The structural engineer has just sized the building, and the concrete bid comes back with a number that lands twice: once on the budget, and once on the carbon ledger. A 220,000 SF higher-ed academic building targeting LEED v5 Platinum can hold roughly nine thousand cubic yards of concrete in its frame and slabs, and at a baseline mix that is a few thousand metric tons of carbon dioxide equivalent embedded in the structure before a single light fixture is energized. LEED v5, released April 28, 2025 and the sole LEED registration option as of July 1, 2026, shifts roughly half its credit weight to decarbonization, and for Platinum BD+C it sets an embodied-carbon reduction target around 20 percent against a baseline. That number is now a design constraint, not a sustainability footnote. This lesson is about using AI to query EPDs in EC3 at the assembly level, run whole-building comparisons in Tally and One Click LCA, and stage early-design alternates in Carbon Designer 3D, fast enough to steer the design while the human verifies the EPD data and owns the material decision, ending in the embodied-carbon optimization memo: a named baseline, three alternates, and the GWP reduction percentage tied to the credit threshold.

The Expensive Concrete Moment

The concrete frame is where embodied carbon concentrates, and it is also where the design decisions get locked early and cost the most to unwind. A baseline ready-mix at a high cement content carries a Global Warming Potential (GWP) that a higher supplementary-cementitious-material mix, with slag or fly ash substituting for a portion of the Portland cement, can cut substantially, but the substitution affects strength gain, cure time, and the structural schedule, so it is a real design decision with real consequences, not a free swap. The expensive moment is that this decision wants to be made at the right time: early, when the frame is still notional and the alternate is cheap to evaluate, rather than late, when the rebar is detailed and the formwork is sequenced and changing the mix ripples through the whole structural package.

Done manually, the carbon accounting that would inform this decision is slow. An estimator or sustainability consultant has to find the relevant Environmental Product Declarations (EPDs), read the GWP figures off them, map each product to the right assembly and quantity, total it across the building, and then do the same again for each alternate mix or system to compare. That is days of careful, error-prone spreadsheet work for a single comparison, and the design team needs many comparisons to find the alternates that hit the reduction target without blowing the budget or the schedule. So the carbon question has the same structure as the cost question in Target Value Design: it is a number that should steer the design, but the manual accounting is too slow to steer with, so the design gets made first and the carbon gets counted after, which is exactly backwards if the target is a 20 percent reduction.

AI as the Analytic Engine for LCA Querying

This is where AI earns its place: as the analytic engine that runs the life-cycle assessment querying and the whole-building comparison fast enough to steer with. Used well, AI queries EC3, the Embodied Carbon in Construction Calculator, at the assembly level, pulling the EPDs for the concrete mixes, the structural steel, the rebar, the envelope assemblies, and surfacing the GWP figures and their variance so the team can see which products and which assemblies carry the carbon. It can run the whole-building comparisons in Tally, the LCA plug-in that reads the building model, and in One Click LCA, the whole-building assessment platform, totaling the GWP across the structure and envelope and comparing the baseline against each alternate. And it can stage the early-design alternates in Carbon Designer 3D, which generates and compares structural and envelope schemes at the concept stage before the model is detailed.

The value is the same value the program has named across every domain: the AI compresses the analytic work so the answer arrives in time to steer the decision. The manual carbon accounting that took days now takes hours, so the design team can run many alternates, the higher-SCM concrete, the lighter steel frame, the lower-carbon envelope assembly, and see the GWP for each against the baseline and the target, which is the comparison the reduction goal requires and the manual process could not deliver in time. This is the analytic engine running many material and assembly alternates fast, the same role AI played in the takeoff and the schedule, applied to the carbon ledger, and it is truly valuable here because the speed is what lets the carbon number steer the design instead of merely scoring it after the fact.

The Carbon Number Is a Signal, Not a Decision

But the GWP figure the AI produces is a signal, not a decision, and this is the load-bearing distinction. The embodied-carbon numbers steer the design the same way cost steers the design in Target Value Design: they are metrics that point toward the lower-carbon alternate, telling the team which mix or system to favor, but they do not make the material decision, because the material decision belongs to the engineer and the design team who must weigh the carbon against the strength, the cost, the schedule, the durability, and the constructability. A mix with the lowest GWP that does not gain strength fast enough for the pour schedule is not the right answer just because it scores best on carbon, so the carbon number informs the decision without making it.

