SELECT ProductName, UnitPrice
FROM Products
WHERE UnitPrice > 50Analysis Report One - What’s Your Data Strategy
Executive Summary
This report looks at how a healthcare organization can use data and artificial intelligence in a way that improves decision-making while still protecting privacy, security, and data quality. The readings for this week show that an effective data strategy must balance data defense, which focuses on protecting and governing information, with data offense, which focuses on using data through analysis, reporting, and visualization to improve performance.
The readings also make it clear that technology alone is not enough. Artificial intelligence can increase productivity and strengthen analysis, but it is becoming widely available and may not provide a lasting competitive advantage by itself. Long-term success will still depend on human creativity, judgment, relationships, and the ability to use technology in meaningful and original ways. The TikTok example further shows how algorithms can create harmful patterns when they operate without enough oversight, while the reading on data poisoning demonstrates how false or manipulated training data can lead to unreliable results, misinformation, and security risks.
For this report, the Northwind database is treated as the operational database of a healthcare supply organization. Products represent medical supplies or equipment, suppliers represent healthcare vendors, and orders represent purchasing activity. SQL queries and visualizations are used to evaluate product activity, supplier relationships, and inventory levels. This analysis shows how trusted and well-organized data can help managers identify operational risks and make more informed decisions.
Based on the research and analysis, the organization should maintain a reliable source of truth, assign clear responsibility for data management, strengthen security and access controls, and carefully evaluate the sources used by AI systems. It should also include human oversight in important automated decisions, monitor inventory and supplier risks, and gradually expand its use of advanced analytics as its data quality and governance practices improve. Overall, the organization should take a defense-led approach while still using data and AI responsibly to improve efficiency, reduce risk, and support better healthcare outcomes.
Introduction
Organizations today have access to more data, technology, and artificial intelligence than ever before. However, simply having these tools does not mean they will automatically lead to better decisions or better results. The readings for this week show that organizations need a clear plan for how data is collected, protected, managed, and used. They also need to consider the ethical and security risks that can come with artificial intelligence, algorithms, and automated decision-making.
In What’s Your Data Strategy?, DalleMule and Davenport explain that organizations need to balance data defense and data offense. Data defense focuses on protecting information through privacy, security, accuracy, compliance, standardization, and strong data governance. Data offense focuses on using data through reports, analysis, and visualizations to improve performance and support better decisions. The authors also explain the importance of having a single source of truth. This means that important information should come from one trusted and reliable source so that everyone in the organization is working with consistent data. Different departments may use that information in different ways, but their reports should still be based on the same accurate foundation.
The article Why AI Will Not Provide Sustainable Competitive Advantage adds to this discussion by explaining that artificial intelligence can improve productivity, support analysis, and make business processes more efficient, but AI alone is not likely to create a lasting competitive advantage. As AI becomes more widely available, competitors will have access to many of the same tools and capabilities. The authors explain that a sustainable advantage must be valuable, unique, and difficult for others to copy. Because AI is becoming easier to access, long-term success will still depend on human creativity, innovation, relationships, passion, and the ability to use technology in original and meaningful ways.
The TikTok reading shows the possible harm that can occur when algorithms are designed to keep users engaged without enough attention being given to the effects of the content they recommend. In the study discussed in the article, accounts that appeared to belong to young teenagers were quickly shown repeated content related to eating disorders, body image, self-harm, and suicide. This shows that algorithms do more than organize information. They can influence what people see, reinforce certain behaviors, and create harmful patterns. It also shows why organizations need human oversight, transparency, accountability, and stronger protections, especially when technology affects children or other vulnerable groups.
The reading about data poisoning introduces another major concern involving the information used to train generative AI systems. Since many AI models are trained using large amounts of data collected from the public internet, false, misleading, or harmful information can sometimes become part of the training data. Malicious actors may even intentionally add incorrect data or hidden instructions in an attempt to influence an AI system’s output. This could cause the system to spread misinformation, produce unreliable results, or expose sensitive information. This reading shows why organizations need to know where their data comes from, monitor its quality, protect access to it, and carefully evaluate AI systems before connecting them to private or confidential information.
Together, the readings show that a strong data strategy involves much more than collecting large amounts of information or adopting the newest technology. Organizations need to protect the privacy, security, and quality of their data while also using reliable information to improve performance and decision-making. They must understand the benefits of AI without assuming that it is always accurate, safe, ethical, or enough by itself to create a lasting advantage. These concerns are especially important in healthcare, where inaccurate information, weak security, harmful recommendations, or manipulated data could affect patient privacy, safety, supply availability, and quality of care. This report applies these ideas to the Northwind database by treating it as the operational database of a healthcare supply organization and using SQL queries and visualizations to examine products, suppliers, inventory, and purchasing activity.
Research and Extensions
The assigned readings show that data and artificial intelligence can create value, but recent research makes it clear that organizations also need stronger plans for managing the risks that come with these tools. A modern data strategy now involves more than protecting information stored in a database. It must also address how AI systems are selected, what information they are trained on, how their results are reviewed, and who is responsible for monitoring them after they are placed into use. This is especially important in healthcare because unreliable information or an incorrect automated recommendation could affect patient privacy, safety, and quality of care.
Recent research supports the idea that AI can improve performance without necessarily giving an organization a lasting competitive advantage. Brynjolfsson, Li, and Raymond studied the introduction of a generative AI assistant among 5,172 customer-service employees. They found that access to the AI tool increased productivity, measured by issues resolved per hour, by an average of 15%. The results were not the same for every employee, showing that the value of AI can depend on a worker’s experience, skills, and ability to use the system effectively. This supports the argument from Why AI Will Not Provide Sustainable Competitive Advantage. AI may be valuable, but it is becoming widely available and can be copied by competitors. A stronger advantage will come from how an organization combines the technology with employee knowledge, creativity, training, and relationships.(brynjolfsson2025generative?)
