California Management Review

What seems to be happening today more than ever is the increasing role of artificial intelligence (AI) in the world economy. AI is becoming the driving engine of the great transformation in business to make the world economy more productive and efficient. However, we found that many initiatives to develop AI fail. These findings result from quantitive research among more than 340 executives and senior executives in Germany, Australia, Hong Kong, and Taiwan. These findings show that the cultural and structural barriers and the approach top managers adopt to AI, in which instead of continuous improvement, they look for rapid transformation and replacement, are among the most critical obstacles to developing AI in companies. This article is to help executives to develop an effective AI-powered implementation strategy to resolve these obstacles and reduce the probability of the failure of AI development projects. AI is increasingly changing the form of business all over the world.1, 2, 3 The way of decision-making focusing on gaining insights from data from the lowest to the highest levels of the organization is dramatically changing.4, 5, 6 Organizational processes are also becoming increasingly automated to bring extensive savings to companies worldwide and transform customer relationships.7, 8, 9 Google is a high-tech company that has built all its services based on digital and AI technologies. Google uses AI and automation extensively in its operations to create and provide solutions for developers with Google Cloud’s AI and for every user about research, marketing.10 This technology has given Google the extraordinary power to break free from all the limitations of traditional processes and quickly expand the scope of its processes to benefit from unique market opportunities. AI has also provided Google with better opportunities for continuous learning, allowing this company and customers who use this technology to stay ahead of the competition.11 Another successful company that has developed AI at scale is Alibaba. The Chinese high-tech company, now ranked among the most valuable companies in the world, is built on a digital core that has removed the limits of traditional business processes and created an incredible impact for this company worldwide.12 Digital Processes in Alibaba are exploited by algorithms that consider data the primary criterion for informing and supporting internal and customer decision-making to exploit e-commerce sells.13 However, despite the business successes that digital and AI have brought to big high-tech companies, our research shows that only 12% of the executives and senior executives participating in this research have included AI initiatives in their corporate strategies. Many of these participants have considered AI a tactical rather than a strategic factor, eventually failing to develop it in their companies. Forty-eight per cent of these executives and senior executives have expressed that the reason for projects failure is the approach that CEOs have adopted towards AI in which they consider AI as a rapid transformation and replacement, not a continuous development process14, 15, 16, 17 as it happens with an MVP.18 This approach prevents CEOs from developing an effective AI-powered strategy development. In this article, we are to make evident how the approach needs to change and present solutions for the cultural, structural, and lack of capability-building barriers that prevent organizations from moving toward the development of AI. By changing this approach and removing barriers, it will be possible to develop an AI-powered strategy. Organizations can more effectively benefit from the opportunities related to AI to better align with the new requirements of world business. AI-powered strategy is relatively new, with AI based on data and digital core. The emerging and consolidating positions of high-tech companies such as Google and Alibaba, initially built on a digital core, have caused a new era for businesses worldwide. We call this the new era of AI. Many small companies across the globe have been inspired and supported by companies such as Google and Alibaba. These small businesses are thriving in their new markets with AI’s development. All services which customers receive from small and big companies are based on data analysis algorithms and automated processes. The critical point in these new services is the extraordinary business capability created by these companies. In particular, data analysis provides these companies with a high predictive power about the future purchases of customers, which can ultimately improve the effectiveness of marketing and sales for these companies. Therefore, to change the approach of many CEOs who see AI as an early-return technology capability, we suggest that they consider AI not only a technological but also a critical business capability that can influence the expansion of sales and market share.19 The key point here is to focus on this critical business capability as a dependent variable requiring continuous organizational learning, developing technological infrastructures, and implementing structural and cultural changes. Obstacles to Implementing AI Many executives and senior executives participating in this research have noted that bureaucratic or hierarchical structure is a roadblock to developing a collaborative culture and forming cross-functional teams with business administration and IT/AI specialists. The research participants have declared that their bureaucratic structures prevent replacing decisions based on the intuition of CEOs with decisions based on data analysis. In contrast, flatter structures can contribute to the development of AI, in which decisions in the form of algorithms resulting from data analysis should be implemented by first-liners. Therefore, we suggest redesigning the corporate organizational structure to activate the delegation process related to decision-making from higher to lower levels in organizations about algorithms and chatbots. In addition, delegating authority to first-liners to implement algorithms requires the development of risk culture in companies. In developing this form of corporate culture, a learning approach should be replaced with a short-term profit-oriented approach. CEOs should consider algorithms as an essential resource of feedback for continuous learning for companies. This risk culture can also improve innovation in the algorithms of the company. Developing an Effective AI-Powered Implementation Strategy With the introduction of AI to a company, strategy development will change, and a new strategy development process will be redefined based on data analysis and digital applications. The first step is to assess human and technological infrastructure capabilities for AI, avoiding pitfalls in data analysis and further elaboration. Secondly, implementing an effective knowledge management system is one of the most significant technological and human infrastructures companies need before developing AI. Insights related to data analysis are usually available at operational levels, but the lack of an effective knowledge management system causes these insights to not pass through the bottlenecks of communication channels and are not available to upper levels. Here, developing chatbots and using other AI tools can lead to developing a data-oriented approach in companies and eventually strengthen the data analysis side in AI-powered strategy development. Another critical pillar of AI-powered strategy development is the digital core knowledge, which refers to the software on which algorithms derived from data analysis are applied. This step creates a more scientifical baseline for decision-making, and algorithms for hybrid automated processes are presented[. It is advisable to avoid software and technology choices that can act on the current CEO’s perception and research of rapid transformations and adoptions. The accelerated decision about technologies could create errors in the data to be utilized in strategy development and delays in effective AI implementation. AI requires processes redesigned to get advantages of automation (AI and RPA) along critical processes using chatbots (Customer service, Supply chain, HR,). This part of AI implementation is the opportunity to make the participation of internal resources effectively, especially those at the bottom line, to work on RPA coding and algorithmics. This can happen if a hybrid change process is allowed, which, under an effective and active sponsorship from the top, can remove the fear of technology from internal resources. The CEO’s role is to communicate technology’s scope and benefits with employees. As said earlier, a bottom-up approach with employees’ participation and decision-making power can lead to minor resistance and create a culture which, in addition to considering experimentation, can