This is the metric-as-signal dynamic the program established with cost and schedule, applied to carbon: the number is a steering signal that points the design toward the target, and the human reads the signal and makes the design decision, weighing the carbon against everything else the design must satisfy. The AI accelerates the production of the signal, running the alternates fast so the signal is available in time to steer, but the steering, the actual choice of mix and frame and envelope, is the human's, because the choice is a design decision with consequences the carbon number alone does not capture. The carbon GWP is a signal that steers, the human reads it and owns the material decision, and the AI's value is producing the signal fast enough to steer with, not making the decision the signal informs.

The embodied-carbon number is a signal that steers the design, the same steering dynamic as cost in Target Value Design, so the AI accelerates the LCA querying to produce the signal fast, but the human verifies the EPD data and owns the material decision, because a fast wrong GWP missteers the design exactly the way a fast wrong cost does.

The Danger: Steering on a Mis-Mapped EPD

The danger in this workflow is precise and it follows directly from the metric-as-signal dynamic: steering the design on an unverified or mis-mapped EPD. If the AI pulls the wrong EPD for a product, or maps an EPD to the wrong assembly or the wrong quantity, or compares a product-specific EPD against an industry-average baseline so the two are not measured on the same basis, then the GWP figure it produces is wrong, and a wrong GWP steers the design wrong, favoring an alternate that is not actually lower in carbon or rejecting one that is. This is the exact analog of the fast wrong cost in Target Value Design: a fast wrong number missteers the decision, and the speed makes the error worse because it arrives with the confidence of a finished comparison and the design team acts on it.

The mis-mapping failures are specific and they are where the human verification concentrates. An EPD has a declared unit, a scope (cradle-to-gate versus cradle-to-grave), and a reference service life, and comparing two products requires that these match, so an EPD compared on a different unit or scope produces a meaningless difference that looks like a real one. A product EPD is specific to a manufacturer and plant, while an industry-average EPD covers a category, so substituting one for the other changes the number without changing the product. And the assembly mapping, which EPD applies to which quantity of which assembly, is where a fast tool can attach the right declaration to the wrong line, totaling correctly off a wrong premise. So the verification is not a rubber stamp on the AI's GWP total but a check that each EPD is the right declaration, mapped to the right assembly and quantity, compared on the same basis, because the carbon decision rests on that mapping being right.

The Human Verifies the Data and Owns the Decision

So the division of labor is the program's standard division applied to carbon: the AI accelerates the LCA querying and the whole-building comparison, producing the GWP signals fast across many alternates, and the human verifies the EPD data and owns the material and design decision. The verification is the EPD check: confirming each EPD is the correct declaration for the actual product, that its declared unit and scope and service life are right for the comparison, that the assembly mapping attaches it to the right quantity, and that the baseline and the alternates are measured on the same basis so the reduction percentage is real. This is the data-verification gate applied to the carbon ledger, the same discipline as verifying a takeoff quantity or a cost figure before it drives a decision.

And the decision is the human's: the engineer and the design team weigh the verified carbon signal against the strength, cost, schedule, durability, and constructability, and choose the mix, the frame, and the envelope, owning that choice as a design decision they are responsible for. The AI did not choose the higher-SCM mix; it produced the signal that showed the mix's lower GWP, and the team chose it after weighing the schedule effect of the slower strength gain, so the decision is theirs and the carbon number is one input they verified and weighed. This is the responsible-charge discipline applied to embodied carbon: the design professional owns the material decision, the AI accelerates the analysis that informs it, and the verification of the EPD data is what makes the accelerated analysis trustworthy enough to steer with. The human verifies the data so the signal is real, and owns the decision the signal informs, which is the only arrangement in which the AI's speed helps rather than misleads.

Staging the Workflow Across the Tools

The four tools stage across the design timeline, which is how the workflow is sequenced. Carbon Designer 3D comes first, at the concept stage, generating and comparing structural and envelope schemes before the model is detailed, so the early big moves, the frame type and the envelope strategy, are steered by carbon when they are cheapest to change. As the design develops and the model takes shape, Tally reads the building model to run the whole-building LCA inside the modeling environment, and One Click LCA runs the whole-building assessment for the comparison and the LEED documentation, totaling the GWP across the structure and envelope. Throughout, EC3 is the EPD source queried at the assembly level, supplying the product-specific declarations that make the comparison real rather than generic.

