Make Digitalk Crispy Again
Most of us are aware that Artificial Intelligence (AI) is becoming a more prevalent aspect of our daily lives. There are AI that can beat the world chess champion, others that control automobile automatic driving systems or some that provide personalized health guidance. But many of us would be quite surprised to learn that AI will soon be implemented in potato chips manufacturing. Indeed, AI, in particular its machine learning applications, has caught the interest of both large and small industrial firms.
According to the Wall Street Journal (WSJ), “The companies have long relied on research-and-development practices that took years to yield new products—but now consumer tastes are changing faster than ever, as people seek new flavors and find a widening array of specialty foods online. Brands can doom themselves by not adjusting quickly to new trends or changing preferences.”(AI flavors for Frito-Lay’s, 2021a, para 1)
But if you’ve been staying up to date with machine learning developments, you’ve probably heard that, as advanced as the technology has become, it’s still a long way from being “ready for prime time” in industry (How Frito-Lay Applies Machine Learning, 2019). However, it has the potential to become a vital component of food production and delivery. With this blog entry, let us engage in a thought experiment: Imagine that we, as Digitalk, are a consultancy business on the matter of AI. Recently, the American chips ‘company Lay’s called upon our consulting services to implement AI in their mass production. Let’s make Digitalk crispy again!
A Tasty History
Frito-Lay is an American PepsiCo subsidiary that produces, promotes, and sells corn chips, potato chips, and other snack foods worldwide. Frito-Lay began as two independent companies, “The Frito Company” and “H.W. Lay & Company,” in the early 1930s, until merging in 1961 to establish “Frito-Lay, Inc” (Newton & Miller, 2009). In the 1970s, Frito-Lay faced greater competition from rival potato chip brands such as Pringles. Today, the company is the world leader in the salty snack category, hiring more than 92,000 employees and controlling more than 35% of the global snap chip market (Frito-Lay Company, 2019). To keep up with the trends, Frito-Lay’s want to change the way they create and market their products. Since food industries are directly affected by technological developments, AI can have a significant impact on food production, improving the manufacturing process and enhancing customer satisfaction. Using AI wisely could help the brand to enhance their understanding of the value of machine learning and its application of new technology. As a fictional consultancy firm, Digitalk proposes several ways where technology could assist innovation in snacking:
- Augmenting human intuition by assessing “store potential” and delivering information and projections to local sales representatives to assure full shelves week after week.
- Measuring trends, detect online activity, monitor in-store activity, and watch social media to seek “flavor inspiration” — giving an insight into what tastes or items consumers want, as well as revealing seasonal and weather-related increases in demand (Larson & Teichman, 2009).
- Decreasing development time in half by tracking consumer behavior and speeding up the testing processes by up to 50%. Without the use of time-consuming and expensive evaluations, top sellers can be forecasted based on historical examination of flavor and texture characteristics.
- Improving time-to-shelf via graphical representations that reveal specific demand in local areas enabling Frito-Lay’s to shift production to facilities closer to the consumers — resulting in providing clients with more fresh snacks while reducing environmental and community impact
The Taste of AI
AI is usually not the first thing that comes immediately to mind when discussing the food industry. However, AI’s ability to drive higher efficiency and profits, decrease wastage, and protect against supply chain interruptions is being recognized by an increasing number of companies in the food industry. One aspect of artificial intelligence that gained attention from several companies is cognitive technology. Cognitive technology refers to a wide range of technologies that aid, augment, or simulate cognitive processes or can be utilized to achieve cognitive goals. Considered as “the next-generation artificial intelligence”, it has the potential to enhance technological innovation and significantly benefit companies (Ienca, 2018, para 1). According to Davenport and Ronanki (2019), companies should consider AI in terms of business capabilities rather than technological skills. In their study, the authors explore the different types of AI that are being used and present a framework for how firms should start building up their cognitive skills in the coming years to meet their business goals (3 Things AI Can Already Do for Your Company, 2019). Three critical business challenges can be addressed by AI: Automating business processes (process automation), gaining insight through data analysis (cognitive insight), and engaging with customers and employees (cognitive engagement). In an effort to provide Frito-Lay’s the best possible recommendation, Digitalk discusses the three types of AI inspecting which one would best suit the food company’s requirements.
(1) Process Automation Robotic
Process automation (RPA) is a technology for performing high-volume, repeating activities that replicates how humans interact with software. It is becoming more popular as a way to reduce spending, streamline processes, and improve user experiences. RPA is designed to mimic how people interact with and think about digital systems (Lawton, 2021). RPA can benefit several aspects of a company’s services such as finance and accounting (invoice processing), human resources (hiring and onboarding) or retail (inventory management). RPA is the best value and easiest AI to implement, with a rapid and significant return on investment.
