Disrupting with generative AI:
How 3 innovators are using artificial intelligence to elevate business operations and protect the planet.
From healthcare to education, from renewable energy to environmental conservation, generative AI is emerging as a powerful tool to address complex challenges and improve lives. Generative AI helps streamline processes, introducing new forms of automation, new ways to decrease waste and new approaches to simplify reporting.
When applied to some of the biggest challenges facing our planet around climate change, this technology can make a big impact. Consider supply chain or fulfillment – where saving one package return or the spoilage of one container of fish can add up to big savings in emissions.
Here, we profile three companies that have worked with AWS to create generative AI solutions that are benefiting customers, workers and the planet.
Parsyl unlocks supply-chain resilience
The supply chain for perishable goods and medicines stretches worldwide. A fish caught in Japan could be on a plate in New York a couple of days later, while a vaccine made in Germany may need to be at a hospital in the Midwest immediately. Ensuring that sensitive goods like these reach their destination on time while being maintained in the proper conditions is key to preventing spoilage, maintaining safety and reducing waste. But it’s no easy task.
Looking to provide customers handling temperature-sensitive goods with better coverage, Denver-based insurer Parsyl is innovating the cargo insurance industry with a generative AI-powered solution that proactively incentivizes good risk management practices. This helps to ensure supply chain integrity and efficiency. In turn, this helps the environment by reducing product waste and lowering shippers’ fuel and electricity consumption and emissions.
AWS consultants worked with the company to design a scalable and secure architecture. To accelerate their development process, they utilized Amazon Bedrock, which gave Parsyl access to high-performing foundation models and lets developers test performance for different use cases. “That was really helpful to the team, to be able to prototype really quickly, test things out, move stuff into production really quickly,” says Mike Linton, Parsyl co-founder and chief technology officer.
“There’s a lot of information out there, so we are using generative AI to structure and understand much more of that data.”
Amazon Bedrock enabled Parsyl to accelerate time to market, moving from proof of concept to production in weeks. “They really want to innovate and move fast,” says Rahul Sareen, director for Industry and AI Solutions at AWS.
With the solution in place, Parsyl can now guide customers toward more resilient and sustainable supply-chain outcomes. The insurer incorporates emerging supply chain and transition risks into its data- and expertise-driven underwriting practices. These include data from strategic partnerships, as well as real-time and historical data sources, such as monitoring devices within shipments and cold storage facilities, public data and more.
This data provides information on shipping lanes, modes of transit, the goods to be carried and more so Parsyl can assess the likelihood of disruptions. For example, a shipment arriving at certain ports may have increased risk of delays or seasonal variations of temperature damage.
Parsyl’s customers – who typically ship perishable goods – get customized insurance policies, risk insights over time, and incentives to improve their supply chains and avoid disruptions before they happen. “The amount of data that has historically been used to underwrite those risks is pretty limited,” says Linton. “But there’s a lot of information out there, so we are using generative AI to structure and understand much more of that data.”
With risks quantified, Parsyl can offer personalized, appropriately priced insurance policies. It can also recommend ways to reduce risk and offer lower rates and improved coverage to customers that follow that advice. “When we organize and structure all that data, we can start seeing new types of trends or insights about where there might be issues,” says Linton.
As more powerful models become available, Parsyl will continue to innovate and stay true to its purpose to create value for the world by understanding and reducing risk. “We try to predict what’s going to happen, and when something does happen we want to know as much as we can about it,” says Linton.
Listen to the AWS Executive Insights interview with Parsyl co-founder and CTO, Mike Linton, to learn more about how they are transforming the cargo insurance industry and creating more supply chain resilience for their customers using real-time data and AI.
FlexZero makes ESG easier
Companies around the world are under pressure from regulators, investors and customers to be more transparent about activities that contribute to climate change. But tracking environmental data, such as electricity and fuel usage and greenhouse gas emissions, can be challenging and complex.
Enter FlexZero. The startup has built a suite of generative AI-based tools that simplify emissions tracking and reporting so more companies can meet their environmental obligations, which, in turn, helps the planet.
“We used Bedrock to learn on the side and iterate and ultimately improve the platform.”