This staging matters because the carbon decision wants to be made early, when the alternate is cheap, and refined late, when the model is detailed, so the tools that work at the concept stage steer the big moves and the tools that work at the model stage refine and document. The AI accelerates the querying and comparison at each stage, but the verification regime is constant: at every stage, the EPD data that drives the GWP figure must be verified, because a mis-mapped EPD at the concept stage missteers the big move and a mis-mapped EPD at the model stage produces a wrong reduction percentage in the LEED documentation. So the workflow runs the AI-accelerated analysis at the right stage with the right tool, and the human verifies the EPD data at each stage before the GWP signal steers the decision, which is the staging that puts the carbon signal in front of the design decision at the moment the decision is cheapest to make right.

The Applied Problem: The Embodied-Carbon Optimization Memo

Here is the exercise. Produce the embodied-carbon optimization memo for the 220,000 SF higher-ed academic building targeting LEED v5 Platinum BD+C. Establish a named EC3 baseline: the structure and envelope assemblies with their product-specific EPDs and the total baseline GWP. Then develop three structural and envelope alternates, for example a higher-SCM concrete mix, a lighter or hybrid steel frame, and a lower-carbon envelope assembly, run them through Tally or One Click LCA as whole-building comparisons against the baseline, and stage the early big moves in Carbon Designer 3D, with the AI accelerating the querying and the comparison so all three alternates can be evaluated fast.

For each alternate, the memo states the GWP and the reduction percentage against the baseline, and it ties that percentage to the credit threshold, the roughly 20 percent embodied-carbon reduction LEED v5 sets for Platinum BD+C, showing which alternates clear the target and by how much. The verification is the EPD check on every figure: each EPD confirmed as the right declaration for the actual product, the declared unit and scope and service life confirmed consistent across the baseline and the alternates, the assembly mapping confirmed correct, so the reduction percentages are real and not artifacts of a mis-mapped or mis-compared declaration. The memo records that the carbon figures are signals the design team weighed against strength, cost, schedule, and constructability, and that the material decision is the team's, owned as a design decision, with the carbon number one verified input among several.

The deliverable is the embodied-carbon optimization memo: the named EC3 baseline, the three structural and envelope alternates, and the GWP reduction percentage for each tied to the Platinum threshold, with the EPD verification recorded so the numbers can be trusted to steer. The lasting product is a workflow that uses AI to run the LCA querying and the whole-building comparison fast enough to steer the design toward the LEED v5 reduction target, while the human verifies the EPD data so the carbon signal is real and owns the material decision the signal informs, which is the only way the carbon number can steer the design without missteering it on a fast wrong figure, because the embodied-carbon decision, like the cost decision in Target Value Design, is a design decision the professional owns and the AI's accelerated analysis informs but does not make.

Key Takeaways

• LEED v5, released April 28, 2025 and the sole LEED registration option as of July 1, 2026, shifts roughly half its credit weight to decarbonization, including an embodied-carbon reduction target around 20 percent for Platinum BD+C, which makes embodied carbon a design constraint rather than a sustainability footnote.
• Embodied carbon concentrates in the concrete frame and the early big moves, where decisions lock cheaply if made early and expensively if made late, so the carbon decision wants the analysis in time to steer the design rather than score it after.
• AI is the analytic engine that runs the LCA querying and whole-building comparison fast: querying EC3 EPDs at the assembly level, running Tally and One Click LCA whole-building comparisons, and staging early-design alternates in Carbon Designer 3D, compressing days of carbon accounting into hours so many alternates can be evaluated.
• The GWP figure is a signal, not a decision: the embodied-carbon number steers the design the same way cost steers it in Target Value Design, pointing toward the lower-carbon alternate, while the engineer and design team own the material decision, weighing carbon against strength, cost, schedule, durability, and constructability.
• The danger is steering the design on an unverified or mis-mapped EPD: a wrong EPD, a wrong assembly mapping, or a comparison on inconsistent unit, scope, or service life produces a wrong GWP that missteers the design exactly the way a fast wrong cost missteers a Target Value Design decision.
• The human verification concentrates on the EPD data: confirming each EPD is the right declaration for the actual product, that its declared unit, scope, and reference service life are consistent across the baseline and the alternates, and that the assembly mapping attaches the right declaration to the right quantity, so the reduction percentage is real.
• The four tools stage across the timeline: Carbon Designer 3D steers the early big moves at the concept stage, Tally and One Click LCA run and document the whole-building comparison as the model develops, and EC3 supplies the product-specific EPDs throughout, with the EPD verification constant at every stage.
• The deliverable is the embodied-carbon optimization memo: a named EC3 baseline, three structural and envelope alternates, and the GWP reduction percentage for each tied to the roughly 20 percent Platinum BD+C threshold, with the EPD verification recorded so the carbon signal can steer the verified, owned material decision.