better align people and technology, leading to a successful implementation. In Conclusion AI will transform and enhance decision-making and organizational processes. These transformations will bring extensive benefits to companies. Companies that use this emerging technology have a higher competitive advantage when compared to companies that only focus on one of the two aspects of machines and humans. The change in the approach of CEOs as well as structural and cultural changes will become a basis for developing an effective implementation strategy to better respond to new needs. This AI-powered Business strategy, relying on data analysis and AI and digital technology, has a high potential to respond effectively to the emerging needs of today’s evolving business environment. References Enholm, I.M., Papagiannidis, E., Mikalef, P. et al. (2022) Artificial Intelligence and Business Value: a Literature Review. Information Systems Frontiers, Vol. 24, pp. 1709–1734. https://doi.org/10.1007/s10796-021-10186-w Demir, F. (2022). Artificial Intelligence. In: Innovation in the Public Sector. Public Administration and Information Technology, vol 39. Springer, Cham. https://doi.org/10.1007/978-3-031-11331-4_4 Davenport, T., Guha, A., Grewal, D. et al. (2020) How artificial intelligence will change the future of marketing. Journal of the Academy of Marketing Science, Vol. 48, pp. 24–42. https://doi.org/10.1007/s11747-019-00696-0 Sousa, M.J., de Barros, G.O., Tavares, N. (2022). Artificial Intelligence Trends: Insights for Digital Economy Policymakers. In: Guarda, T., Anwar, S., Leon, M., Mota Pinto, F.J. (eds) Information and Knowledge in Internet of Things. EAI/Springer Innovations in Communication and Computing. Springer, Cham. https://doi.org/10.1007/978-3-030-75123-4_8 De Mauro, A., Sestino, A. & Bacconi, A. (2022) Machine learning and artificial intelligence use in marketing: a general taxonomy. Italian Journal of Marketing, Vol. 2022(4), pp. 439–457. https://doi.org/10.1007/s43039-022-00057-w Sarker, I.H. (2022) AI-Based Modeling: Techniques, Applications and Research Issues Towards Automation, Intelligent and Smart Systems. SN Computer Science, Vol. 3, pp. 1-20. https://doi.org/10.1007/s42979-022-01043-x Itagi, S., Gowda, S., Udupa, T., Shylaja, S.S. (2022). Future Frame Prediction Using Deep Learning. In: Raje, R.R., Hussain, F., Kannan, R.J. (eds) Artificial Intelligence and Technologies. Lecture Notes in Electrical Engineering, vol 806. Springer, Singapore. https://doi.org/10.1007/978-981-16-6448-9_21 Sharma, M., Kumar, C.R.S. (2022). Machine Learning-Based Smart Surveillance and Intrusion Detection System for National Geographic Borders. In: Raje, R.R., Hussain, F., Kannan, R.J. (eds) Artificial Intelligence and Technologies. Lecture Notes in Electrical Engineering, vol 806. Springer, Singapore. https://doi.org/10.1007/978-981-16-6448-9_19 Choudhary, R., Karmel, A. (2022). Robotic Process Automation. In: Raje, R.R., Hussain, F., Kannan, R.J. (eds) Artificial Intelligence and Technologies. Lecture Notes in Electrical Engineering, vol 806. Springer, Singapore. https://doi.org/10.1007/978-981-16-6448-9_3 GOOGLE_Inc. (2023). Tools and Resources. ai.google: ttps://ai.google/tools/ GOOGLE_Inc. (2023). Tools and Resources. ai.google: ttps://ai.google/tools/ Zeng, M. (2018, September -October). Strategy- Alibaba and the Future of Business. Harvard Business Review. https://hbr.org/2018/09/alibaba-and-the-future-of-business Zeng, M. (2018, September 11). How Alibaba Is Leading Digital Innovation in China. Harvard business Review. https://hbr.org/podcast/2018/09/how-alibaba-is-leading-digital-innovation-in-china Kersting, K. (2018). Making AI Smarter. KI - Künstliche Intelligenz, Vol. 32, pp. 227–229. https://doi.org/10.1007/s13218-018-0562-8 Davenport, T. H., & Seseri, R. (2020, December 15). What Is a Minimum Viable AI Product? MIT Sloan Management Review. https://sloanreview.mit.edu/article/what-is-a-minimum-viable-ai-product/ Barriga, A., Rutle, A. & Heldal, R. (2022). AI-powered model repair: an experience report—lessons learned, challenges, and opportunities. Software and Systems Modeling, Vol. 21, pp. 1135–1157. https://doi.org/10.1007/s10270-022-00983-5 Daley, S. (2023, February 17). 36 Artificial Intelligence Examples Shaking Up Business Across Industries. BUILT-IN. https://builtin.com/artificial-intelligence/examples-ai-in-industry Belhadi, A., Mani, V., Kamble, S.S. et al. (2021). Artificial intelligence-driven innovation for enhancing supply chain resilience and performance under the effect of supply chain dynamism: an empirical investigation. Annals of Operations Research. https://doi.org/10.1007/s10479-021-03956-x Davenport, T., & Ronanki, R. (2018, Jan-Feb). Artificial Intelligence for the Real World. Harvard Business Review. https://hbr.org/webinar/2018/02/artificial-intelligence-for-the-real-world

Recycling can be a powerful tool to keep waste out of landfills and even turn it into profitable manufacturing inputs. Unfortunately, the American recycling system has five major shortcomings that result in only 32.1% of waste being either recycled or composted. This article describes the current shortcomings of the American recycling system and explains how those shortcomings cause recyclable material to be landfilled (called ‘leaks’). This is the first article in a two-part series on America’s recycling industry. The second article, entitled “Is it time for a national recycling standard?” discusses how standards could improve financial and environmental outcomes. Americans were among the early leaders in curbside recycling. However, today the United States lags behind in global recycling efforts. The current regulations governing solid waste management in the U.S. were codified under the Resource Conservation and Recovery Act1 in 1976. That means it’s been more than 47 years since the U.S. updated its approach to waste management. The dated approach of the U.S. recycling system is bad for American businesses. For example, the U.S. lacks geographic consistency in which products are accepted for recycling. Research, including our survey of nine major cities, shows that inconsistent recycling practices lead to consumer confusion and sorting errors. These sorting errors are costly – particularly putting a non-recyclable product into the recycling stream, called contamination.  Contamination costs the material recovery facilities that sort recyclables at least $300 million per year in additional labor, processing, and machinery repairs.2 Ultimately, this increases the cost to downstream manufacturers. America’s dated approach to recycling is also bad for the environment. About 1.8 million acres of land in the U.S. is lost to landfills. Recycling diverts waste from these landfills, yet only 32.1% of American waste is currently recycled or composted4 There is so much opportunity for the recycling market to create more financial value and conserve our natural resources.  The U.S. recycling market has five critical shortcomings. Before we dive into the details of these shortcomings, it’s helpful to see a quick overview of how the industry works. Understanding America’s Recycling System Figure 1: America’s recycling system Notes: The double arrows are meant to represent the decision making process at the sorting facility (e.g. accept the load and begin sorting, or reject the load because it is too contaminated and send it to landfill.) Step 1: A material recovery facility decides which products to accept. The waste management process begins when a municipality selects a material recovery facility to service their region. The material recovery facility is a private or public company that sorts recyclables into clean streams of products that can be sold for remanufacturing. The recovery facility makes decisions about which products it will accept based on (1) the market for recyclables, (2) the equipment and labor it possesses, and (3) the scale of potential materials that can be collected. Step 2: Consumers buy products, then dispose of waste and recyclables. Consumers purchase goods and dispose of the resulting waste at home. This process involves sorting recyclables from non-recyclable waste. The consumer’s ability to dispose of waste accurately depends on their knowledge of which products are recyclable in their region. Consumers typically don’t spend much time thinking about how to correctly dispose of their waste, yet knowing which items are recyclable is critical, as we will show in this article. Step 3: A logistics company does curbside recyclable collection. A logistics company (sometimes owned by the same parent company as the material recovery facility) is responsible for collecting curbside waste. The logistics company collects household recyclables and delivers the truckloads to a material recovery facility to be sorted into individual streams. Step 4: A material recovery facility screens materials for