According to David Landreman (CPO of Olive, an Information Technology and AI Services company), “RPA is the process by which a software bot uses a combination of automation, computer vision, and machine learning to automate repetitive, high-volume tasks that are rule-based and trigger-driven.”( Robotic Process Automation (RPA), 2020, para. 4)
Many food companies started to implement robotic process automation as a major contributor to improve operational efficiency. There is no large-scale interruption associated with a technology rollout because robotic process automation duplicates a worker’s inputting behavior, utilizing the same software and user interfaces. As an example, Coca-Cola United recently streamlined its order management using robotic process automation in Microsoft Power automate. They partnered with the IT transformation company Happiest Mind Technologies to generate an automated service agent performing routine and rules-based tasks such as getting customer data from many corporate systems, assessing a software’s accuracy, or processing an insurance claim (BizClik Media, 2021). It helped them change their labor equation: By improving their customer experience, they enhanced their ability to meet different requirements of service level accords.
(2) Cognitive Insights
Cognitive insights makes use of AI algorithms to gather massive amounts of data, analyze its significance, and spot patterns in the data. Companies in highly competitive industries have recourse to cognitive insights to predict customer buying habits and trends aiming to personalize commercials and marketing strategies by automating them. Furthermore, the use of cognitive insight accelerates the hiring process by automatizing job researches and onboarding (Content, 2020). Cognitive insights are data-intensive and complex, and the models are often trained on a subset of the data set. Over time, the models improve in their capacity to use newer data to make predictions by classifying entities. Machine learning can also make new data available for improved analytics. While data curation has traditionally been a time-consuming process, machine learning can now discover probabilistic matches across databases—material that is likely to be connected with the same person or firm but appears in slightly different formats (Cognitive Insights Work Where Human Insight Is Constrained, 2018).
Recently, the BMC Helix Platform had recourse to IBM Watson Discovery to gain advantage of cognitive insights facilities. It resulted in enhancing the efficiency, productivity and innovation of the open platform. Indeed, implementing cognitive insights in multiple chatbots of their service increased the customer’s experience and broader their analytics ‘data (BMC Documentation, 2021).
(3) Cognitive Engagement
In the context of AI, cognitive management is considered as the technology to communicate more with coworkers rather than customers. Customer data is input into a cognitive engagement strategy. Predictive analytics uses this information to provide predictions about future customer behavior. Cognitive engagement involves internal websites that provide answers to employee questions on IT, employee benefits, and HR policy (Mills, 2018). Food and beverages companies started to use cognitive management for personalization. Starbucks recently laid the foundation for the integration of digital marketing and physical retail. Since Customers must be analyzed and comprehended in order to be cognitively engaged, Starbucks will now make serving recommendations to consumers approaching their outlets based on location data. Individual preferences and expectations must be considered while developing engagement tactics. The American company also implemented chatbots in their services to provide customers with quality on-time support (Mills, 2018). However, despite the great innovation that cognitive engagement represents, companies tend to be cautious when it comes to customer-facing cognitive interaction technology because of its newness and its formative stages (Davenport & Ronanki, 2019).
Step by Chip
Considering Frito Lay’s desire to change the way they create and market their products, process automation and cognitive insight represent a great prospect in leveraging artificial intelligence to unleash innovation. Out of the three different types of artificial intelligence, process automation is the one that best suits Frito Lay’s expectation in implementing AI in their manufacturing process. However, executing artificial intelligence in a company is not a one-day process. Companies must first understand which technologies execute specific types of activities, as well as their strengths and limits, before embarking on an AI program. According to the study of Davenport & Ronanki (2019), Digitalk will advise Frito-Lay’s with the fourth step they need to follow regarding the establishment of AI in their process.
“It’s not like there’s a start and stop to this transformation. It’s a process.”Kevin Buehler – Senior Director, Snacks to You for Frito-Lay North America (IBM, 2021b, para 3)
(1) Understanding the Technologies
Frito Lay’s must first understand which technologies execute specific types of activities, as well as their strengths and limits, before embarking on its digital transformation. Process automation and cognitive insight are the two most interesting AIs regarding their desideratum. Considering the shortage of cognitive technology expertise, Frito-Lay’s should aim to create an organization’s resource in a centralized department such as IT and make experts available to high-priority projects across the company (Davenport & Ronanki, 2019).
