10,000 Fuel Receipts in One Quarter: Why Construction Sustainability Teams Are Drowning in Data Entry

Consider a mid-to-large construction company. Over 10,000 fuel receipts. In a single quarter. From a handful of active project sites.

The sustainability team — two people — manually opening each receipt, finding the litres, recording the fuel type, matching it to a piece of equipment, and entering it into a spreadsheet. Every. Single. One. For over a year.

These are people with environmental science degrees. They didn't take this job to copy numbers from crumpled diesel dockets into Excel. But that's what their job becomes.

Construction carbon accounting automation isn't some nice-to-have upgrade for companies like this. It's the difference between having a sustainability function and having two very expensive data entry clerks.

The maths is genuinely brutal

Here's the thing about construction fuel receipts that people outside the industry don't grasp. A single mid-to-large construction project might run 15 to 30 pieces of heavy equipment — excavators, dozers, graders, compactors, dump trucks, generators. Each one burns diesel. Some get refuelled daily. That's one receipt per piece of equipment per refuelling event.

Multiply by the number of active sites. Add light vehicles — utes, service trucks, site managers' cars. Add generators that run overnight on remote sites. Add the mobile fuel tanker deliveries that come with their own dockets.

10,000 receipts per quarter isn't even extreme. It's just what happens when you have eight or ten active projects running simultaneously.

And each receipt isn't just a number. You need the date. The fuel type (diesel, petrol, sometimes biodiesel blends). The quantity in litres. The vehicle or equipment ID it was allocated to. The project site for allocation. Sometimes the odometer or hour-meter reading if you're tracking fuel efficiency. Some of these receipts are thermal-printed dockets from bowsers that are already fading by the time they hit the sustainability team's desk. Some are handwritten by a subcontractor who filled up at a servo on the way to site.

At 2 minutes per receipt — and that's generous, assuming the receipt is legible and the equipment ID is clear — 10,000 receipts equals 333 hours of manual processing. That's 8.3 full-time weeks of work. Per quarter.

Let that sink in. Two sustainability professionals spending more than half their quarter typing fuel volumes into a spreadsheet.

This is a Scope 1 problem hiding in plain sight

Diesel combustion is the single biggest source of Scope 1 emissions for most construction companies. The NGA Factors workbook puts diesel at roughly 2.71 kg CO2-e per litre (scope 1 only, before you add the 0.2 kg scope 3 component for upstream extraction and transport). A mid-size construction company burning 500,000 litres of diesel per quarter — which is realistic across ten active sites — is generating about 1,355 tonnes of CO2-e in Scope 1 emissions from fuel alone. Per quarter.

That's not a rounding error on someone's ESG report. That's a material emissions source. And the accuracy of that number depends entirely on whether a tired sustainability analyst correctly transcribed 10,000 individual fuel dockets.

Research published in the Journal of the American Medical Informatics Association found manual data entry error rates of 1-4%, with one study finding 3.7% of 6,930 entries contained discrepancies. Apply even a conservative 2% error rate to 10,000 fuel receipts and you're looking at 200 wrong entries. On a portfolio generating over 5,000 tonnes of CO2-e per year from diesel, a 2% data error could shift your reported emissions by 100 tonnes.

Under ASRS, Scope 1 emissions carry full liability from day one. No modified liability protection. No safe harbour provision for getting the numbers wrong. If your NGER or ASRS submission is off because someone misread a faded thermal receipt, that's a compliance exposure — not a data quality footnote.

The opportunity cost nobody talks about

Here's what makes this truly maddening. While your sustainability team is buried in fuel receipt processing, they're not doing any of the work that actually reduces emissions.

They're not analysing fleet fuel efficiency to identify which excavators are burning 30% more diesel than they should. They're not building the business case for electric or hybrid equipment on your next tender. They're not mapping your emissions by project to figure out which site configurations generate the most fuel waste. They're not preparing for the ASRS assurance requirements that are coming for Group 2 entities from July 2026.

They're not doing sustainability work. They're doing data entry.

Australia's public infrastructure pipeline has climbed back to $242 billion over the next five years, according to Infrastructure Australia's 2025 Market Capacity Report. Construction output is projected to grow at 3.3% annually through 2029. Every major builder, every tier 1 and tier 2 contractor, is going to face increasing pressure on emissions reporting — not less. And the industry already has a 141,000-worker shortage on public infrastructure projects alone.

Nobody has spare people to throw at this problem. The answer isn't "hire another sustainability analyst to do more data entry." The answer is to stop doing data entry altogether.

Why construction fuel receipts are harder than utility bills

We've written before about AI document processing for utility bills — electricity invoices, gas bills, water statements. Those are hard enough. But fuel receipts in construction are a different beast.

Utility bills are structured documents from a relatively small number of retailers. AGL, Origin, EnergyAustralia — different formats, but at least they're properly typeset PDFs with consistent layouts. Fuel receipts in construction come from everywhere.