contamination. The raw recyclable waste that comes in from logistics companies will cost less to process if there is little to no contamination.5 At the recovery facility, the logistics vehicle is weighed on a scale and a visual inspection is performed. The inspection is intended to evaluate the quality of the raw recyclable waste and determine if the contamination will impact the recovery facility’s processing capabilities. If the contamination exceeds the tolerable threshold, the load is rejected and brought to the landfill. Step 5: Raw recyclables are sorted into individual streams. Once a load is accepted, the recovery facility separates the raw recyclable waste into its individual products, for example, corrugated cardboard or #2 plastics. The process for this varies from facility to facility. Some facilities depend heavily on human labor to sort products while others have more technology to automate the sorting process. Step 6: Sorted recyclables are sold as non-virgin inputs for manufacturing. Once the recyclables are sorted, the individual products can be sold for re-manufacturing. The material recovery market is divided into six regions across the U.S. The market is similar to other commodity markets where the price is determined by a bid system.6 Long-term contracts also exist, albeit less popular because this may be more costly to the processor. The cost a material recovery facility pays to collect, sort, compact, and bail recyclables remain fixed, but the demand and the price they can charge for their product fluctuates. This means that sometimes recovery facilities are profitable – and sometimes they are not.7  Step 7: Manufacturers make products from non-virgin materials Manufacturers interested in closing the loop can commit to sourcing non-virgin materials as inputs to their production. Yet, they will only do so if they find that purchasing recycled material is more cost-effective or profitable. To close the loop, the recycling industry needs to scale and become more efficient at collecting, segregating, and reselling recycled material. Five shortcomings of the Recycling System: A Study of Recycling Practices across Nine Cities To better understand how the U.S. recycling system is performing, our team conducted a survey of recycling practices across nine major American cities in 2020. We selected cities with variations in how they manage residential recycling and waste. The nine cities include San Francisco, Los Angeles, Seattle, Austin, Columbus, Nashville, Boston, Portland, and Chicago. We then recruited 282 participants who reported living in one of the nine cities through Amazon Mechanical Turk. We included 25 items in our survey. We purposefully chose some products that were landfilled across all nine cities, some that were recycled across all nine cities, and some with municipal variation in end-of-life practices. Table 1: Products included in the survey The sections below use our survey findings to characterize the five core shortcomings of the U.S. recycling system. Figure 2: Five shortcomings of the recycling system 1. Supply and demand volatility Some products, such as corrugated cardboard and rinsed plastic jars, have predictable processing costs, stable demand from manufacturers, and stable supply from consumer waste streams. This is good for all businesses in the recycling supply chain. These products offer a consistent return on investment for processing facilities and reliable availability for manufacturers. Unfortunately, market stability is not the reality for most recycled products. Supply volatility happens because recycling practices vary widely from region to region and over time. For example, our study included nine products that were accepted in some cities and not in others, including glass jars, plastic clamshells, and laminated cartons. This variation happens because regional recovery facilities make autonomous choices about which products to accept based on the expected return on investment, which introduces geographic differences in supply availability. Processing facilities can also change the products they accept over time, which introduces temporal volatility in supply. There are also inconsistencies in the classification of grade or quality. A manufacturer may pay for non-virgin material of a certain grade, measured using the weight of recyclable content to total weight of the bale, but receive material of lower quality. In some cases, the quality of a bale is not guaranteed prior to shipment, and buyers are aware of these risks in purchasing from processors. Short-term demand volatility can result when manufacturers buy in batches. For example, if a paper manufacturer commits to using 50% recycled paper in their products, that 50% is an annual average. This means they can choose to purchase no recycled material when prices are high, then buy in bulk later in the year if prices drop. Consumers also pay the cost of volatility through sustainability/recycling adjustment fees and rebates, which gives customers a credit when prices at the market are favorable, and charges customers a fee when prices at the market are unfavorable. 2. Consumer confusion The consumers that participate in recycling play the critical role of sorting recyclables at the point of disposal. The accuracy of consumer disposal decisions directly influences the performance of the recycling system. In one survey, 94% of U.S. residents said they support recycling,8 reflecting a strong willingness to participate but a striking mismatch with the actual rate of recycling at 32-34%.9 The effectiveness of recycling further deteriorates due to contamination. In a separate study, roughly one out of four items (or 25%) are incorrectly placed in the recycling bin.10 Our findings in this study indicate that the error rate for recycling is even higher than these two studies suggest. The accuracy with which consumers correctly dispose of items into the recycling dramatically impacts the effectiveness and cost of the recycling system. In waste disposal, not all errors are the same. There are two types of errors that consumers can make as they sort their waste – false negatives and false positives. A false negative refers to the scenario where a consumer throws a recyclable product into the garbage. This represents another leak in the recycling system, whereby an item that could have been recycled does not get recycled. A false positive, also called ‘contamination,’ occurs when a consumer puts a non-recyclable product into their recycling bin. The cost of contamination is much higher than the cost of throwing a recyclable in the garbage. We will return to this cost argument in the next subsection. The geographic consistency of recycling practices helps consumers to know with more certainty which products are recyclable. Our results show that geographic consistency in which products are accepted for recycling is a key factor in improving sorting accuracy. Our data show that products that are recyclable across all nine cities had a very low error rate of 8%. Products that were not accepted in any of the nine cities had a higher disposal error rate, at 33.7%, but this was still lower than the 52% error rate of products that were accepted in some cities but not in others (See Figure 3). Additionally, our study shows that the more cities that recycle a particular item, the lower the error rate becomes. The correlation between disposal accuracy and the number of cities that recycle an item is 0.63, supporting the idea that a more uniform recycling practice across the U.S. could lead to lower disposal errors. Figure 3: Disposal errors by product 3. Contamination costs When a consumer throws a recyclable item into the garbage, it results in an opportunity cost; one item whose value could have been captured through recycling, which will instead sit in a landfill. There are also landfilling costs of throwing away recyclable materials, but landfilling costs are substantially smaller than the cost of processing a contaminated recycling supply chain. Contamination cost is the cost when a consumer throws a non-recyclable into their recycling bin. The contamination cost is higher than the opportunity cost and the cost to landfill combined. Contamination is costly for logistics companies. Collected curbside recycled materials are screened by the recovery facility. If the screening reveals excessive contamination, the entire load is sent to the landfill. This creates another leak in the recycling system, whereby large volumes of recyclable materials do not get recycled.  