Lodestellar Uses AI to Boost EPD Quality in Construction

The era of vague environmental claims in construction is fading, as manufacturers increasingly turn to data-driven tools to prove their sustainability credentials. Lodestellar, an Estonian startup, has developed an AI-powered platform that acts as a quality assurance system for Environmental Product Declarations (EPDs), helping companies reduce delays, cut verification costs, and avoid compliance risks.

“Manufacturers are moving away from meaningless eco slogans and self-declared green certificates,” says Lodestellar CEO Anni Oviir. “Sustainability is leaving the marketing department and entering data science, where it belongs.”

Oviir leads a team of Life Cycle Assessment (LCA) experts who created Lodestellar to address a growing bottleneck in the construction industry. With around 50 EPDs published daily worldwide, independent verifiers are overwhelmed by complex declarations that often contain errors. A single EPD can cost manufacturers over €10,000 and take more than six months to complete, with multiple revision rounds during verification , and no guarantee of success if standards aren’t met.

The platform works like a “linter” for sustainability data, automatically scanning draft EPDs against all relevant standards, such as EN 15804+A2 or ISO 21930. It returns a detailed, line-by-line report highlighting missing explanations, ambiguous assumptions, inconsistent wording, omitted disclosures, or poor traceability between data and results.

“When we catch quality issues early, even before submission for verification, we eliminate significant cost, hassle, and delay for manufacturers,” Oviir explains.

The tool is already helping major building product companies raise their EPD quality and accelerate publication. Its development was sparked by a 2022 Estonian Economic Ministry hackathon that brought software developers and construction experts together. The team, now led by CTO Tanel Teinemaa, modeled Lodestellar on the automated quality checks programmers use on their code.

“Even the best developers run linters before publishing,” Teinemaa says. “EPD creators are developers of a different kind, with equally complex work where small details matter enormously.”

Industry experts see strong potential. Dr. Roger Singleton, CEO of Riskoa, notes that independent review will become critical as AI accelerates EPD production. “Speed only creates value if results can be trusted. Lodestellar adds an independent quality layer, so providers aren’t marking their own homework.”

Professor Callum Hill, an independent LCA verifier, calls it “an extra pair of eyes” that catches issues easily missed in complex models. “As EPDs are increasingly used in tenders, mistakes can become disputes. This tool helps fix basic issues early, reducing unnecessary back-and-forth.”

The stakes are high. Errors caught during verification add cost and time, but errors that slip through could create legal risks as EPDs become more valuable in procurement decisions. Lodestellar’s team has examined years of verifier feedback and found that quality issues follow predictable patterns.

“We’ve checked hundreds of EPDs and haven’t found a single one , even among published declarations , without issues that automated checks could have spotted,” Oviir admits. “That includes EPDs I created myself. The standards are incredibly complex, so we all benefit from an extra quality layer early on.”

The team is now developing internal benchmarking methods to track performance transparently and rolling out the ability to analyze EPD background reports. They aim to keep the tool low-cost and intuitive, ensuring accessibility for widespread adoption.

“We want to raise EPD quality even beyond compliance,” Teinemaa says. “That required codifying every possible requirement in the standards and developing our own extensive checklist. The technology has only caught up with our ambitions in the past year.”

While built on AI, Oviir emphasizes that results must speak for themselves. “The companies extracting real value from AI never use it as a buzzword. In both AI and sustainability, ‘eco-friendly’ and ‘AI-powered’ are equally meaningless without proof.”

She argues that AI’s broader impact will come through empowering buyers. “Specifiers must weigh carbon cost alongside fire-proofing, acoustics, and financials. As AI advances, they can easily compare all these factors across products. Everything is moving toward credible, verified data as the advantage. Ditch the unsubstantiated slogans and make sure your data is properly calculated and declared.”