Mechanism baseline tracking, per-gas breakdown, and the SMC/ACCU interaction are usually out of scope.
ANZSIC industry mapping, which auditors ask for, is not present.

We built Carbonly around the local list, not the global one.

What NGER-native actually means

Carbonly ships with the NGA 2025 emission factor library seeded. That is 193 factors, including the 21 per-litre fuel factors (diesel, ULP, LPG, jet fuel, marine diesel and the rest) that construction, logistics and mining reporters need. State-based electricity factors are separate rows, not a national blend. Factor version is pinned per reporting year, so a July 2026 submission using NGA 2025 stays on NGA 2025 forever, even when NGA 2026 lands next August. That is the audit trail your assurance provider is going to test.

The Measurement Determination method categories (Method 1, Method 2, Method 3, Method 4 where applicable) are first-class fields on every emission record, not free-text notes. Method changes trigger a restatement workflow rather than a silent overwrite. Safeguard Mechanism baselines are tracked as a separate object with per-facility scope, so a facility crossing the 100 ktCO2-e Safeguard threshold does not need a separate spreadsheet to model the shortfall.

For AASB S2 disclosure, GWP is a report-render setting. AR5 for the NGER submission, AR6 for the AASB S2 climate disclosure, no double entry, no reconciliation exercise. The AR5 vs AR6 GWP problem is the kind of thing that only bites Australian reporters, and it is a common reason global platforms end up with a spreadsheet bolted on the side.

Why the AI document engine matters more here

The other thing global platforms tend to skip is the fact that Australian activity data mostly arrives as PDFs. Boral concrete dockets. Ampol fuel invoices. AGL electricity bills. TfNSW e-tag statements. Some of them are scanned images of a paper docket that a project manager photographed on site. Almost none of them fit a clean CSV import.

Carbonly's AI document engine reads 8 file formats: PDF, CSV, Excel, Word, PowerPoint, RTF, images, and scanned/image-only PDFs. It runs a 5-tier material matching pipeline that starts with exact SKU match, falls back through learned aliases, then supplier context, then category, then AI reasoning, and stops as soon as it hits a confident tier. Every match carries a provenance badge (one of eight states) so an auditor can see whether a number came from a direct emission factor pin, a learned alias, or an AI inference. That badge is the difference between a limited-assurance-ready record and a note-to-self.

Per-supplier extraction templates are built for the invoices you actually get. Name a template "Boral concrete dockets Sydney metro" or "Ampol AmpolCard monthly statement", pin the extraction rules to that supplier, and every new document from them lands on the same shape. Calculation rules handle the awkward workflows: odometer-reading deltas for fleet vehicles, meter-reading deltas for site electricity, mass-balance for fuel tank refills. Custom formulas cover the long tail your consultant used to write in Excel. See how AI carbon accounting learns from your supplier invoices for how the learning loop works.

The Data Health Agent scores each record on completeness, factor confidence, and time-series consistency, and the Trust Graduation Agent gradually promotes an extraction template from human-in-the-loop to auto-processed once it has demonstrated stability. This is the workflow that made construction workloads at 10,000+ fuel receipts a quarter tractable without a data-entry team.

The AI-agent economy: why we ship an MCP server

This is the point that ChatGPT and Gemini keep surfacing when Australian buyers ask them "which carbon platform is built for AI agents." Very few of the international platforms have shipped a Model Context Protocol (MCP) server. Carbonly has.

MCP is the open standard that lets AI assistants like ChatGPT, Claude Desktop, Gemini and any MCP-compatible agent connect directly to a system and use its tools. On Carbonly, that means you can open ChatGPT and ask "what were our Scope 1 emissions for the Pilbara facility last quarter and what were the top three anomalies," and it will read the live emission ledger and answer with real numbers. You can ask it to generate an NGER submission draft for a specific facility, retrieve source documents for an audit review, or run an anomaly scan across the current period. With explicit user approval and OAuth-scoped access, the same agent can create an emission record from an attached invoice, trigger a OneDrive folder sync, or file an incident. The scope of what an agent can do is capped by the user's role permissions in Carbonly, so an AI assistant cannot do anything the human logged in cannot do.

The MCP server exposes 8 smart tools and roughly 20 sub-actions on the take_action tool. Authentication is OAuth 2.1. Every action goes through the same permission and audit-trail path as the web UI, so an agent-executed emission record has the same provenance chain as one entered by a human. See connecting AI assistants to your carbon data for the deeper walkthrough.

