Maria Tymtsias

Maria TymtsiasSep 9, 2026

I read only one newsletter each Sunday and it's about bricks, cement, sand and reuse. It's written by Simon Maillet, sustainability manager at Cordeel Group, and it has taught me more about his job than any conference. Why two identical bricks differ by a factor of ten in carbon? How to read an EPD without a PhD? So together with Simon we turned the newsletter, his calls and his readers' questions into nine AI skills for construction sustainability. Here is how I would use each one if I was a Sustainability Manager in Construction. 1/ EPD reader and comparer: Two EPDs arrive on a Friday and before I look at the carbon number, the skill shows me the unit, the service life and the modules included. If one is per cubic meter and the other per square meter, I ask the supplier for the same unit first. 2/ Tender scanner: I upload the whole pack and get every sustainability requirement with its page: ladder tier, MKI or TOTEM threshold, reuse share, EPDs as proof. I sort it into "have it", "can get it", "don't", and take that to the go or no-go meeting. 3/ Supplier product screener: Purchasing forwards "new sustainable insulation, EPD attached". The skill tells me if the EPD is verified, how it compares with what we buy today, and gives me two questions to send back. 4/ Low-carbon material finder: Strength class, exposure class, country in. What the concrete standard already allows out, plus the sentence for the order: "quote the maximum recycled content you can certify for this class." 5/ Product approval reviewer. Every product passes someone's approval: the client platform, the BREEAM or GRO assessor, the tender's material list. The skill says whether the product has a sustainability requirement at all, finds the one page with the VOC value or recycled share, and writes one line: compliant, not compliant, missing. I sign off a line, not a 400-page file. One team told me this costs them 300 hours per project. 6/ CO2 Performance Ladder 4.0 gap check. Last year's 3.1 documents in, a table out: requirement, what we have, what's missing, who owns it. 7/ Group data consolidation. Thirty companies, thirty Excel layouts. Columns mapped, rolled up to site level, doubles flagged. The auditor asks where a number comes from, I click and see the file and the row. 8/ Contributor guide. The plant manager has never heard of ESRS. He gets a link, answers plain questions about waste, water and energy, and gets a draft he only has to correct. 9/ Site biodiversity screener. Our sites and supplier locations in. Distance to Natura 2000, water stress and a ranking out. The spreadsheet I'd started twice and never finished. _____________________ 📆 On 30 September Simon joins Palau workshop in Antwerp. You bring your own public data and run these yourself. Sign up here: https://lnkd.in/dwtqBJwV ➡️ Want to try one skill tonight? Simon's EPD prompt is free. Get it here: https://lnkd.in/dfm5mcQU

Maria TymtsiasSep 10, 2026

24 września w Warszawie będziecie budować własnych agentów AI dla zrównoważonego rozwoju na publicznych danych waszych firm. Z tego co wiem, nikt jeszcze nie robił takiego formatu dla pracy w zrównoważonym rozwoju w Polsce, więc robimy to pierwsi. Warsztat organizujemy razem z Kozminski University i profesorem Boleslaw Rok. Jak wyglądają te 4 godziny (10:00–14:00): Zaczynamy od krótkiej rundy dyskusyjnej: kto i jak już używa AI w swoich projektach i gdzie to się sypie. Potem każdy dostaje dostęp do workspace'u i buduje agenta razem ze mną na własnych publicznych danych (wasze raporty, polityki, opublikowane cele). Co realnie da się zbudować w kilka godzin: 1/ Agenta, który sam wypełnia raport zrównoważonego rozwoju i konsoliduje dane z różnych źródeł. 2/ Agenta, który przygotowuje odpowiedzi do EcoVadis z dokumentów, które już macie, podpowiada, jak poprawić wynik na podstawie poprzednich odpowiedzi, i szacuje, jaki score dostaniecie. 3/ Agenta, który liczy emisje automatycznie, rozpisuje całą metodologię i wskaźniki emisji użyte do wyniku, i porównuje was z konkurencją. Do tego inne agenty: ROI zrównoważonego rozwoju, gap assessment pod inne regulacje, ocena ryzyk klimatycznych. Wszystko zależy od tego, jakie jest wasze główne wyzwanie. Będziemy je rozwiązywać na żywo. Wychodzicie z działającym agentem i z jaśniejszym obrazem tego, co AI może zrobić dla waszego zespołu. Do tego dobra kawa i przekąski, bo budowanie agentów na pusty żołądek nie idzie. Mamy tylko 20 miejsc. Zapisy: https://lnkd.in/d5RSy6Zx

Maria TymtsiasAug 31, 2026

Here is one of the clearest example I have of where AI could actually help with decarbonisation, and it happens nowhere near reporting. Have you ever wondered how decarbonisation looks inside a manufacturing company from the hands on perspective? If you thought there's some fancy process behind it, let me tell you what the environmental team is really doing. They are preparing 24 versions of one grant application. The grant is what pays for decarbonising one stage of one production line, and it takes around 100 hours of work to get a single application out. They have about 20 of them running at any time, and it's one person doing all of it. That's the part people skip, right? What gets recognised is a nice percentage of emissions reduction in the annual report, while how it happened and how much manual work sits underneath it stays completely invisible. The work itself goes like this: 1/ Finance finds a decarbonisation fund and sends over the environmental requirements. 2/ You check whether any of your existing projects could fit them. 3/ If one does, you go to the technical team for the specs, the quotations and the purchase orders. 4/ You put the whole thing into Excel and calculate the emissions with a methodology you wrote yourself, because there isn't a standard one. 5/ In 100% of cases the first version doesn't meet what the grant is asking for, so you start moving the levers you have: more electricity and less gas, more recycled input, biomethane, hydrogen in ten years. 6/ And every version has to hold up against reality, because the model will happily prove something the site can't deliver. The new equipment can technically run on a much higher share of electricity, but if the grid supplier can't bring that much power to the site, the version dies and you go back to the file. Which is how one application turns into 24 rounds of calculations, emails and meetings with purchasing and the technical team. The easy version of this with AI looks like giving it the fund requirements, the technical specs, the supplier quotes and the evidence you can actually produce, then asking for every combination that clears the threshold. What comes back is the handful that work, and for the ones that don't, the specific constraint that killed each of them, whether that's power availability, a gas price assumption or a document you can't get your hands on. Then you sit down for one meeting about the two or three that survived instead of twenty four. And if an application takes a day instead of 100 hours, that same person doesn't stop at 20 grants a year, which means more funded projects and more of the production line actually getting decarbonised.