Most architects can tell you the embodied carbon of their last project to one decimal place. Far fewer can tell you which line in the specification put it there. That gap is the whole problem. Embodied carbon gets discussed at concept, modelled in spreadsheets at planning, and then quietly decided, clause by clause, in a document much of the team still treats as paperwork. The specification is where the carbon target either survives contact with reality or it doesn't.
The early RIBA stages get the attention. That's where the carbon strategy is presented, the massing is justified, and the client nods along to a figure on a slide. None of it is binding yet.
The decisions that fix the number come later, usually around Stage 4, when someone writes down the concrete mix, the steel grade, the insulation type, the cladding build-up, and the timber source. A C32/40 mix with thirty per cent cement replacement is a different carbon outcome to the same strength with none. Recycled-content steel is a different outcome to virgin. Those are specification choices, not concept choices, and they live in clauses that rarely get the same scrutiny as a planning render.
The RIBA 2030 Climate Challenge asks for embodied carbon of 300 kgCO2e/m² for homes and 500 kgCO2e/m² for non-domestic buildings, measured over a 60-year life across modules A to C. Set that against typical baselines of around 1,000 kgCO2e/m² for housing and 1,100 for commercial work, and you are looking at a cut of fifty to seventy per cent. That is not a tweak. It is a different way of choosing materials.
The regulatory direction is just as clear. Part Z, the industry-drafted amendment to the Building Regulations, would require whole-life carbon reporting on projects over 1,000m² or ten homes, with actual limits expected to follow around 2027 to 2028. It isn't law yet, but the Greater London Authority already demands whole life-cycle carbon assessments for referable schemes, and the UK Green Building Council's Net Zero Carbon Buildings Standard, published in 2024, treats embodied and operational carbon as one assessment. The reporting habit is arriving before the legal limits do, which means firms that can produce credible carbon-aware specifications now will be ready when the limits land.
There is a second number that changes how you think about all of this. For a reasonably efficient new building, upfront embodied carbon, the modules covering raw materials, transport, and construction, can account for roughly half of the whole-life total. Operational carbon you can keep improving over decades, as the grid decarbonises and plant gets replaced. Upfront embodied carbon is spent the day the building goes up. You get one chance to get it right, and that chance is the specification you issue before anything is ordered.
Concept-stage carbon estimates almost always use generic data. You pick a material category, the tool applies an industry-average figure, and the project looks reasonable on paper. The estimate isn't dishonest. It just isn't binding.
Then procurement happens. The generic low-carbon block becomes whatever the merchant can supply on the programme. The specified mix gets value-engineered. A substitution request comes in, gets approved on cost and lead time, and nobody checks what it did to the carbon figure. By completion the building can be a long way from the slide that won the job, and the only place that drift could have been caught was the specification.
If you want a carbon target to hold, you have to write it into clauses that a contractor and a quantity surveyor are obliged to follow. That means requiring a product-specific Environmental Product Declaration for every major element rather than accepting category averages. It means setting a maximum A1 to A3 figure for key materials, mandating cement replacement percentages, and writing substitution clauses that force any proposed swap to prove carbon parity, not just cost parity.
This is detailed, repetitive work, and it is exactly the kind of work that gets cut when a deadline tightens. Tools like Avoice are built for this part of the job. By generating specifications structured around recognised standards such as Uniclass and CAWS, and grounding clauses in a firm's own product choices rather than a generic clause library, the carbon requirements stop being an afterthought bolted onto a boilerplate document and become part of how the spec is assembled in the first place.
Work in the right order. The cheapest tonne of carbon is the one you never specify, so the first question on any element is whether it needs to be there at all, then whether it can be lighter or thinner, and only then which product to name. Exposed soffits that remove the need for suspended ceilings, structural grids that avoid transfer beams, and retained existing fabric all cut carbon before a single low-carbon product is chosen. The specification should reflect that hierarchy rather than jumping straight to swapping one material for a greener version of the same quantity.
Start by writing performance, not just brand. A clause that names a single product locks you to that product's data and that product's supply chain. A clause that sets a carbon limit and a performance requirement lets the team find the best available option without losing the target. Ask for EPDs as a submittal, not a nice-to-have.
Resist over-specifying. A lot of embodied carbon is simply material you didn't need: concrete thicker than the structure requires, strength grades higher than the loads justify, finishes doubled up out of habit. Specify what the building needs and no more. And keep the specification honest against the schedules. If the window schedule says one thing and the spec says another, the carbon assessment is built on sand. Avoice flags inconsistencies between specifications and the drawings and schedules before they reach site, which matters because a coordination error doesn't just cost rework, it quietly invalidates the carbon numbers you reported.
Take a mid-rise residential block. On most schemes like it, the structure and substructure carry the largest share of upfront carbon, often around half, with the facade and finishes making up much of the rest. That tells you where the specification effort should go. Fussing over the carbon of an internal paint while leaving the frame and the concrete unexamined is effort spent in the wrong place.
So the structural specification is where you push hardest. Cement replacement with GGBS or fly ash, a realistic strength grade rather than a cautious one, and a clause requiring product-specific declarations for the mixes actually supplied. On the facade, the decision between a unitised aluminium system and a lighter, lower-carbon build-up can move the number meaningfully, and that decision has to be written down with enough precision that it can't quietly revert during procurement. None of this is exotic. It is ordinary specification work, done with carbon as one of the things the clauses are protecting, alongside performance, durability, and cost.
The point is that carbon reduction is not a separate document or a consultant's report that sits beside the spec. It is the spec, written slightly differently. Once you see it that way, the question stops being whether to do carbon specification and becomes how to do it without doubling the hours your team already doesn't have.
The honest difficulty with carbon specification is not ambition. It is data. EPDs sit in PDFs nobody has time to read. The genuinely useful information, what your practice specified on the last three similar buildings and how it performed, is scattered across old project folders, superseded drawings, and the heads of whoever happened to run those jobs.
That is the gap worth closing. Avoice ingests a firm's existing documentation, the sheets, specifications, schedules, material libraries, and historical projects, and turns them into structured, searchable data you can actually reuse. Instead of starting each carbon-aware spec from a blank clause library, you start from what your own practice already knows works, classified the way your projects are already classified. The carbon story becomes consistent across jobs because it is drawn from the same evidence base, not reinvented each time by whoever is free that week.
Carbon is following the same path as every other technical requirement architects have absorbed. Thermal performance used to be a specialist concern, then Part L made it everyone's. Accessibility, fire, airtightness, all started as add-ons and became fluency. Embodied carbon is next, and the architects who can specify it as confidently as they specify a fire rating or a U-value will be the ones clients trust with the projects that have real carbon scrutiny attached.
The numbers are already public, the assessment methods are settled, and the limits are coming. The only open question is whether your specifications are ready to carry the weight. If you want to see how carbon-aware specification works on a real project rather than in theory, Avoice runs demos tailored to how your practice actually works.
