Measuring sustainability is complex. AI can help companies capture and analyze more data, leading to better estimates of sustainability performance. However, the quality of AI measurements is only as good as the underlying data itself. Unreliable data can lead to misleading results. Additionally, the way AI measures sustainability is driven by algorithms whose objective functions determine which factors are prioritized. Changes to these functions can shift the focus among social and environmental factors, affecting overall assessment results. Therefore, leaders need to stay actively engaged, understanding and critically evaluating the AI criteria used to analyze impacts, instead of relying solely on automated systems. Maintaining high data quality and close oversight is key to ensuring that AI-driven insights remain reliable and aligned with organizational goals.

AI’s ability to identify patterns makes it a powerful tool for forecasting and modeling. Modeling helps pinpoint the most pressing sustainability problems and explore feasible solutions. However, what looks good in theory may not always be practical. Managers therefore need to weigh what is technically possible against what is viable, given the nuanced context of the physical world we inhabit.

While monitoring, measuring, and modeling provide diagnostic insights, managing AI-driven initiatives means transforming insights into tangible actions. The main hazard here is determining which decisions AI can and should be used for. AI has enormous potential to optimize processes and lead to the development of new technologies, but companies may not always prioritize sustainability over other goals, such as operational efficiency. Leaders must recognize that while AI can deliver financial benefits, the temptation to use AI for other areas of the business may lead to unintended consequences, such as higher energy consumption or neglecting long-term sustainability goals.

The 'D's in the Butterfly Framework

The ‘3D’s in the framework – data, drivers, and decisions – act as the wings of the butterfly, with social and environmental effects represented on either wing. These components feed on the one hand into the ‘4Ms’, forming on the other hand the core mechanisms through which AI-driven actions impact both society and the environment.

The potential for AI for sustainability to result in unintended, adverse outcomes hinges on several variables. These include the data utilized to train and deploy AI tools, the factors that emerge from how algorithms optimize data analysis, and the decisions made based on potentially flawed or biased data. The likelihood of responsibility issues arises from the specific tasks AI is applied to and any errors or biases present in these components. For each M and on each wing of the butterfly, managers are encouraged to pause and reflect on the following:

• What is the quality of the data being used? The quality and availability of data are foundational to AI's effectiveness. Data is the raw material for AI learning, and insufficient or biased data can undermine outcomes. For instance, data gaps in carbon emissions reporting can lead to inaccuracies that influence decisions in harmful ways.

• How do algorithms process and prioritize data? AI algorithms drive how data is interpreted and utilized. Algorithms determine what gets optimized, whether it's cost, speed, or environmental benefit. As firms design their AI systems, they must be deliberate about defining their objectives to ensure that sustainability goals are prioritized rather than sidelined.

• How is AI being used to support decision-making without taking full control? AI-generated decisions can carry biases or overlook nuances that a human might notice. Firms need to maintain human oversight to ensure that AI's recommendations align with ethical and sustainability goals.

Putting the Framework into Practice

Now it is time to act. Using the butterfly framework is easy. First, define the ‘4Ms’: what to monitor, how to measure it, how to model impacts, and how to manage outcomes. Second, evaluate the ‘3Ds’ – data, drivers, and decisions. Assessing data quality, scrutinizing what drives AI outputs, and understanding which decisions AI will influence ensures that processes remain transparent and free of unintended bias. Third, identify key activities and anticipate the butterfly effect, that is the possibility that small changes can lead to largely altered and potentially harmful outcomes. Recognizing which actions could significantly impact society and the environment allows for proactive management of unintended consequences.

The Final Takeaway

Implementing AI solutions into sustainability strategies thoughtfully and responsibly will lead to meaningful change, helping businesses address today’s environmental and social challenges while preparing for those of tomorrow. The butterfly effect framework is intended to help managers do just that – to establish safeguards that prevent seemingly harmless AI solutions from causing cascading adverse effects for both society and the environment down the line.

Bibliography

Pinkse, J., & Bohnsack, R. (2024). Harnessing AI butterfly effect for sustainability: Digital boost or recipe for disaster? Amplify, 37(10), 44-51. https://kclpure.kcl.ac.uk/ws/portalfiles/portal/310667524/Pinkse_Bohnsack_2024_Amplify.pdf