(2) Creating a Portfolio of Projects
Creating a portfolio of projects will then enable the food company to review in detail the evaluation of needs and capabilities in establishing an AI program. Frito-Lay’s expressed their desire to facilitate the several aspects of the manufacturing process. Brainstorming ideas could lead them to seek and identify opportunities better. Indeed, using RPA and cognitive insight in their process could lead them to many advantages such as: streamline the hiring process, better leverage the massive amounts of data collected, or even implement machine learning and robotization in mass-production.
(3) Determining the Use Cases
The third step consists of examining the scenarios in which cognitive applications could add significant value and help a company succeed (Davenport & Ronanki, 2019). Determining where it is primordial to implement AI and which tasks need to be automatized is of crucial importance. By asking themselves questions such as: Would the benefits of launching Ais be worth the time and effort? How could we optimize efficiency and provide better services? – Frito Lay’s will give greater importance to the sectors where AI could be implemented.
(4) Selecting the Technology
Lastly, the food company will have to examine whether AI tools are substantially effective to the tasks assigned for each use case. Frito-Lay’s manufacturing plant could benefit from machine learning, and indeed, it is already happening. One project utilizes lasers to sharpen chips, then listens to the sounds they make to determine if the texture is adequate or not. To automate the quality assessment for Frito-chip Lay’s processing facilities, algorithms evaluate sound and identify chip’s quality (How Frito-Lay Applies Machine Learning, 2019). However, they need to be cautious about certain pitfalls of AI: Technology such as RPA which might speed up procedures like invoicing, may actually slow down more complicated manufacturing systems. While deep learning visual recognition systems can distinguish images in photos and videos, they require a large amount of labeled data and may struggle to understand a complex visual field (Davenport & Ronanki, 2019). Frito-Lay’s will have to weigh the benefits and risks of each Ais before starting to incorporate them in their process.
A Strong Recommendation for Weak AI
The technology that we advise Frito-Lay’s to implement is called weak AI since it only requires a computer system to carry out activities that have previously been undertaken by humans (Competing in the Age of AI, 2021). Nevertheless, despite its name, the AI factory can already make a wide variety of key decisions using weak AI. It may function and manage information and analyze data for firms in some situations. In some others, it may direct how a corporation creates, transports, or operates tangible things. As an example, Pep Worx, PepsiCo’s cloud-based data and analytics platform, assists the corporation in advising retail locations on which items to stock, where to position them, and what promotions to utilize. In an effort to rebrand its staple product Quaker Oats, the firm was able to pick 24 million families from a dataset of 110 million in the United States that it thought would be ideal to sell the product to. PepsiCo then identified the shopping areas that these households were more likely to frequent and devised campaigns to appeal to them. In the first 12 months following introduction, this data-driven approach and emphasis on a very particular market helped drive 80 percent of the product’s sales growth (Marr, 2019).
In their implementation of AI, Frito-Lay’s should consider that scale, scope, and learning have all become important factors in a company’s success. However, in today’s world, AI-driven processes may be scaled up far faster than traditional scaling methods. It also expands the scope by linking to other digital businesses and provides significant learning and improvement tools, such as extrapolating customer behavior models and tailoring services accordingly (Competing in the Age of AI, 2021). When compared to traditional processes, scale is limited since it ultimately approaches a point of diminishing returns, but AI-driven processes can scale to stratospheric heights.
Collision occurs when a traditional firm competes with an AI-driven firm that serves the consumer. Such collisions are generated by a whole different type of organization that fundamentally alters industries and reshapes the structure of competitive advantage, rather than by a specific technology. It should be highlighted that, in comparison to traditional operating models, AI-driven operational models will take longer to deliver economic value. This is due to the algorithm’s cold start problem, which requires it to gather sufficient data at first.
Traditional network effects, such as connecting businesses, collecting data flows, extracting value, and AI, can reinforce and increase each other’s benefits. However, specialization is no longer as crucial as the ability to acquire data, process it, conduct analytics, and design algorithms, thanks to machine learning. As a result, industry expertise is becoming less important (Competing in the Age of AI, 2021). Because of their capability of getting data, processing it, conducting analysis, and constructing algorithms, strategies are increasingly focusing on the benefits of employing cognitive insights and RPA in their process, such as PepsiCo’s snack delivery robot. These self-driving robots, nicknamed Snackbot, are a collaboration between Robby Technologies and PepsiCo delivering fresh snacks in universities – a service that the authors of this blog, college students themselves can greatly appreciate. Students can order snacks through the Snackbot app, and the robot will deliver them to more than 50 locations across campus for no extra charge. The bots have a 20-mile range on a single charge, and with inbuilt headlamps and all-wheel drive, they can maneuver at night, in the rain, and up curbs. Snackbot is a solution to the needs of time-strapped college students, as well as their preferences, as revealed by PepsiCo’s study (Marr, 2019). In the same way, Frito-Lay’s should reflect on assessing machine learning to facilitate their mass-production process.