Bowser dockets from site fuel tanks — thermal printed, often partial, sometimes illegible after two weeks in a glovebox. Service station receipts from dozens of different petrol station brands. Fuel delivery invoices from bulk diesel suppliers like Viva Energy or Ampol. Handwritten logs from subcontractors who filled up from a mobile tanker and scribbled the litres on a piece of paper.

Some arrive as phone photos. Some get stapled to timesheets. Some exist only as line items buried in a subcontractor's monthly invoice. It's not unusual to find a shoebox of fuel dockets under someone's desk at the end of the financial year. A shoebox.

And here's the part that kills traditional OCR solutions: there is no standard format. You can't build a template for "Australian diesel fuel receipt" because there are hundreds of variations. The quantity might be labelled "Litres," "Volume," "Qty," or just a number next to a dollar amount. The fuel type might say "Diesel," "ULP," "B20," or just a pump number. Template-based extraction breaks constantly. You'd need a developer maintaining templates full-time — which defeats the whole purpose.

How we think about this differently

We built Carbonly's document processing pipeline specifically because template-based approaches don't work for messy, real-world documents. Our system uses multimodal AI vision to read a fuel receipt the way a human would — looking at the whole document, understanding context and layout, extracting what matters regardless of format.

When you upload a batch of 500 fuel receipts, each one goes through the same 7-phase pipeline. Classification (is this a diesel receipt, petrol receipt, or a lunch receipt someone accidentally included?). Vision-to-text extraction. Structured data extraction — litres, fuel type, date, equipment ID if present. Validation against expected ranges (did someone really put 9,000 litres in a ute?). Normalisation to standard units. Emissions calculation using the correct NGA Factor for that fuel type. And an audit trail linking every calculated emission back to the source document.

The part that matters most for construction companies: it doesn't care about format. A thermal-printed bowser docket gets the same treatment as a properly formatted PDF invoice from Ampol. A phone photo of a handwritten fuel log gets read the same way a human would read it — by understanding what the numbers mean in context.

We won't pretend it's perfect. Handwritten receipts where the ink has smudged are hard for AI too. Our system flags low-confidence extractions for human review rather than guessing. And equipment ID matching — linking a fuel receipt to a specific excavator or truck — sometimes requires project-level context that isn't on the receipt itself. We're still iterating on that piece.

But the difference between "AI processes 9,500 receipts correctly and flags 500 for review" and "a human manually processes all 10,000" is the difference between a 2-day task and an 8-week task. That's not an incremental improvement. That's giving your sustainability team their jobs back.

What your sustainability team should actually be doing

Consider what a construction company's sustainability function could look like if they weren't spending half the year transcribing fuel dockets.

They could be analysing fuel burn rates by equipment type and age to build a fleet replacement business case. They could be identifying which projects use 40% more fuel per square metre than comparable sites — and figuring out why. They could be modelling the emissions impact of switching from diesel generators to battery energy storage on remote sites. They could be getting ahead of the Safeguard Mechanism's declining baselines instead of scrambling to file accurate reports after the fact.

A sustainability analyst in Australia costs between $80,000 and $105,000 per year in base salary. Loaded with super, leave, and overheads, that's roughly $110-$120 per hour. Spending 333 hours per quarter on data entry means you're burning about $36,600 to $40,000 per quarter — over $150,000 per year — on a task that produces zero strategic value. That's before you count the correction cycles when errors surface during assurance.

And if you're still running this on spreadsheets, the problem compounds. Ray Panko's research across multiple audits found that 94% of spreadsheets contain errors. The ANAO's performance audit of NGER reports found 72% of 545 reports contained errors, with 17% having significant ones. Manual processes and spreadsheets aren't just slow. They're unreliable at exactly the point where reliability matters most — when an assurance provider or the Clean Energy Regulator comes asking questions.

Construction companies can't afford to wait on this

ASRS Group 2 reporting kicks in from 1 July 2026. That pulls in entities with $200 million or more in revenue, $500 million in gross assets, or 250-plus employees — and every NGER-registered corporation not already in Group 1. A significant chunk of Australia's Tier 1 and Tier 2 construction companies fall squarely into that group.

Scope 1 emissions from diesel fuel are going to be front and centre in every construction company's climate disclosure. These aren't Scope 3 estimates you can hedge with qualifiers and methodology notes. These are direct combustion emissions from fuel your company purchased and burned. Assurance providers will want to trace every reported tonne back through the calculation chain to a source document. If that source document is a row in a spreadsheet with no link to the original receipt, you've got an audit trail gap.

The companies that figure out construction carbon accounting automation now — while there's still time before their first ASRS reporting period — will have clean, auditable data when it matters. The ones that don't will be hiring consultants at $200-plus per hour to retrospectively reconstruct their fuel records from a shoebox of faded dockets.

We know which side of that we'd rather be on.

If you're processing thousands of fuel receipts per quarter and want to see what automated extraction looks like on your actual documents, talk to our team. We'll run a batch through the pipeline so you can see the output before you commit to anything.

Related reading: