How Unilever is using AI to make its products more sustainable
On a daily basis, three billion consumers use at least one of Unilever’s products – whether it’s spreading some Marmite on toast, filling the washing machine with Comfort fabric softener, or slathering on Dove moisturiser.
The mass production and use of Unilever’s fast moving consumer goods make its goal of reducing 70% of scope one and two emissions by 2025 – before completely hitting net zero in 2039 – quite a daunting task.
Just about half of those emissions come from the ingredients that Unilever uses in its foods, soaps, and other products, 17% of the emissions come from distribution, manufacturing and logistics, 12% come from the packaging materials, and 10% come just from the running of their ice cream cabinets.
“With that kind of size comes responsibility, and you’ve got this target, which is very ambitious” acknowledges Alberto Prado, Unilever’s R&D’s head of digital and partnerships.
“When I see that, suddenly the only option that we have to fill our commitments and therefore transition at speed, is to not only work in collaboration with partners, but leverage technology.”
Prado, who has been in his role at Unilever for two and a half years after working as vice president of digital innovation at healthcare technology firm Philips, states that his mission at Unilever is to help the entire group of R&D transition from traditional research and development to using technology such as computer models and analytics in their work.
“Why? Because it’s much more effective at driving new scientific discovery more efficiently,” he says. “If you look at the contribution of R&D through the use of digital technologies as change agents, what we can do to decrease emissions is massive.”
Keeping the ice cream from melting
Pumping out 10% of Unilever’s commissions, the three million ice cream cabinets Unilever owns are proving to significantly burn into its net zero goals.
To address this, Unilever has set out a roadmap to reduce its emissions which includes reducing cabinet energy consumption such as powering them with renewable electricity, innovating the technical components to be more energy efficient, and also, ‘warming up’ the freezers.
Raising the temperature of its last mile freezers from -18°C, the industry standard, to -12°C reduces power consumption, but it can only be done if the ice creams stay frozen.
What this means for Prado, is using machine learning technology to help recreate the formulation of Unilever’s Cornetto ice creams so that they can withstand the higher temperature without dramatically changing the taste.
“If you did it in a lab, it would take another 100 years, but if you do it with AI and machine learning models, and leveraging emerging techniques like genetic AI, then we can really speed up what we call the material transition,” says Prado.
The material transition essentially outlines taking carbon-emitting manufacturing and production processes, as well as unrecyclable materials used in packaging and more, and transitioning these to more carbon-friendly alternatives.
“We don’t think that reaching net zero is just a food supply chain manufacturing thing, it’s an R&D thing as much as it is a marketing thing,” says Prado. “It’s a bit of a team sport.”
In the research and development department, this means telling the manufacturers how they should manufacture the next product.
“So we need to come up with not only the ingredients that make up the specification for that product, we also use AI to optimize waste reduction, and use the lowest possible energy levels required.”
Reformulating ingredients doesn’t always mean for the food, but also for Unilever’s homecare products such as cleaning and laundry supplies.
“The good thing about these products is that they are very effective, the bad thing is that they traditionally, and it’s not just within Unilever but everyone in the industry, have fossil fuel-derived ingredients in them.”
For example, typically laundry detergents use an ingredient called surfactants which are used for creating the foam in most laundry and household cleaning products – typically derived from fossil fuels.
To find an alternative, Unilever partnered with carbon recycling company LanzaTech and India Glycols to manufacturer greener, technology-based chemicals to replace the surfactants in their China-based laundry detergent OMO (also known as Persil in the UK).
The greener surfactant replacement used in the product now is made by capturing the industrial emissions at a steel mill in Beijing that is then converted into ethanol. Then, the ethanol is converted to ethylene oxide, which is then used to make a range of ingredients including surfactants that goes into the laundry capsules at OMO’s factory in China.
This cuts the production of surfactants emissions by 82%.
Using biotech to create replacements such as this “is not easy,” says Prado, so using machine learning and artificial intelligence will make these discoveries much faster than in the lab.
“It’s not just about making your energy sustainable through solar and alternative energy sources, it is also about making materials sustainable.”
Prado hails the use of partnering with specialist start-ups, firms, and academia in order to help recreate each of its products to be more environmentally friendly: “It’s about building that ecosystem of companies that can all bring a unique skill the can solve, and that’s the way we secure access to talent that is out there.”
Plus, while using AI and machine learning has dramatically helped cut emissions, it is also the power consumption of these technologies that is important to balance.
“We work a lot with Microsoft who have their own targets,” he says, “and we make sure that they’re aligned with ours as well.”
Subscribe to our Editor's weekly newsletter