Model fast with AI. Get a number that holds up.
Build with AI drafts a full connected system model from your description and your documents in 30–60 seconds. Every node arrives unapproved until a person reviews it — and third-party verification is what makes the number defensible.
- Human approval on every node
- Every suggestion explained
- Third-party verification, not AI sign-off
A CarbonSig System Builder model at carbonsig.app, just drafted by Build with AI. The system title carries an AI estimate of 1.9 kg CO₂e per kg, shown for reference only. Four nodes — Iron ore (input, Scope 3), Blast furnace (process), Furnace combustion (direct emission, Scope 1) and Steel billet (output) — are all Draft: they have AI-generated data that nobody has reviewed yet. Reviewing and approving a node turns it green. Illustrative model.
Illustrative model; the title estimate is for reference only.
How does CarbonSig use AI in carbon accounting?
CarbonSig uses AI to speed up the modelling work, not the judgement. Build with AI drafts a full connected system model from your description and documents in 30–60 seconds, and AI suggests ranked emission factors with the reasoning behind each one. Every AI output arrives unapproved until a person reviews it, and an accredited verifier reviews the registered model.
Modelling is the slow part — and it needs an expert.
A product carbon footprint starts as a blank canvas. Someone has to know the process well enough to lay out every input, process and emission source, then find a defensible emission factor for each one. Traditional consulting-led projects run 6–10 months and $50,000–$200,000 per project.
-
A blank canvas
Every input, process, direct emission and output has to be laid out and connected before a single number exists.
-
The emission-factor hunt
Then each input needs a factor you can defend — searched out of the Reference Emission Data Library by hand, one input at a time.
-
Gaps stall everything
Until the data is in, there is no number to react to — and no clear view of which gaps actually move the result.
Describe the process. AI drafts the system.
Build with AI is in the Models section, on the Templates page. You describe what you make; it returns a connected system model you can work on.
The Build with AI form asks for a system name, a detailed description of the process, optional supporting documents — up to 5 files of 20 MB each, such as bills of materials, EPDs, specifications and drawings — and an optional setting to show an estimated carbon footprint in the system title, for estimation purposes only.
A first draft of the whole system, in under a minute
AI reads your description and your documents and drafts a connected system model — inputs, processes, direct emissions and outputs — in 30–60 seconds. It opens straight into the Builder, where you rename, reshape, add and remove exactly as you would with a model you built by hand.
The more specific you are, the closer the draft lands: name your sector and location, your main inputs and energy flows, and the standard you report to. Attach what you already have — a bill of materials, an EPD, a specification, a drawing.
It is a draft, and CarbonSig says so. Every node arrives Draft — orange — which in the Builder means “has data, needs your review.” Nothing AI produces is treated as reviewed until a person approves it.
Stop hunting for factors. Review a ranked shortlist.
An input in the Builder, Iron ore, with the “Ask AI to suggest emission” action. A ranked shortlist is returned; here it holds three: Iron ore concentrate at mine, from Ecoinvent, global, high confidence; Iron ore pellets, from Ecoinvent, Europe, medium confidence; and Iron ore, unspecified, from an EPD International record, global, low confidence. Each carries a “Why This EF?” explanation report. Illustrative suggestions.
Ask AI on any input — then check its work
On an input that needs emission data, ask AI to suggest it. AI reads the input name, your notes and any documents you uploaded, and searches the Reference Emission Data Library and the Emission Data Marketplace — Ecoinvent, IPCC, US EPA and EPD International, 40,000+ emission factors.
You get a short ranked shortlist in 2–5 seconds. Each suggestion is scored High, Medium or Low confidence and shows its database, its geography and its source, so you can see at a glance which are worth a second look.
“Why This EF?” opens the reasoning: what matched in the name, how close the geography is, where the data came from and how fresh it is. That report is the answer to a verifier’s “why this factor?” — on the record, per input.
Accept a suggestion or override it. The choice, and the reason for it, stays yours.
