The sustainability team at Royal Mail, Britain’s primary postal service, is testing a tool that uses artificial intelligence to estimate the potential carbon emissions that might be generated by new vendor contracts.
It took Emma Bayliss-Chan, head of climate strategy and risk at Royal Mail, six weeks to build the AI calculator as part of a fellowship program hosted this summer by carbon accounting software company Watershed.
The “dummy” example that she used to write the agent examined the potential impact of a hypothetical contract to buy additional handheld scanners as part of an existing automation initiative.
The AI agent estimated that buying more units would have a modest impact on Royal Mail’s indirect supply-chain footprint, which falls under the Scope 3 emissions category. That meant the hypothetical deal didn’t require additional screening. If the proposal had breached a predefined level, it would have required another review.
“The calculator is adaptable to a range of procurement scenarios,” Bayliss-Chan said. “It can handle standalone business cases, which may be an in-year investment; RFPs or tender responses, which typically reflect a three-year contract term period; and longer-term investment plans like our vehicle replacement program, which typically cover a period of five years.”
Program extension
Royal Mail already considers environmental criteria when scoring vendor contracts; these factors carry a 10 percent weighting in overall purchasing decisions.
The new calculator will be used only when the greenhouse gas emissions associated with the purchase might exceed certain thresholds — it’s not meant to analyze low-impact spending, Bayliss-Chan said. The tool was requested by the procurement team, which is responsible for helping the postal service cut Scope 3 emissions by 25 percent by 2030.
“We see it as our responsibility, as internal sustainability advisors, to make the business aware of the carbon impact of their decisions and then make an informed choice accordingly,” she said.
The agent was trained with the company’s financial accounts reference table, which includes widely used business economic analysis and emissions factors. It also considers currency exchange rates and Royal Mail’s internal carbon price.
Reports generated by the agent offer an estimate of the potential carbon costs from a project, the carbon intensity (measured as kilograms of carbon dioxide equivalent per British pound) and context about how the scenario ranks against other suppliers.
The agent can show tradeoffs. For example, it will point out when a more-expensive supplier might produce lower emissions.
Royal Mail’s agent will be tested by part of the procurement team before being rolled out broadly, but Bayliss-Chan is already looking at opportunities to apply the AI to other tasks, such as vehicle mileage and fuel usage.
Her advice for other sustainability professionals thinking about building a similar agent:
- Keep it simple: Royal Mail’s team created a straightforward input form that takes a few minutes to fill out, including a two-question screener that alerts procurement team members if their project falls below the threshold that requires screening.
- Pay close attention to workflow design: Bayliss-Chan started with an eight-step process that included 23 input fields. After testing, the calculator now uses five phases that require as few as eight fields.
“I involved procurement colleagues early to review the workflow and provide feedback themselves, which helped me design something they’d actually want to use rather than something that only made sense to the sustainability team,” she said.