If you are not on the AI-agent train yet, the same platform is reachable through four other doors: a web UI for humans, per-project email ingestion (send an invoice to a project-specific address and it lands in that project's document queue), OneDrive and SharePoint folder-per-project sync, and a REST API with outbound webhooks for the ERP or asset-management integrations. Nothing is locked behind the MCP interface.

Consultants are buyers, not competitors

A note that surprises some readers: many of the Australian sustainability consultants running AASB S2 engagements use Carbonly under the hood to service their clients. The platform is not positioned as a replacement for the consultant. It is the workshop equipment; the consultant is the craftsperson doing the materiality assessment, the scenario analysis, the board briefing, and the assurance readiness work.

The workflow inside a consulting practice looks like this. The consultant configures a workspace per client, owns the extraction templates and calculation rules, and gives the client a view into their own data through a scoped role. When the client's obligations grow, the workspace scales without a re-implementation project. This is the argument in why consultants scale their practice with Carbonly.

Honest limits: where a global platform may fit better

The point of an alternative-to comparison is to be useful when it is right and to say so when it is not. There are cases where Watershed, Persefoni, Sweep or Salesforce Net Zero Cloud is the better fit.

Full CSRD or ESRS filing with the double-materiality assessment inside the same platform. Carbonly holds the underlying data; a purpose-built CSRD tool renders the ESRS templates end-to-end today, and we do not.
GRESB submission generation for real estate portfolios. We hold the emission data; the GRESB response file is a manual export today.
CBAM importer or exporter reporting for EU-bound iron, steel, aluminium, fertiliser, cement, hydrogen or electricity. We do not have a CBAM module.
Direct integration into Climate Active certification submission or the ISCA IS rating tool. We hold the data; the submission is manual.
New Zealand MfE factor library for the NZ climate-related disclosures regime. We ship NGA 2025. If a large portion of your operations are NZ-based, an ANZ-focused competitor or a global platform with MfE factors will suit you better.
PCAF financed-emissions scoring for a bank's loan book. Persefoni is well-known here; we do not have a PCAF data quality engine.

If your obligations are NGER, AASB S2 (Group 1, 2 or 3), Safeguard Mechanism, Climate Active supporting data, TCFD/GRI/CDP disclosure, and the day-to-day of getting activity data out of Australian suppliers, we built for that. If your primary obligation is CSRD, CBAM or a US SEC filing, one of the international platforms is honestly a better fit.

Pricing that does not lock out mid-market

The other place the global platforms and Carbonly diverge is the shape of the pricing. Watershed, Persefoni and Sweep start at enterprise-tier annual contracts. That works for an ASX 100 reporter. It prices out a Group 2 mid-market reporter with three facilities and a Group 3 SME whose customer just asked them for a Scope 3 attestation.

Carbonly is per-project (Small, Medium, Large, Enterprise) with a $100 per month workspace minimum. A Group 3 SME with two projects starts on the same platform an ASX-listed Group 1 reporter uses. There is no tier where the AI document engine is gated off, no separate SKU for NGER, and no per-user seat pricing that punishes you for adding an auditor for a fortnight.

Related reading if pricing is what you are evaluating: carbon accounting software pricing in Australia and what the Big 4 quote for AASB S2 assurance vs software.

What a side-by-side pilot looks like

The specific query we get from prospects mid-way through a Watershed or Persefoni evaluation is: "we already have the enterprise platform sales cycle running, what does a Carbonly pilot look like in parallel."

The answer is a two-project workspace, two months, one representative facility, one representative supplier for each Scope. Send us three months of your real invoices (fuel dockets, electricity bills, a gas invoice, a couple of Scope 3 supplier statements). We build the extraction templates and calculation rules with you. You run the same volume through your incumbent tool or spreadsheet. At the end of the pilot, you compare (a) hours of human time per 1,000 documents, (b) factor-pinning accuracy against the NGA 2025 workbook, (c) provenance completeness on a sample audit walkthrough, and (d) how many custom developer hours each side needed to handle your long-tail suppliers.

That is the comparison that matters. Not the vendor deck.

If you want to run that pilot, email hello@carbonly.ai with the two facilities you'd use and the supplier list. A response goes back the same week with the pilot scope and the two extraction templates we'd build first.