The Burnt Side of the Chips, or the Dark Side of AI
Do you think that we could, one day, be woken up by a revolution of walking humanoid potato chips, claiming that the processing AI in charge of their production is about to take control of the food industry? This, or a similar dystopian scenario comes to the mind of many AI skeptics. As the reader knows from the above analysis, implementing AI in a company can offer advantages, but Frito-Lay’s should also take into consideration the potential danger of it. The term dark AI refers to a malicious autonomous system that can execute given the correct inputs. For example, AI contributes to the widening of the social divide. However, AI that manipulates human purchasing, selling, and transaction patterns, as well as deep fakes, already exist (Minevich, 2020). Face surveillance, for example, can be employed as a compromising deep fake image by using facial recognition that gives individual characteristics. Spying conduits could be opened up by smart home devices and IT technology. There are several ways to combat dark AI: putting a greater emphasis on explainable artificial intelligence, for example. Engineers responsible for Frito-Lay’s digital transformation must be able to explain/take responsibility for how their creations behave in order to avoid unintentional attacks. A self-contained system should be able to explain its motivations and behaviors to avoid dark AI to gain control or utilize its power for malicious purposes.
Frito-Lay and IBM collaborated to develop a Salesforce-based solution and implemented AI in their process. Snacks to You is a cutting-edge e-commerce solution that automates front-line staff delivery routes and gives drivers and supervisors with a useful mobile app to boost productivity and visibility. Customers can use the e-commerce platform to streamline their ordering and delivery processes while also expanding their product options. The app is fully connected with Frito- Lay’s snacking insight AI engine, which means it can offer ordering suggestions based on seasonal tastes using data-driven insights. Aspirational customer experience visions ran throughout the entire operation. The technology can also detect when a retailer’s stock is running short and suggest customized assortments. Historical data, predictive analytics, and even a snack score that indicates how likely buyers are to adore a product are all displayed on a dashboard. The creative team would use this backlog of interactions to come up with solutions to the problem, unrestricted in their hunt for the perfect technology for such a job. The transformation teams would use scaled agile principles to quickly design, evaluate, and launch the solution that best fulfilled customers’ needs once a workable minimal viable product (MVP) had been validated. The end product was a stunning user experience with a simple architecture (IBM, 2021).
Time to crush the crispy chips
The different types of AI are directly linked to the first blog entry: IT doesn’t matter, can artificial intelligence be proprietary technology? What about the products of AI? Codification of data is important for streamlining AI learning resources. Frito-Lay’s should consider having a great knowledge management in order to implement AI correctly in their manufacturing process. The main concern regarding cognitive technologies is that they will obliterate large numbers of jobs. Indeed, industry leaders focus on data research and management, not expertise specific skills. Tasks are completed by cognitive systems, not entire jobs. Human employment losses have mostly been caused by attrition of workers who have not been replaced or by automation of outsourced jobs. However, AI represents other risks, companies need to be careful of data management, try to prevent and mitigate breaches damage as much as possible concerning the dark side of IT. As regards the three articles discussed in this blog entry, we think that the first one has too much emphasis on accountability, ethics, and other vague principles with less concretes efforts such as implementation. To enable Frito-Lay’s to gain a deeper insight of introducing AI in their process, it would have been interesting to further discussed the implementation aspect putting less prominence on accountability. Concerning others articles, there should be a logical sequences of how a regular AI can invertedly turn into Dark AI and its threshold. Transboundary Dark AI Side Effects should be further discussed for a better understanding of its risk.
- Companies should consider AI in terms of business capabilities rather than technological skills.
- Despite their growing experience with cognitive tools, businesses confront major development and implementation challenges.
- Information-intensive fields including marketing, health care, financial services, education, and professional services could become both more valuable and less expensive to society when AI is applied.
- Collision occurs when a traditional firm competes with an AI-driven firm that serves the consumer. Such collisions are generated by a whole different type of organization that fundamentally alters industries and reshapes the structure of competitive advantage, rather than by a specific technology.
- Understanding algorithms, constructing data pipelines, experimenting with it, and other challenges for traditional organizations are not the only ones stated above. The difficulties also need reorganizing the company’s structure and operational model.
- Digital enterprises can also be vulnerable to cyber-attacks, or algorithms that, if left uncontrolled, can lead to bias and misinformation.