According to data from New York City, the average cost to collect recyclables is $686 per ton.11 Contaminated recyclable loads are disposed of at an average of $80 per ton.12 This means that contaminated recyclable loads cost logistics companies as much as $766 per ton. In contrast, the average cost of collecting waste without segregation is $126.03 per ton. This data is from New York, so it cannot be directly extrapolated to other cities. There is lack of data for other municipalities, commanding the need for more research in this area. The total contamination costs across all recovery facilities in the U.S. is at least $300 million per year in additional labor, excess processing, poor material quality, longer downtime due to equipment damages,13 and increased safety hazards. This results in recovery facilities often being cost centers instead of profitable businesses. Ultimately, this increases the cost to downstream manufacturers. Unfortunately, past research shows contamination (false positives) is a far more common disposal error than false negatives. This happens because of ‘wish-cycling,’ a phenomenon where well-intending consumers encounter an item that they are not sure how to sort and end up putting it in the recycling bin in the hopes that the item can be recycled. In our study, within the group of recyclable items accepted in some cities but not others, the overall error rate was 52% (the middle bar graph in Figure 3). Of those errors, 87% were false positives (or contamination). This captures wish-cycling in action. When people are unsure how to dispose of something, they typically try to recycle it. Our study also sheds light on a potential solution to contamination. Although there is still a significant amount of disposal error amongst products that are not recycled in any of the cities, the false positive rate of this category of products is 12% lower than the false positive rate amongst products that are accepted in some cities and not in others (see the middle and rightmost cluster of products in Figure 4). This finding suggests that more uniform policies regarding what is and is not accepted could reduce contamination costs. Figure 4: Average disposal error rates across different categories 4. Barriers to international markets Contamination also poses a unique challenge for U.S.-based recycling exporters. Prior to 2018, China was a leading global importer of recycled materials and had very high tolerances for contamination. This meant that American recycling exporters could send even highly contaminated loads of recyclables to China for profit. As of 2017, 31% of all U.S. scrap went to China.14 In 2018, China implemented a strict waste import policy called National Sword. With the implementation of the National Sword, Chinese inspectors could only accept loads if15 they contain less than 0.5% contaminated product in plastic bales, 1% in ferrous and nonferrous metals, and 1.5% in scrap paper.16 The average quality requirement in the U.S. is 25% contamination. American recyclers cannot meet these standards, so the profitability of recycling crashed overnight.17 The effects were particularly acute in coastal recycling markets, which depended more heavily on export revenue. 5. Inconsistent data collection Low-quality and contaminated recycling materials are pre-competitive challenges. These issues impact the profitability all companies in the recycling industry, including logistics companies, material recovery facilities, and manufacturers and reduce the competitiveness of the U.S. recycling industry on an international scale.  Yet there is very little data available and minimal sharing of best practices to improve collective performance. During our study, our team discovered an extreme lack of data on core topics, such as which products are accepted in each municipality. The only way to find this data was to examine each municipality’s website individually. There is even less data available related to the contamination rates of raw or processed materials. Reporting and accessing these data is the first step to understanding which processes cause and prevent contamination and disseminating these best practices nationally. Cost estimates, the volume of traded recyclables, and contamination rates are hard to compare, so it is very difficult to conclude what is and isn’t working.   Shortcomings Create Leaks in the Recycling System The shortcomings identified above ultimately create three major ‘leaks’ in the recycling system, whereby potentially recyclable materials end up in the landfill. Those leaks are: Leak 1: Material recovery facilities don’t accept products with volatile demand or price, meaning those products are sent to landfill. Leak 2: Confused consumers throw recyclable products into the garbage. Leak 3: Contaminated loads or raw recycling collected by a logistics company are rejected by material recovery facilities and sent to landfill.   Figure 5: Three Leaks in the American Recycling System These leaks are so significant that only 32.1% of American waste is currently recycled or composted.18 These leaks hurt the U.S. financially, as companies miss the opportunity to capture financial value from waste. Other hidden costs of landfills are hard to quantify, including Greenhouse Gases that contribute to climate change, the release of toxic and forever chemicals that pose health issues, and the deterioration of property and land values.19,20 The leaks also hurt the U.S. environmentally, as recyclable materials needlessly end up in landfills. There is so much opportunity to patch leaks in the American recycling system, so we can create value and conserve our natural resources. Patching the leaks will require a market that is able to carefully and efficiently balance supply and demand and radically reduce the costs of contamination. In the second article, we discuss the potential role of a national recycling standard in strengthening the U.S. recycling system. References EPA, “Resource Conservation and Recovery Act (RCRA) Regulations,” August 2022, https://www.epa.gov/rcra/resource-conservation-and-recovery-act-rcra-regulations#nonhaz. Recycling Partnership, “2019 West Coast Contamination Initiative Research Report,” May 2020, http://recyclingpartnership.org/wp-content/uploads/2020/04/The-Recycling-Partnership_WCCI-Report_April-2020_Final.pdf. Vasarhelyi, Kayla. “The Hidden Damage of Landfills.” University of Colorado Boulder, Environmental Center. April 2021, https://www.colorado.edu/ecenter/2021/04/15/hidden-damage-landfills. EPA, “National Overview: Facts and Figures on Materials, Wastes and Recycling,” December 2022, https://www.epa.gov/facts-and-figures-about-materials-waste-and-recycling/national-overview-facts-and-figures-materials. EPA, “Frequent Questions on Recycling,” Last accessed November 2022, https://www.epa.gov/recycle/frequent-questions-recycling#recycling101.  Waste 360, “Reducing Recycling’s Risks: Share the Pain, Share the Gain,” January 1998,  https://www.waste360.com/mag/waste_reducing_recyclings_risks Iyer, A. V., Vedantam, A., & Lacourbe, P. (2023). Recycled content claims under demand benefit and supply uncertainty: Multi-period model and application to glasswool insulation. European Journal of Operational Research. Storymaps ArcGIS, “Contamination in Recycling,” July 2021, https://storymaps.arcgis.com/stories/145063284501409cb2516770ebaf0865. Storymaps ArcGIS, “Contamination in Recycling,” July 2021, https://storymaps.arcgis.com/stories/145063284501409cb2516770ebaf0865. Recycling Partnership, “2019 West Coast Contamination Initiative Research Report,” May 2020, http://recyclingpartnership.org/wp-content/uploads/2020/04/The-Recycling-Partnership_WCCI-Report_April-2020_Final.pdf. Husock, Howard. “The Declining Case for Municipal Recycling.” June 2020. https://www.manhattan-institute.org/recycling-cost-benefit-analysis. Ibid. Recycling Partnership, “2019 West Coast Contamination Initiative Research Report,” May 2020, http://recyclingpartnership.org/wp-content/uploads/2020/04/The-Recycling-Partnership_WCCI-Report_April-2020_Final.pdf. Miller, Randy. “Contamination in Recycling Costs Business Money.” January 2020. https://millerrecycling.com/recycling-contamination-costs-money/. Miller, Randy. “One Month into China’s National Sword Recycling Program.” February 2018. https://millerrecycling.com/one-month-china/. Katz, Cheryl. “Piling Up: How China’s Ban on Importing Waste Has Stalled Global Recycling.” Yale Environment 360. March 2019, https://e360.yale.edu/features/piling-up-how-chinas-ban-on-importing-waste-has-stalled-global-recycling. Vedantam, Aditya, Nallan C. Suresh, Khadija Ajmal, and Michael Shelly. “Impact of China’s National Sword Policy on the U.S. Landfill and Plastics Recycling Industry.” Sustainability 14, no. 4 (2022): 2456. EPA, “National Overview: Facts and Figures on Materials, Wastes and Recycling,” December 2022, https://www.epa.gov/facts-and-figures-about-materials-waste-and-recycling/national-overview-facts-and-figures-materials. Recycling Partnership, “The Hidden Cost of Landfilling vs. Recycling,” January 2023, https://recyclingpartnership.org/the-hidden-cost-of-landfilling-vs-recycling/. EPA, “Basic Information about Landfill Gas,” September 2022, https://www.epa.gov/lmop/basic-information-about-landfill-gas.