FlexZero’s system makes Environmental, Social and Governance reporting faster, more efficient and more accurate. And it gives organizations real-time insights into their emissions so that they can make changes without delay. “You don’t have to wait until you’ve done your annual reports,” says David Johnsen, FlexZero co-founder and chief product officer.
AWS Professional Services experts helped the startup create a proof-of-concept model and bring it to production within 10 weeks. “We actually moved it to production while they were already using it for a few customers,” says Sareen.
FlexZero used Amazon Bedrock to streamline the development process, maintain flexibility and experiment with various models. “We used Bedrock to learn on the side and iterate and ultimately improve the platform,” says Johnsen.
FlexZero also uses Bedrock to ingest data at scale – including data from utility bills and shipping logs – and apply it to LLMs to define which emissions factors to map. FlexZero additionally uses Bedrock for easy access to LLMs that power a chatbot and other services.
The company used AWS Sustainability Insights Framework, which helps to automate carbon footprint tracking, to create back-end data structures and access patterns.
Finally, FlexZero leverages the AWS cloud for improved visibility and speed. “Because you’re bringing everything into the cloud and into the platform, you have a full trail of all your data and you have quick responses,” says Johnsen.
Thanks to all this, FlexZero’s customers can now use the platform to automatically create industry standard ESG reports and analyze patterns to calculate future emissions. Through the generative AI-powered chatbot, users can ask the system in plain English about their emissions trends and identify key areas that most significantly impact progress toward environmental goals, such as achieving net-zero emissions by a specific target year.
FlexZero plans to continue adding capabilities to the system, such as custom decarbonization strategies that are based on a company’s specific environmental data. “I have a lot of different ideas about how to use tech to improve sustainability workflows,” says Johnsen.
Amazon eyes perfection
Shipping mistakes are costly. Studies show that rectifying a single incident typically costs between $38.50 and $58.50.1 That can quickly add up for large enterprises. At the same time, errors lower customer satisfaction and raise transportation emissions. What if there was a way to use generative AI to catch mistakes before they happen? That’s what Amazon is doing through an effort that could take fulfillment to the next level.
Amazon processes millions of products per year and aims to ensure that every order contains the right items and is damage-free. To ensure customers receive the products they expect, Amazon launched a quality control initiative known as Project P.I., which stands for Private Investigator.
Project P.I. uses cutting-edge technologies, including generative AI, computer vision and Optical Character Recognition, to automatically spot and isolate imperfect items to prevent them from shipping. Project P.I. also works to identify the root causes of errors to stop them from happening again.
“This is important for our planet because shipping imperfect products results in unwanted returns, wasted packaging and unnecessary emissions.”
With Project P.I. up and running, Amazon fulfillment centers across North America now have a powerful tool to boost accuracy. Items pass through imaging tunnels, where the system works to uncover defects like damaged products. It can also scan labels for details including product size and color. If a defect or incorrect order is found, Amazon isolates the product so it stays put and investigates further to determine if there is a wider issue with similar items.
“Project P.I. is very critical to Amazon’s customer-obsessed culture, and it’s a way to provide a positive shopping experience,” says Pingping Shan, director of perfect order experience at Amazon.
To preempt errors, Project P.I. uses a multimodal large language model to identify the root causes of negative customer experiences. The model compares customer feedback on erroneous orders to pictures from the imaging tunnel to confirm what led to the problem.
In the case of, say, an incorrectly sized sweater, the system can then ask questions like, “Does the label say ‘medium’ or ‘large’?” If the technology finds that the Amazon selling partner mislabeled the product, Amazon would alert the seller to help prevent the error from recurring.
Project P.I. is allowing Amazon to quickly root out problems, and it’s helping the company meet its environmental commitments. “This is important for our planet because shipping imperfect products results in unwanted returns, wasted packaging and unnecessary carbon emissions from additional transportation,” says Shan.
The system is enabled by technologies such as Amazon Bedrock, a fully managed service that lets users access a wide array of high-performing foundation models from leading AI companies and Amazon. This makes it easier for users to find the best model for their use case. Amazon is also using Amazon SageMaker, which gives developers access to powerful tools.
“These services have made it easier for us to develop and build gen AI applications, such as Project P.I.,” says Shan.
1 F. Curtis Barry & Company