- Overconfidence might result in a perilous situation. This must be avoided, limiting AI decision-making capabilities, and making it a legal need for humans to have the last word in the process of finding results.
Frito-Lay’s meets Digitalk
This blog began with a podcast on how the UCM writing center can rebrand itself to attract more attention from UCM students. Now that we are in the end of our digital journey, Digitalk introduces you to the final podcast on Lay’s. The following audio excerpt is a fictional, yet not unrealistic situation: The FRito-Lay’s company wants to rebrand itself to implement AI in their mass-production to keep up with the trends and customers’ expectations. To do so, they contact Digitalk® to inquire about the most suitable type of AI and the best strategy to achieve their goals. What does the Digitalk® consultant suggest? Hear for yourself:
3 Things AI Can Already Do for Your Company. (2019, November 27). Harvard Business Review. https://hbr.org/2018/01/artificial-intelligence-for-the-real-world
BizClik Media. (2021, May 20). Coca-Cola Bottling Company Streamlines Operations with RPA. Manufacturing Global. https://manufacturingglobal.com/ai-and-automation/coca-cola-bottling-company-streamlines-operations-rpa
GrayMatter (2021, March 8). AI Flavors for Frito-Lay, Paper Industry Goes Digital, OT Cybersecurity.https://graymattersystems.com/ai-flavors-frito-lay-paper-industry-ot-cybersecurity/
Cognitive Insights Work Where Human Insight is Constrained. (2018, October 12). Chrobis. https://www.chrobis.co.uk/cognitive-technologies/cognitive-insight
Competing in the Age of AI. (2021, January 21). Harvard Business Review. https://hbr.org/2020/01/competing-in-the-age-of-ai
Content, P. (2020, October 27). Artificial Intelligence (AI) Automation & Cognitive Insight Breaking The Business Mold. The Tech Portal. https://thetechportal.com/artificial-intelligence-ai-automation-cognitive-insight-breaking-the-business-mold/
Frito-Lay Company. (2019). History of Frito-Lay Company – FundingUniverse. http://www.fundinguniverse.com/company-histories/frito-lay-company-history/
How Frito-Lay Applies Machine Learning. (2019, March 25). Automation World. https://www.automationworld.com/products/data/blog/13319607/how-fritolay-applies-machine-learning
How to explain Robotic Process Automation (RPA) in plain English. (2020). The Enterprisers Project. https://enterprisersproject.com/article/2019/5/rpa-robotic-process-automation-how-explain
I. (2021b). How a snack empire stays fresh. IBM. https://www.ibm.com/case-studies/frito-lay-ibm/
Ienca, M. (2018). Democratizing cognitive technology: a proactive approach. Ethics and Information Technology, 21(4), 267–280. https://doi.org/10.1007/s10676-018-9453-9
Kumar, I., Rawat, J., Mohd, N., & Husain, S. (2021). Opportunities of Artificial Intelligence and machine learning in the food industry. Journal of Food Quality, 2021, 1–10. https://doi.org/10.1155/2021/4535567
Larson, A., & Teichman, W. (2009). Frito-Lay North America: The Making of a Net Zero Snack Chip. SSRN Electronic Journal. Published. https://doi.org/10.2139/ssrn.1417218
Lawton, G. (2021, April 23). robotic process automation (RPA). SearchCIO. https://searchcio.techtarget.com/definition/RPA
Leveraging cognitive insights in your application – Documentation for BMC Helix Platform 20.02 – BMC Documentation. (2021). BMX Helix Platform. https://docs.bmc.com/docs/helixplatformprevious/leveraging-cognitive-insights-in-your-application-888488930.html
Marr, B. (2019, April 8). The Fascinating Ways PepsiCo Uses Artificial Intelligence And Machine Learning To Deliver Success. Forbes. https://www.forbes.com/sites/bernardmarr/2019/04/05/the-fascinating-ways-pepsico-uses-artificial-intelligence-and-machine-learning-to-deliver-success/?sh=124fd282311e
Mills, T. (2018, February 1). Enterprise AI Is About Cognitive Engagement. Forbes. https://www.forbes.com/sites/forbestechcouncil/2018/02/01/enterprise-ai-is-about-cognitive-engagement/?sh=6411ea08513d
Minevich, M. (2020, March 1). How To Combat The Dark Side Of AI. Forbes. https://www.forbes.com/sites/markminevich/2020/02/28/how-to-combat-the-dark-side-of-ai/?sh=527abc82174b
Newton, D., & Miller, A. (2009). Frito-Lay, Inc. SSRN Electronic Journal. Published. https://doi.org/10.2139/ssrn.1419843
Word count: 3780