This is the second article in a two-part series on American recycling. The first article described the current shortcomings of the American recycling system and explains how those shortcomings cause recyclable material to be landfilled (called ‘leaks’). In this article, we discuss how a national recycling standard could improve financial and environmental outcomes. We also explore the limitations of such a standard and identify places where more research and discussion are needed. We believe that a solution to a broken recycling industry in the U.S. is developing recycling standards, which include a common, minimum set of recycled products across all major cities, and standardizing data collection and reporting. There are many other industries where standards have paved the way for a more vital, robust industry, including healthcare, electricity markets, airlines, information technology, and telecommunications. We hope our two-part issue will spark critical conversation and action in the U.S. recycling sector. We may not have all the nuts and bolts on what the perfect standard should look like, but we attempt to set a framework for future research and articulate open questions in the recycling market. We hope that this article challenges the status quo and spurs discussions on whether it’s time to consider a national recycling standard. Fixing the Broken System: The Need for a National Recycling Standard We believe that the U.S. recycling system needs updating and that the best first step is the creation of a national recycling standard for curbside recycling programs. Such a standard may have the power to: Reduce market volatility and create new demand for non-virgin materials. Reduce consumer confusion. Lower contamination rates and related financial costs. Improve access to international markets. Encourage the dissemination of best practices, thus improving the performance and profitability of the entire American recycling sector. Below, we describe what such a standard could look like and provide the rationale for our suggestions. The goal of this paper, however, is not to provide ready-to-implement policy recommendations. Instead, the goal of this paper is to summarize the crucial ingredients to create a strong recycling market, spur discussion amongst researchers, recycling businesses, and policymakers, and identify gaps in the industry that require more research or market development.   Recommendation 1: Uniform Recycling Requirement Description We recommend that major cities standardize a minimum set of recycled products in their curbside pickup programs. This minimum list of products should be shaped by: Which common materials that have high demand from manufacturers. Which products are most financially valuable to the circular economy. Which products are most commonly disposed of by U.S. residents, including paper, cardboard, glass, and plastic bottles. Rationale We describe at least four potential benefits of minimum recycling standards: It would send a strong market signal to recyclers and manufacturers. Having a steady supply of recycled materials will build confidence among manufacturers to commit to sourcing a significant portion of their raw materials from recyclables. Manufacturers could better predict and negotiate price and volume. Having a minimum set of recycled products makes the volume for these products more predictable, creating a more stable recycling market. This, in turn, may increase demand. A minimum recycling requirement would create economies of scale through volume and pooling of resources. The reason why some municipalities do not collect certain types of material is that they may not have the resources or do not have the volume. In contrast, if municipalities have minimum recycling standards, several neighboring municipalities can pool their collections to better justify volume. Pooling collections of recyclable materials can also create economies of scale when transporting them. Recycling standards can encourage larger investments in better technology brought by (1) and (2). Recycling standards can lower consumer confusion. The results we discussed in the first article reveal that items that are recycled across all cities we surveyed had a very low error rate of 8%. Additionally, our study shows a correlation between disposal accuracy and the number of cities that recycle an item is 0.63. This suggests that the more cities recycle a particular item, the lower the error rate. Introducing a minimum recycling standard is not likely to impose substantial costs because most (if not all) large municipalities are already collecting most of these products. Instead, the primary value of a standard would be to encourage a predictable supply of materials to recovery facilities so that manufacturers can commit to sourcing more low-cost recycled content. The U.S. has already used minimum recycling requirements outside of curbside recycling programs with positive results. For example, we have seen this with cardboard and newspaper waste streams. The United States EPA’s Code of Federal Regulations 40 CFR Part 2461 recommends the following requirements, which denote the minimum required actions: (1) High-grade paper must be recycled at office facilities over 100 office workers; (2) Newspapers must be recycled at all facilities in which more than 500 families reside; and (3) Corrugated cardboard must be recycled at any commercial establishment generating 10 or more tons of waste corrugated containers per month. These requirements have created stability in the supply and demand of these products. One could posit that the requirements for consumers to separate specific, profitable, recyclable items has the potential to reduce the volatility in supply and demand. Furthermore, because the supply of these three streams has been stable over time, the processes to segregate from even single-stream waste streams have been developed and fine-tuned, further diminishing lost value.   Recommendation 2: A Common Communications Campaign Description A common communications campaign would teach consumers across the U.S. which products are universally accepted or rejected, and encourage them to throw products in the garbage if they find themselves uncertain about regional disposal practices. Rationale Currently, local collection companies or municipalities shoulder the burden of educating local consumers about which products are accepted for recycling in their area. This reflects the geographically fragmented market that creates recycling collection variation. Once a recycling standard is established, there would be significant efficiencies in having a common communications campaign to educate all U.S. consumers.  Part of the campaign could also be the creation of a new symbol to denote “universally accepted” products. Currently, the recycling logo alone doesn’t provide useful sorting information, as the consumer still has to understand whether their region actually has the capability to recycle that item. But a ‘universally accepted’ logo would clarify that confusion easily for items included in the minimum recycling requirement.   Recommendation 3: Standardized Data Collection and Reporting Description We recommend that municipalities collect the same data regarding their recycling processes and share their data publicly at consistent time intervals. Here is a short list of data that we believe would be valuable to collect for each municipality: Amount of incoming raw recyclable material Average contamination rate of raw inputs and processed outputs Average landfill rate and load rejection Post-recovery volume and yield of each product stream Facility-level technology inventory The process used to collect and calculate these figures should also be standardized to ensure that data are comparable across municipalities. Rationale There are at least three benefits of creating a standardized approach to what gets measured and how frequently this data is collected. First, measurement will allow municipalities and facilities to measure progress against targets and ensure performance metrics are comparable across facilities and regions. Having a common metric will allow different locations to benchmark against each other to identify leaders and laggards. Identifying industry leaders can help facilitate the diffusion of best practices. Identifying industry laggards can direct support or resources to areas that need improvement. Second, manufacturers that buy recycled material as inputs can better compare quality and cost across different municipalities and recycling facilities, creating a more competitive, healthy recycling economy. Third, having a national standard on data reporting can enable better assessments of policies that may benefit or harm recycling markets. For example, without having standard measures of contamination, it may be difficult to compare the effect of a policy implemented in one geographic area and its impact on contamination in other areas if data is not available or comparable.     We have already seen standardized reporting work well in other commercial and industrial sectors of the U.S. One example is the Toxic Release Inventory (TRI) Program managed by the Environmental Protection Agency (EPA). Each year, industrial facilities of a certain size submit data to the EPA. Only certain facilities that meet the threshold report to the EPA to avoid imposing undue burden on smaller facilities.2 The TRI was a success because it made it easy for companies to identify compounds that should be avoided. In turn, this made it possible for buyers to align with their suppliers on what types of compounds are not good for the environment. The industry uses TRI data to set reduction targets and track progress in reducing harmful chemicals. Recycling Standard Limitations and How to Address Them We acknowledge there are limitations, but with proper transition planning and implementation, the benefits can substantially outweigh the costs. Limitation 1: Cost to Material Recovery Facilities A national standard may impose a cost on facilities that don’t recycle a required material. This can be burdensome for that municipality, which is why we recommend that a national standard only be applicable to large municipalities, similar to how toxic reporting is only limited to facilities of a certain size. The population threshold would help avoid the burden of requiring small municipalities to collect items that they currently cannot cost-effectively recycle. An alternative is to develop market mechanisms to help smaller recovery facilities transition into the standard. Smaller municipalities can financially and environmentally benefit from widening their recyclable collections, but they may need private and public support to get there.  Although it is beyond the scope of this work to identify which items should and should not be nationally recycled, we have started to lay the groundwork on what type of data is necessary to identify candidates for items that should be recycled regardless of the location in the U.S.. Many of these items are already recycled on large scales (e.g., paper, cardboard, glass, aluminum, PET, and HDPE to name a few). Standard can encourage manufacturers to increase their commitment in purchasing these recycled materials because its supply can be more consistent and predictable.    Limitation 2: Lack of Incentives or Penalties Standards can make it more convenient to recycle. Making recycling easy to follow can increase the scale and cost-effectiveness of recycled products. Standards can reduce or eliminate the confusion around what is and is not recyclable, even when residents move to a different location. However, most of the existing literature on reducing recycling contamination has focused more on behavioral interventions.3 Behavioral interventions may be cost-effective, but they have limits.4 It may be better to focus on how to make recycling easier for the entire population. Future studies can compare whether non-pecuniary measures for recycling (e.g., standards) are more or less effective than pecuniary interventions.   Limitation 3: Lack of Data to Inform Standard Creation. Our study collected data on which products are recycled across nine major U.S. cities, but more data is needed. Future studies can collect more comprehensive data on what is and is not recycled across the U.S. to get a fuller picture of how close or far the U.S. is on implementing a national recycling standard. Based on our preliminary data, there are many promising directions for a national standard because there are already a lot of items currently recycled in all nine cities we examined. We speculate that this pattern may hold for most if not all of the major cities in the U.S. Our exercise in gathering recycling data shows how fragmented the U.S. recycling industry is but it also shows the potential value that data can unlock for the U.S. recycling market. Acknowledgements We want to thank Chelsea Hicks-Webster for helping us write this two-part series on the U.S. recycling industry and for providing the illustrations. Code of Federal Regulations. “Part 246 – Source Separation for Materials Recovery Guidelines.” Last accessed January 2023, https://www.ecfr.gov/current/title-40/chapter-I/subchapter-I/part-246. EPA, “What is the Toxic Release Inventory?” July 2022, https://www.epa.gov/toxics-release-inventory-tri-program/what-toxics-release-inventory. Rosenthal, S., & Linder, N. (2021). Effects of bin proximity and informational prompts on recycling and contamination. Resources, Conservation and Recycling, 168, 105430. Donnelly, G. E., Blanco, C. C., Spanbauer, C., and Stienecker, S. (2023). The Effects of Item Dirtiness on Disposal Decisions. Journal of the Association for Consumer Research.

