Apple made some waves when it said that the newest Apple Watch would be carbon-neutral, which is not an insignificant feat, given that the company sold over 50 million smartwatches last year.
But at the same time, Apple brought in almost $400 billion in revenue last year. It has plenty of money to study the impact its products have on the environment in great detail. And perhaps more importantly, it has the leverage required to demand the sort of changes that make a carbon-neutral hardware product possible.
So it’s reasonable to assume that smaller companies have a much harder time estimating the carbon footprint of their products. For most tangible products, everything from materials to manufacturing to distribution leaves a carbon footprint, and not all vendors report their emissions.
“The majority of a company’s emissions are within their supply chain — more than 70% for the average corporation,” Harris Chalat, co-founder and CEO of Muir AI, told TechCrunch+. “It’s the largest opportunity for reductions. But because the global supply chain is incredibly complex, these corporations don’t understand how to even go about beginning to reduce those emissions.”
Getting a handle on supply chain emissions — also known as Scope 3 emissions — is by far the most complex part of this problem. Companies don’t exert any direct control over such emissions. They can ask nicely or they can have their contracts require vendors to supply auditable emissions data, but most companies don’t have the sort of leverage over suppliers that Apple does to accomplish that.
What’s more, an expert can take several months to perform a life cycle analysis on a single product. If there ever was a business problem crying out for a software solution, it’s this.
Muir AI hopes to be that solution. The startup works with companies to understand their supply chains and glean what data it can from them. If that data is unavailable, Muir AI uses the information in their databases combined with clues from the rest of the customer’s supply chain to make “probabilistic” assumptions about the carbon impact of the component, said Peter Williams, the company’s CTO.
Artificial intelligence comes in at a few points, Williams said. One is in the “product-breakdown stage,” in which the company virtually disassembles the product into components and materials to create a map of the supply chain. It also uses generative AI to fill in some of the gaps.
Much of Muir AI’s data comes from open repositories on the web, peer-reviewed journal articles, and from customers themselves. But where data is missing or where they want to validate something, they’ll go ahead and study satellite imagery of the facilities that produce a component.
“When we do know the actual physical location that a good is being manufactured or produced at, such as the actual factory, we use satellite imagery to extract relevant details about that facility that we then feed into our estimation model,” Williams. said.
For example, the company might look for on-site renewable power or a factory’s smokestack, which can give clues about its fuel source. They also look for clues about factory utilization rates. Williams didn’t go into detail about this, but clues might include things like the number of trucks at loading docks or cars in an employee parking lot.
“That allows us to get a more nuanced, more accurate estimate of the emissions that are associated with that step,” he said.
No estimate is perfect, of course, so Muir AI intends to give customers details on the uncertainty around its figures. It also plans to help customers trace where the biggest sources of error are in their supply chains.
Muir AI has raised a $3.25 million seed round led by Base10 Partners, the company exclusively told TechCrunch+. Existing investors Madrona Venture Labs and Soma Capital joined. The round will fund the technical and business development parts of the company.
For sustainability executives — or more realistically, for their teams — such software promises to be a boon for productivity. Software like Muir’s should deliver on AI’s promise to augment existing jobs rather than replace them, though that’s partially because “life cycle analyst” isn’t a booming job title yet.
Like with any seed-stage company, it’s still too early to say whether Muir will deliver. But a lot of the company’s success will depend on how trustworthy its data is. Data that comes from probability-based models should be on firmer ground. After all, humanity has decades of experience with such models and generally understands how they work and their limitations.
AI is trickier. In some cases, we have good confidence in its results. In others, we don’t know why a model spat something out. Conveying that uncertainty and doing so transparently should help Muir AI build customer trust.