An estimate to work with — labelled as an estimate.
Real data arrives late and in pieces. AI keeps the model moving in the meantime, and is explicit about what is still an estimate and what is still missing.
AI carbon estimation
When you build with AI you can ask it to show an estimated carbon footprint in the system title — for estimation purposes only. It is there for scenario work, early assessments and workshops, so a conversation can start before every figure is in. It is not a reporting number, and it never presents itself as one.
Reference and comparison only. Reporting needs complete data entry, human review and third-party verification.
System Status Report
To see what is actually missing, run the System Status Report: an AI gap report in 30–90 seconds covering the health of the model, the gaps in it and the next steps — and shareable with whoever owns the data you are waiting on.
Gaps named and prioritised, so the review effort goes where it changes the result.
AI assists. People approve. A third party verifies.
Speed is only worth having if the number survives scrutiny. Every AI contribution to a CarbonSig model is labelled, reviewable and attributable — and none of it is verification.
-
AI drafts
Draft, and marked as draft
Build with AI creates the structure and populates emission data. Every node arrives Draft — orange — and each suggested factor carries its “Why This EF?” report.
-
You approve
A person reviews every node
Open each node, check the factor, the geography and the quantity, read the AI’s reasoning, then approve. Reviewed nodes turn green. Nothing skips this step.
-
Verified on CarbonSig
An independent verifier confirms it
Register the model and an independent, accredited verifier reviews that locked snapshot round by round, raising findings pinned to the exact node, through to a formal opinion.
CarbonSig’s AI does not verify anything, and neither does CarbonSig. Verification is performed by an independent accredited third party, in the Verifier Hub, against the standard you report to — that is the step that makes the number hold up.
Faster to a model. Faster to a number a verifier can review.
AI takes the blank canvas and the factor hunt off your desk, so expert time goes where expertise actually pays: reviewing, correcting, defending. What comes out is a model an independent verifier can work through — because a person approved every node, and every suggestion arrived with its reasoning attached.
Then a person approves every node, and an independent accredited verifier reviews the registered model.
CarbonSig AI, answered.
Does CarbonSig’s AI verify my carbon footprint?
No. CarbonSig’s AI drafts your system model and suggests emission factors, and every AI output arrives unapproved until a person reviews and approves it. Verification is a separate, independent step: an accredited third-party verifier reviews your registered model round by round and issues the formal opinion — “Verified on CarbonSig.” CarbonSig is a pre-verification platform.
How does Build with AI create a system model?
You give Build with AI a system name and a detailed description of the process — sector and location, main inputs, energy, outputs, and the standard you report to — and you can attach up to 5 supporting documents of 20 MB each. AI drafts a full connected system model in 30–60 seconds and opens it in the Builder for you to refine.
How does AI suggest emission factors, and can I see its reasoning?
On any input you can ask AI to suggest emission data. It returns a short ranked shortlist in 2–5 seconds, each suggestion scored High, Medium or Low confidence, drawn from the Reference Emission Data Library and the Emission Data Marketplace — Ecoinvent, IPCC, US EPA and EPD International, 40,000+ emission factors. “Why This EF?” shows the reasoning behind a suggestion: what matched, the geographic fit, and the data source and its freshness.
Can I use AI-estimated numbers for compliance reporting?
No. An AI carbon estimate is for reference and comparison — scenario work, early assessments, workshops — and CarbonSig labels it as an estimate. For reporting and compliance you complete the data entry, have a qualified person review and approve every node, and then have the result independently verified by an accredited third party.
What does an AI-generated node look like in the Builder?
Orange. In CarbonSig, orange means Draft: the node has data, but nobody has reviewed it. Everything Build with AI creates arrives Draft, and a node only turns green — Reviewed — once a person opens it, checks the emission factor and the quantity, and approves it. Red means Incomplete: no data yet.
See it draft your system, then pull it apart.
Bring a product and a description of how you make it. We will build it with AI in the session, and you can check every node it produces.