Introduction In the domain of management practice and research, a substantial transformation is unfolding in contemporary history, primarily attributable to the advent of machine learning methodologies (George et al., 2014). These advanced techniques harness extensive data repositories to project individual and group behaviors with remarkable accuracy. This progression is intrinsically linked to the broader field of artificial intelligence (AI), an area that has witnessed tremendous growth and enhancements (Brynjolfsson & McAfee, 2014; Haenlein & Kaplan, 2019), profoundly reshaping the modus operandi of businesses, organizations, and societies as a whole. A noteworthy development within this arena, is the emergence of natural language processing models or NLPs (Hirschberg & Manning, 2015), which possess the ability to decode, analyze, and construct text that bears an uncanny resemblance to human-generated prose. Among these innovative models, the GPT architecture by OpenAI is distinguished as a state-of-the-art breakthrough (Clarke et al., 2021), showcasing exceptional proficiency in producing text that closely mimics human authorship. The recent introduction of latest generation of GPT architecture unveiled in March 2023, GPT-4, promises to disrupt traditional work structures and processes in numerous ways. Its ability to perform tasks such as content creation, data analysis, and customer support has led to a reevaluation of human and AI roles within organizations. The GPT-4 model, and similar advanced NLPs, exhibits a remarkable semblance to human-like performance, thereby revolutionizing diverse industries, augmenting decision-making processes, and fostering unparalleled opportunities for ingenuity and synergy across multiple sectors. This pivotal development has been aptly christened as the ‘inception of authentic artificial intelligence’ (Economist, 2023). As a cutting-edge language model devised by OpenAI, GPT-4’s potential to engender a profound influence on management practice and research is rapidly becoming evident. As AI models like GPT-4 become increasingly integrated into businesses, organizations need to adapt their strategies and human resources practices accordingly (Agrawal et al., 2019). This involves redefining job roles, redesigning organizational structures, and reskilling employees to capitalize on the unique capabilities of both humans and AI (Bessen, 2019; Manyika et al., 2017). Synergistic Frontiers: A Typology In this post, we delve into the tensions surrounding the relationship between human expertise and AI-driven language models, specifically focusing on the GPT-4 language model and its integration in management practice and research. Rather than perceiving these NLP models as threatening or substituting human capabilities, our framework envisions a synergetic interplay between human expertise and the AI-driven instruments in the realm of management practice. Drawing inspiration from existing literature on digital technologies and AI within organizational contexts (Brynjolfsson & McAfee, 2014; Kaplan & Haenlein, 2019), we argue that AI-driven mechanisms will serve to enhance, rather than supplant, human faculties. Consequently, the trajectory of management practice and research may be contingent upon cultivating a reciprocal alliance with AI, wherein both human and machine intelligence collaborate in a synergistic manner to further the boundaries of knowledge in the discipline.   While the rapid evolution of AI-generated content raises ethical and legal concerns, such as intellectual property rights and the potential for misinformation or manipulation (Brundage et al., 2018), our primary objective is to explore typologies of user types in the context of GPT-4 usage and similar advanced NLP models. By focusing on information retrieval and creative applications, we aim to provide a comprehensive framework for understanding and employing this technology in diverse contexts. In examining the nascent schools of thought surrounding the implications of these technological advancements for businesses and management scholars, we acknowledge the ongoing debates concerning ethics, workplace implications, and changes in management routines. However, our focus remains on contemplating typologies of user types, represented by a two-by-two matrix. By comparing the dual dimensions of GPT-4 usage—information retrieval versus creative applications—against the categories of users who can harness its potential, we aim to provide a comprehensive framework to better understand and navigate the diverse ways in which this groundbreaking innovation can be employed across various contexts. In Figure 1, we propose a 2x2 matrix of ChatGPT-4 (and similar NLP models) user archetypes, based on two primary dimensions: the purpose of the interaction (information vs. creativity) and the level of refinement or leveraging of the model’s output (raw output vs. refined output). As the technological landscape continues to evolve at a rapid pace, the integration of AI-driven tools like ChatGPT-4 into various business and management practices has become crucial for staying competitive. Our user archetypes elucidate the varying ways that these tools can be employed to improve productivity, generate novel ideas, and enhance the quality of output in diverse professional domains.    Figure 1: Synergistic Frontiers: The confluence of Human Expertise and AI-Driven Language Models Informed Explorers (Information + Raw Output): These users turn to ChatGPT-4 for quick answers or overviews of various topics. They are not necessarily seeking high-quality final output but rather general knowledge or insights. Informed Explorers might include casual users, students looking for quick explanations, or professionals needing immediate information. In the realm of business and management, Informed Explorers might comprise of managers or executives seeking quick (and reliable) insights to inform decision-making, entrepreneurs researching new market opportunities, or team members addressing immediate queries during meetings. For instance, a project manager might utilize ChatGPT-4 to rapidly acquire an overview of Agile methodologies, or a marketing professional may seek a cursory understanding of a competitor’s latest campaign. A researcher, much as this author of the post, can leverage NLPs to use the potential of AI in quickly seeking insightful and informative content. Insightful Refiners (Information + Refined Output): These users engage with ChatGPT-4 to gather information but then refine and synthesize the AI’s output into polished, high-quality final products. They might be researchers, analysts, or writers creating detailed reports or articles, relying on ChatGPT-4 for initial information and then applying their expertise to elevate the content.For individuals in roles where precision and depth are of paramount importance, the Insightful Refiner archetype becomes quite relevant. These users recognize the potential of advanced and intelligent NLPs as a foundation to build upon, rather than as a source of final products. Market analysts, for example, may employ ChatGPT-4 to gather raw data on industry trends, subsequently refining the information to create comprehensive reports. Similarly, management consultants could utilize ChatGPT-4 to obtain preliminary data on best practices, subsequently crafting tailored recommendations for their clients. Creative Catalysts (Creativity + Raw Output): These users employ ChatGPT-4 as a source of inspiration for creative projects, leveraging the AI’s raw output as a starting point. They might be writers, artists, or designers who use ChatGPT-4 to brainstorm ideas, generate rough drafts, or create initial designs, understanding that further human refinement will be necessary.This archetype can also represent users who use AI-powered tools as a springboard for their creative endeavors. In the context of business and management, a product designer might engage for instance with ChatGPT-4 to generate an array of innovative concepts for a new product line, later refining and curating the ideas through human intervention. Advertising executives could also benefit from the Creative Catalyst archetype, employing ChatGPT-4 to brainstorm potential taglines or campaign slogans before selecting and refining the most compelling option. Artful Architects (Creativity + Refined Output): This archetype is characterized by users who tap into the creative prowess of ChatGPT-4, elevating and refining the AI-generated content to produce polished and tailored final products. In the business world, this might include public relations professionals who use ChatGPT-4 to draft press releases, subsequently editing and optimizing the content to align with their company’s unique voice and messaging. Social media managers may also exemplify this archetype, employing ChatGPT-4 to generate a range of post ideas, then refining and customizing the content to resonate with their target audience and support specific marketing objectives.Artful Architects harness ChatGPT-4’s creative capabilities and then build upon and refine the AI-generated content to create high-quality final output. They might include content creators, marketers, or social media managers who use ChatGPT-4 to draft content and then meticulously polish and optimize it to meet their specific goals and requirements. This conceptual model of Synergistic Frontiers between AI-powered tools and humans, highlights the diverse range of applications for tools like ChatGPT-4 in the business and management sphere. By understanding and capitalizing on these user archetypes, professionals can effectively leverage the capabilities of advanced and intelligent NLPs to optimize their workflows, enhance the quality of their output, and ultimately drive greater success in their respective fields. Our analysis of these user typologies elucidates the potential challenges and opportunities inherent in AI-driven technologies, equipping stakeholders with the knowledge required to make informed decisions regarding the integration of GPT-4 and similar NLP models into organizational processes. By focusing on user types and their respective applications, we foster a more comprehensive understanding of AI’s role in transforming the future of work and management practices. The typology we present provides an in-depth perspective on the versatile applications of ChatGPT-4 and comparable NLP models across various contexts. By delineating distinct user archetypes based on the interaction purpose and output refinement level, our framework offers insights into the ways in which diverse users can engage with and derive benefits from AI-driven language models. This understanding allows organizations to customize their AI deployment strategies to address the specific needs of each user archetype, optimizing value and efficiency. Additionally, the matrix facilitates the exploration of AI adoption implications across industries, organizational functions, and job roles, shedding light on the challenges and opportunities tied to the incorporation of AI-generated content in everyday work and leisure situations. Conclusion In this exploration, we introduce a two-by-two matrix to elucidate the typologies of GPT-4 user types, underscoring the distinctions between information retrieval and creative applications as the two dimensions that define the varying degrees of AI adoption. This analytical framework serves to elucidate how distinct user categories can capitalize on the transformative power of GPT-4 and analogous NLP models. By examining the interaction between these dimensions and user types, we aim to provide valuable insights for businesses and management scholars alike, offering actionable guidance for effectively integrating AI-driven tools into a diverse array of contexts. Furthermore, rather than engaging in a man vs. machine debate, we emphasize the importance of leveraging technology. This vision promotes the idea of fostering a collaborative environment, where human and machine intelligence coalesce to create a future where the strengths of both entities are leveraged to achieve advancements in the field of management practice and research. The collaboration between human and AI intelligence holds the potential to significantly enhance decision-making processes within organizations (Brynjolfsson & McAfee, 2014; Daugherty & Wilson, 2018). AI-generated insights can complement human judgment and intuition by providing data-driven recommendations and identifying patterns that may be difficult for humans to discern (Davenport & Ronanki, 2018). For instance, AI models can help uncover hidden correlations or trends in large datasets, enabling decision-makers to make more informed choices (Chui et al., 2018). Moreover, NLP models like GPT-4 can assist in analyzing vast amounts of textual information, making it easier for humans to process and synthesize valuable insights from diverse sources (Hirschberg & Manning, 2015). On the other hand, human expertise remains crucial in refining and contextualizing the insights provided by AI models (Kaplan & Haenlein, 2019). Decision-makers can draw on their experience, domain knowledge, and understanding of organizational culture to interpret AI-generated findings and make nuanced judgments (Manyika et al., 2017). This collaborative approach can lead to more effective and innovative decisions, as it combines the strengths of both human and machine intelligence (Davenport & Kirby, 2016). By fostering an environment in which AI-driven tools and human expertise work in synergy, organizations can leverage the power of AI to achieve unprecedented advancements in management practice and research, while also ensuring that decisions remain grounded in the human values and context that define their organizational identity (Bostrom & Yudkowsky, 2014; Daugherty & Wilson, 2018). Despite the remarkable capabilities of the AI driven models, it is crucial to recognize their limitations to avoid overreliance and maintain a balanced perspective (Amodei et al., 2016; Bostrom & Yudkowsky, 2014). One of the main challenges is the models’ inability to fully understand context and the nuances of human language, which may lead to the generation of inappropriate or nonsensical content (Bender & Koller, 2020). Additionally, these models can sometimes produce plausible-sounding but incorrect or misleading information, posing potential risks when used in decision-making processes (Radford et al., 2019). Furthermore, ensuring transparency and explainability in AI models remains a challenge, particularly as they become more complex (Arrieta et al., 2020). Acknowledging these limitations is vital to guide future research and development in the field, as well as to inform organizations on how to best utilize AI-driven tools in their operations.   References Amodei, D., Olah, C., Steinhardt, J., Christiano, P., Schulman, J., & Mané, D. (2016). Concrete problems in AI safety. arXiv preprint arXiv:1606.06565. Agrawal, A., Gans, J. S., & Goldfarb, A. (2019). The Economics of Artificial Intelligence: An Agenda. University of Chicago Press. Arrieta, A. B., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., … & Herrera, F. (2020). Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI. Information Fusion, 58, 82-115. Bender, E. M., & Koller, A. (2020). Climbing towards NLU: On meaning, form, and understanding in the age of data. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (pp. 5185-5198). Bessen, J. E. (2019). AI and Jobs: The Role of Demand. NBER Working Paper No. 24235. Bostrom, N., & Yudkowsky, E. (2014). The Ethics of Artificial Intelligence. Cambridge Handbook of Artificial Intelligence, 1, 316-334. Brundage, M., Avin, S., Clark, J., Toner, H., Eckersley, P., Garfinkel, B., … & Anderson, H. (2018). The malicious use of artificial intelligence: Forecasting, prevention, and mitigation. arXiv preprint arXiv:1802.07228. Brynjolfsson, E., & McAfee, A. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. W.W. Norton & Company. Chui, M., Manyika, J., & Miremadi, M. (2018). What AI can and can’t do (yet) for your business. McKinsey Quarterly, 1, 110-123. Clarke, D., Hao, Y., O’Reilly, C., & Bubeck, D. (2021). OpenAI’s GPT-3: A Technical Overview and Its Implications for the Future of AI. AI Matters, 7(1), 3-10. Davenport, T. H., & Kirby, J. (2016). Only Humans Need Apply: Winners and Losers in the Age of Smart Machines. Harper Business. Davenport, T. H., & Ronanki, R. (2018). Artificial intelligence for the real world. Harvard Business Review, 96(1), 108-116. Daugherty, P. R., & Wilson, H. J. (2018). Human + Machine: Reimagining Work in the Age of AI. Harvard Business Press. The Economist. (2023). The Inception of Authentic Artificial Intelligence: Unleashing the Power of GPT-4. The Economist. George, G., Haas, M. R., & Pentland, A. (2014). Big data and management. Academy of Management Journal 57(2), 321-326. Haenlein, M., & Kaplan, A. M. (2019). A brief history of Artificial Intelligence: On the past, present, and future of Artificial Intelligence. California Management Review, 61(4), 5-14. Hirschberg, J., & Manning, C. D. (2015). Advances in natural language processing. Science, 349(6245), 261-266. Kaplan, A., & Haenlein, M. (2019). Siri, Siri, in my hand: Who’s the fairest in the land? On the interpretations, illustrations, and implications of artificial intelligence. Business Horizons, 62(1), 15-25. Manyika, J., Lund, S., Chui, M., Bughin, J., Woetzel, J., Batra, P., … & Sanghvi, S. (2017). Jobs lost, jobs gained: What the future of work will mean for jobs, skills, and wages. McKinsey Global Institute Report. Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., & Sutskever, I. (2019). Language models are unsupervised multitask learners. OpenAI Blog, 1(8), 9.

Introduction On 14 November 2022 the Business Twenty (B20) presented its policy recommendations to the G20 at a summit in Bali, Indonesia. The B20 is the official business voice that communicates with the G20, a strategic multilateral platform connecting the world’s major developed and emerging economies. G20 members represent more than 80 percent of world GDP, 75 percent of international trade and 60 percent of the world population. It was established in 1999, initially to convene finance ministers and central bank governors. It has evolved into a yearly summit involving the heads of state of member countries. The B20 formulates and presents policy recommendations on designated issues to the G20 on an annual basis.  The summit was an auspicious event, with many prominent government and business leaders present. Given his recent acquisition of Twitter, it was to be expected that the presence of Elon Musk as keynote speaker would steal the limelight. In a way, the issues highlighted by the Twitter transaction – privacy, free speech, job security, etc. – provided the perfect backdrop for the recommendations of the B20’s Integrity and Compliance Task Force, which form the focus of this contribution. Musk’s participation turned out to be virtual, and also a bit bizarre because it was via a rather poor phone link and by candlelight, given a power outage in California at the time. The ESG agenda The summit took place in a context where there is an increased focus on environmental, social and governance (ESG) issues. The ESG agenda seems to have overtaken the more traditional corporate social responsibility (CSR) perspective and provides a more holistic approach. It is very difficult to think of any business issue that cannot easily be identified as either an E, S or a G.  This can be viewed as both a strength and weakness. Strong because it is holistic, weak because it can dilute focus. If ESG is everything, it can also become nothing. ESG is driven primarily by the investment community, which also indicates an interesting shift from the moral case to the business case for corporate responsibility. Mainstream investors mostly ignored earlier moral arguments in favour of ESG until the impact of climate change (droughts, floods and fires) and inequality (refugees and political unrest) became tangible business risks. For example, in his 2022 letter to CEOs, Blackrock CEO Larry Fink states that “the tectonic shift towards sustainable investing is still accelerating”, but makes it clear what the driver is: “Make no mistake, the fair pursuit of profit is still what animates markets; and long-term profitability is the measure by which markets will ultimately determine your company’s success”. Over the last few years there has been a renewed interest in the G of ESG. There is a growing realisation that sound governance and corporate integrity should underpin environmental and social responsibility. This is illustrated by the Agenda for Business Integrity developed by the World Economic Forum’s Global Future Council on Transparency and Anti-Corruption. There is a need to move beyond compliance and to apply behavioural techniques to strengthen corporate culture. In addition, technology (specifically frontier technology) and collective action are both required to increase scale and impact. Although not addressed directly by the World Economic Forum’s agenda, corporate reporting is also critical. Over the last few years we have witnessed an almost frenetic consolidation of sustainability reporting standards, and the general consensus seems to be that the future lies in the global baseline proposed by the International Sustainability Standards Board (ISSB). This is also acknowledged in the B20 Communique. The Global Reporting Initiative (GRI), which was seen as the leading standard setter in sustainability reporting for decades, remains an important player but seems to have been left behind in the process to establish the ISSB. The B20 recommendations The B20 has a number of task forces, including the Integrity and Compliance Task Force, comprising representatives from the international business community. There are also network partners, including the Basel Institute on Governance, Business at OECD, International Chamber of Commerce, International Federation of Accountants, the Institute of Internal Auditors and the World Economic Forum. Examples of companies represented on the task force are Deloitte, Google, GSK, Mastercard, Novartis, Siemens and the World Bank Group. While the activities of the task force are not exactly the same from year to year, the typical process is to prepare a document with policy recommendations, developed through a consultative process, that is then submitted for inclusion in the overall B20 document and eventually presented to the G20 as part of the formal B20 Communique. The core recommendations  of the most recent Integrity and Compliance task force focused on sustainable governance to support ESG initiatives, collective action to alleviate integrity risks and measures to address risks associated with money laundering, terrorist financing and cybercrime.  The recommendations are summarized in the table below.   Table 1: B20 Recommendations and Policy Actions Measuring impact by uptake The traditional way to measure impact of the B20 recommendations is to measure uptake by the G20. According to research performed by the Basel Institute of Governance, the results are mixed, with most progress at high level recommendations, but very little at the country level: “One type of recommendation that has seen very little uptake by the G20 are those that call for specific actions and engagement with the private sector either at the G20 or country level”. This is not a surprise. The G20 comprises diverse nations, incorporating developed and emerging economies, and with one member – the European Union – comprising 27 member states. Recommendations at the B20 level are mostly generic and might not always take into account that some countries might have regulations in place already and that for others it might not be a short-term priority, especially because host country bias often results in at least some of the recommendations being aimed at domestic agendas. Finally, uptake in the form of eventual regulation and implementation is very complex to track over time. This is particularly relevant for the European Union, where policy decisions have to be transposed into law by individual member countries. In a nutshell, the traditional  “top-down” approach is to track recommendations from the B20, through the G20 Communique, and then to follow 46 different countries on a multi-year level, further broken down in terms of government departments and relevant industry initiatives. Measuring impact by action We suggest that there is another way in which impact can be achieved and measured. This is not presented as an alternative but rather a complementary approach. This bottom-up approach requires us to view the B20 recommendations as a framework for action, rather than a wish list of policy recommendations. The United Nations Sustainable Development Goals (SDGs) provide an interesting example. The 17 SDGs were adopted by governments as a roadmap to sustainability. The SDGs are accompanied by 169 targets to track country level progress until 2030. However, many companies have aligned their sustainability strategies with the SDGs in order to demonstrate and measure support for the SDGs. Some believe that the achievement of the goals will not be possible without significant backing from the private sector, which includes both companies and investors. In fact, the SDGs also inform the B20 recommendations, with the most important links to Integrity and Compliance being goal 16 (peace, justice and strong institutions). This approach was alluded to by the report of the Basel Institute on Corporate Governance in 2020: “[T]he B20 anti-corruption task force recommendations remain important even if they do not appear in the G20’s work. They provide ideas and inspiration that can also be picked up by others”. The bottom-up approach starts with companies and what they do, and in addition how they report on what they do. We make the assumption that the B20 will only make recommendations that they believe will be in the interest of the global business community, and that this community will therefore be inclined to take voluntary action. It is interesting to note that the long-term trend of the B20 task force has been to move away from recommendations that require hard regulation. The first two recommendations of the 2022 cycle (sustainable governance and collective action) are more closely aligned with voluntary corporate action, to a large extent dependent on responsible corporate governance and ethical leadership. We believe that the B20 recommendations can therefore also be viewed as a framework for action. This is not a standards framework (like the ISO standards) but rather a flexible, non-certifiable framework that will allow companies to introduce materiality filters and select areas where they believe they are either at risk, where they can make the biggest contributions or indeed where there are the most opportunities. And finally, measuring impact requires metrics. The emerging global baseline for sustainability reporting can play a big role. The first standards issued by the ISSB have been in the climate space, but if future metrics for integrity and compliance could be aligned with the B20 recommendations, in addition to other global initiatives, this could become very helpful for all stakeholders. Proposed corporate actions With reference to the core recommendations, the table below provides an initial list of corporate actions that could support the B20 recommendations. Table 2: B20 Recommendations and Proposed Corporate Actions Conclusion Compliance and integrity have always been priorities and a challenge for business to maximize value creation whilst achieving long-term sustainability and legitimacy. Amidst a constantly changing and dynamic environment, embedding and integrating compliance and integrity at the core of business are now more critical than ever for business success. By taking the initiative through innovative voluntary actions, the private sector can make an important contribution beyond its own performance. We believe that the B20 integrity and compliance recommendations provide an important framework for action to support the G in ESG and to move the whole agenda forward.