Field worker
Kiln manager, yard supervisor, store keeper. Often limited reading, shared phones, weak network.
- Records daily activity by tap, voice or photo
- Hears every screen read aloud
- Sees one simple monthly result
High-level product documentation · v0.1
Carbon accounting a kiln worker can do by voice
A multi-company app that turns what workers already know (sacks of coal, empty gas cylinders, meter readings) into a GHG inventory that managers, buyers and auditors can trust.
- Brick kilns + ship-breaking first
- Bangla-first, works offline
- GHG Protocol · ISO 14064-1
Worker says“আজ বারো বস্তা কয়লা গেছে”
App understandsCoal · 12 sacks · 600 kg
Inventory recordsScope 1 · Stationary combustion · tCO₂e
Who uses it
Three kinds of people, three surfaces
Each person sees only what they need. The person entering data never has to understand carbon; the person reading the report gets full scientific detail.
Company admin
Sustainability lead or office manager at each company.
- Sets up sites, workers and activity types
- Chooses units, conversions and emission factors
- Reviews AI flags and approves data
Management, buyers, auditors
Owners, international brand buyers, banks, third-party verifiers.
- Reads Scope 1–3, categories, gases, intensity
- Tracks progress against a reduction target
- Exports a verifiable inventory with evidence
How it works
From a spoken sentence to a verified inventory
Every entry passes through the same six stages, in this order. Scroll through them.
Capture
The worker records today's activity in whichever way is easiest: tap a number with big + and − buttons, say everything in one sentence, or photograph a receipt, meter or pile of cylinders. Entries save on the phone first, so no network is needed.
Understand and confirm
AI turns speech into items, reads numbers off receipts and meters, and counts objects in photos. It never saves on its own: it shows what it understood, reads it back aloud, and asks a two-choice question when unsure. The worker confirms.
Convert to base units
Local units become standard ones using the conversions each company sets: sacks to kilograms, maunds to kilograms, cans to litres, meter readings to kWh used today. The original entry is kept beside the converted value.
Calculate and classify
Activity data is multiplied by a versioned emission factor and converted to CO₂e with global warming potentials. Each result is tagged with its scope, source category, ISO 14064-1 category and gas.
Check and approve
AI flags unusual values (coal use triple last week's, a meter going backwards). The admin sees each flag with its photo or voice evidence and approves or corrects it. Every change is logged for the audit trail.
Report
Approved data rolls up into two views: one plain result for workers (this month vs last month, in Bangla), and a full GHG inventory for management and auditors with scopes, categories, gases, intensity and evidence.
Demo · a morning at the kiln
One daily entry, screen by screen
A morning at a brick kiln, from opening the app to the monthly result.
Scroll to walk through one daily entry, screen by screen. The green ring shows where the worker taps next. Under each step: what the system does behind the scenes.
- Step 1 · Home
The worker opens today's checklist
Six tiles, one per activity type at this kiln. Coal and bricks are already done today; the worker taps Coal to update it. Or uses one of the two big buttons to speak or photograph.
Behind the scenesThe app loads this site's activity types from the company's configuration and shows which ones still need an entry today.
- Step 2 · Tap entry
Twelve sacks, counted with + and −
The question is one short Bangla sentence with a speaker button. The worker picks sacks as the unit and taps + until it says ১২, then taps Next.
Behind the scenes12 sacks × 50 kg (the company's sack weight) = 600 kg coal. The original entry and the converted value are both stored.
- Step 3 · Or by voice
Or say the whole day in one sentence
Instead of tapping each tile, the worker can speak: “twelve sacks of coal, twenty litres diesel for the generator, two vans of husk.” AI lists what it understood and asks about the item it doesn't know.
Behind the scenesBangla speech recognition, then item and unit extraction with a confidence score. “Husk” has no type yet, so it is kept aside and the admin is told.
- Step 4 · Proof
A photo of the receipt as evidence
The worker photographs the coal receipt. The amount is hard to read, so the app asks a two-button question instead of making anyone type.
Behind the scenesDocument reading pulls quantity, date and seller. The photo is stored as evidence and linked to the entry for audit.
- Step 5 · Review
Hear it back, then confirm
Everything for today in one list, with a pencil on each line. The worker taps the speaker to hear it read aloud, then confirms with one large button.
Behind the scenesNothing counts until this confirmation. Each entry records how it was made (tap, voice, photo) for the evidence share in reports.
- Step 6 · Saved
Saved, even with no network
The screen says plainly that there is no internet right now and that the entry will send itself later. The worker doesn't need to do anything else.
Behind the scenesThe entry waits in the phone's local database. On sync the server converts units, applies versioned emission factors and GWP, and classifies it: Scope 1 · Stationary combustion.
- Step 7 · Result
The month in one sentence: four tonnes less
At any time the worker can see this month against last month, where most carbon comes from, and one tip from the company. The same data feeds the full GHG report for managers.
Behind the scenesApproved entries roll up by month, site, scope, category and gas. The admin has already reviewed any entries AI flagged as unusual.
The calculation
One formula for every industry
The engine never changes per industry. Only the activity types, units and factors change, and those are configuration.
ECO₂e=Σ ( A×EFgas×GWPgas )
A activity data in base units (kg of coal, kWh, litres), after conversion from what the worker entered.
EF emission factor per gas, from a versioned factor library (for example IPCC defaults or a national grid factor).
GWP global warming potential that turns CH₄ and N₂O into CO₂e, from the IPCC assessment report the company reports against.
Worked path for one entry
Entered১২ বস্তা কয়লা
Converted600 kg coal
× factorsCO₂, CH₄, N₂O
× GWPkg CO₂e
ClassifiedScope 1 · Stationary combustion
Factors are stored with a version and source. When a factor is updated, past results stay reproducible and can be recalculated on purpose, never silently.
Industries and companies
New industries are configuration, not code
An industry is a template of activity types. Each company copies the template and adjusts it: its own sack weight, its own grid factor, extra types it needs. Garments can be added later the same way.
Brick kilns
Seasonal: most emissions fall in the six dry-season firing months. Intensity is kg CO₂e per 1,000 bricks.
- Coal
- Bricks made
- Electricity
- Diesel
- Firewood
- Transport
- + add type
Ship-breaking
Tracked per ship. Intensity is kg CO₂e per tonne of steel recovered. Refrigerant release from dismantled systems is a key fugitive source.
- LPG cylinders
- Oxygen cylinders
- Generator diesel
- Electricity
- Steel recovered
- Refrigerants
- + add type
Garments · later
Same engine, new template. Likely types: boiler fuel, grid and captive power, refrigerants, purchased fabric.
- Template to be defined
Industry template→Company→Site→Activity type→Units · factor · classification
AI features
Voice and photo do the typing
AI removes the need to read and write. It is used for understanding input and spotting problems, never for inventing data.
Speak the whole day
Bangla speech recognition plus item extraction. Understands local words and units, including the extra names admins add for each type.
Read receipts and meters
Reads quantity, date and seller from fuel receipts, and digits from electricity meters. The photo is kept as evidence.
Count from a photo
Counts gas cylinders and sorts them by type. The worker fixes the count with + and − if needed.
Spot missing types
When workers keep mentioning something with no type (rice husk, say), the admin is offered a one-click new type.
Flag unusual entries
Compares each entry with the site's history and season, and sends outliers to the admin with their evidence.
The rule: AI suggests, a person confirms. Every AI result is shown back and confirmed by the worker, every entry records how it was made (tap, voice, photo) and its confidence, and the report shows how much of the total is backed by evidence.
Designing for low literacy
If it needs reading, it needs another way too
- One question per screenA short Bangla sentence, a large picture, one main button.
- Units people already useSacks, trucks, maunds, cans, cylinders. Conversion happens behind the scenes.
- Everything can be heardA speaker button on every screen, and a spoken read-back before saving.
- Big, forgiving controlsTouch targets of 56 px and more, + and − instead of keyboards, easy undo.
- Offline firstSaves on the phone and sends when the network returns, and says so plainly.
- Help is one tap awayA call button to a named person at the company, always in the menu.
Standards and classification
Built on the GHG Protocol and ISO 14064-1
Every emission record carries its scope, source category, gas and factor version, so the same data produces either framework's report.
Scope 1 · Direct GHG emissions
From sources the company owns or controls.
- Stationary combustion (kiln coal, generators, biomass CH₄/N₂O)
- Mobile combustion (company vehicles)
- Process emissions (carbonate decomposition in clay)
- Fugitive emissions (refrigerant leaks and releases)
Scope 2 · Energy indirect emissions
From purchased electricity, heat or steam.
- Location-based (grid average factor)
- Market-based (supplier-specific), reported alongside
Scope 3 · Other indirect emissions
In the value chain, from 15 standard categories. Most relevant here:
- Cat. 3 Fuel- and energy-related activities
- Cat. 4 Upstream transportation and distribution
- Cat. 9 Downstream transportation and distribution
| Every record stores | Values | Why |
|---|---|---|
| Scope and source category | Scope 1–3 · stationary, mobile, process, fugitive · Scope 3 Cat. 1–15 | GHG Protocol Corporate and Scope 3 Standards |
| ISO 14064-1 category | 1 Direct · 2 Imported energy · 3 Transportation · 4 Products used · 5 Use of products · 6 Other | ISO 14064-1:2018 inventory and verification |
| Gas | CO₂ · CH₄ · N₂O · HFCs · PFCs · SF₆ · NF₃ | Gas-level reporting and GWP conversion |
| Biogenic flag | Biogenic CO₂ kept as a memo item outside the scopes | Required separate reporting for biomass |
| Factor version and source | Library, year, region, GWP set | Reproducible results and base-year recalculation |
| Evidence and method | Photo, voice clip, meter reading · tap / voice / photo · AI confidence | Audit trail for ISO 14064-3 verification |
Architecture
Offline phone app, one multi-company backend
Each company's data is isolated. The calculation engine and factor library are shared and versioned.
Clients
Field appAndroid · Bangla · local database and sync queue
Admin webTypes, units, factors, sites, approvals
Reports webDashboards and exports
Edge
API gatewayAuth by phone + PIN, per-company access rules
Sync serviceAccepts offline batches, resolves conflicts
Services
ConfigurationIndustry templates and company overrides
AI servicesSpeech, document reading, object counting, anomaly checks
Calculation engineA × EF × GWP, classification
Reporting and exportGHG Protocol, ISO 14064-1, CSV
Data
Company databaseEntries, emissions, audit log, per tenant
Factor libraryVersioned factors and GWP sets
Evidence storagePhotos and voice clips
Suggested rollout
Start small, prove it in the field
Phase 1Now
Pilot two industries
- Tap entry, offline sync, Bangla read-aloud
- Admin setup for bricks and ship-breaking
- Test with a handful of real workers per site
Phase 2
Add AI input
- Voice entry and receipt/meter reading
- Photo counting for cylinders
- Anomaly flags and approval flow
Phase 3
Full reporting
- GHG Protocol and ISO 14064-1 exports
- Targets, intensity and site comparison
- Verifier access to evidence
Phase 4
More industries
- Garments template
- Company-to-company benchmarks
Want to pilot the app at your site?
We're looking for brick kilns and ship-breaking yards for Phase 1.
High-level design v0.1. Factor libraries, GWP set and targets to be confirmed before production.