Recent research also expands the idea of data governance into the area of responsible AI governance. Papagiannidis, Mikalef, and Conboy developed a framework that organizes responsible AI governance into structural, relational, and procedural practices. Structural practices include clear policies, assigned responsibilities, and formal oversight. Relational practices involve communication and cooperation among managers, technology professionals, employees, and other stakeholders. Procedural practices include testing, documenting, reviewing, and monitoring AI systems. For a healthcare organization, this means an AI system should not be approved once and then allowed to operate without further review. The organization should identify who is responsible for the system, document the data being used, and continue checking its accuracy, fairness, privacy, and security.(papagiannidis2025responsible?)
New research also adds to the concerns raised in the TikTok reading. Fassi and colleagues studied social media use among 3,340 adolescents between the ages of 11 and 19, including participants with and without diagnosed mental health conditions. Adolescents with mental health conditions reported spending more time on social media and being less satisfied with the number of online friends they had. Those with internalizing conditions, such as anxiety and depression, also reported more social comparison and a greater effect of online feedback on their moods. The study did not establish that social media caused these conditions, but it showed that different groups can experience the same technology differently.(fassi2025social?)
This research extends the TikTok example because an algorithm should not be judged only by its overall engagement numbers or average results. A recommendation system may appear successful because users spend more time on the platform, even when the content is more harmful to certain users. Organizations should test automated systems across different groups and consider how recommendations affect people who may be more vulnerable. In healthcare, the same concern would apply to systems used to identify high-risk patients, recommend treatments, schedule appointments, or determine which patients receive additional attention.(fassi2025social?)
Data poisoning is another area where recent research shows that the risks have become more serious. Alber and colleagues tested whether medical large language models could be affected by false medical information placed into their training data. They found that replacing only 0.001% of the training tokens with medical misinformation made the models more likely to produce harmful medical errors. The poisoned models still performed similarly to the original models on common medical benchmarks, meaning that regular testing might not reveal the problem. The researchers developed a screening method using biomedical knowledge graphs that identified 91.9% of the harmful content.(alber2025medical?)
This finding is especially important for healthcare because it shows that an AI system can appear accurate while still containing dangerous information. Healthcare organizations should carefully review where AI training data comes from and should not automatically treat information collected from the public internet as reliable. Medical information can be outdated, misleading, or intentionally manipulated. AI-generated information should be checked against trusted clinical sources and reviewed by qualified professionals before it is used to make decisions involving patient care.(alber2025medical?)
These research findings also connect directly to the Northwind analysis in this report. If Northwind represented a healthcare supply organization, accurate information about products, suppliers, prices, inventory, and reorder levels would be necessary for both traditional reporting and future AI systems. Incorrect inventory counts could lead to supply shortages, duplicate supplier records could hide vendor dependence, and inaccurate prices could affect budgeting. An AI system would not automatically correct poor-quality data. It could repeat the errors or make them more difficult to identify.(alber2025medical?)
Overall, the additional research shows that healthcare organizations should take a careful but practical approach to data and artificial intelligence. The organization should maintain a reliable source of truth, assign clear responsibility for data and AI systems, review where information comes from, test results across different groups, and continue monitoring systems after implementation. AI can improve employee productivity and support decision-making, but it should assist rather than replace human knowledge and judgment. The organization’s strongest advantage will come from combining reliable data, responsible technology, trained employees, and effective leadership. (papagiannidis2025responsible?; alber2025medical?)
Data Visualizations
For these visualizations, I used the Products table from the Northwind database. I kept both examples simple by using the SELECT, FROM, and WHERE statements we learned this week.
Visualization One – Offensive This query shows the names and prices of products that cost more than $50. The graph makes the prices easier to compare. This is an offensive example because the company can use the information to help with pricing and purchasing decisions.
Visualization Two – Defensive This query shows products that have fewer than 20 units in stock. The graph makes it easier to see which products are running low. This is a defensive example because the company can use the information to help prevent inventory shortages.
Visualization One - Offensive
#ggplot visualization in this block.
ggplot(data = myquery1,
aes(x = ProductName, y = UnitPrice)) +
geom_col() +
coord_flip()
Visualization Two - Defensive
SELECT ProductName, UnitsInStock
FROM Products
WHERE UnitsInStock < 20#Put your ggplot visualization in this block ```ggplot(data = myquery2, aes(x = ProductName, y = UnitsInStock)) + geom_col() + coord_flip()
Recommendations for Industry
Based on the research and data findings, organizations should focus on creating a data strategy that improves daily operations while also protecting sensitive information. One important step is to use a centralized data management system so that employees and leaders have access to accurate and up-to-date information. Companies should also establish clear rules for how data is collected, stored, accessed, and maintained. This can help reduce errors, duplicate information, and inconsistencies across departments.
Organizations should also strengthen their cybersecurity practices by using multi-factor authentication, limiting access to sensitive information, monitoring systems for unusual activity, and regularly training employees. Technology alone cannot prevent every data breach, so employees need to understand how to recognize phishing attempts and other security risks.
Artificial intelligence and predictive analytics can also help organizations identify inefficiencies, predict future demand, detect unusual activity, and make better decisions. However, companies should not collect data simply because they can. They should focus on collecting useful information that can improve customer service, lower costs, and make operations more efficient.
Finally, executives should treat data as an important business resource rather than something that only concerns the IT department. Business leaders, employees, compliance teams, and technology professionals should work together to make sure data decisions support the organization’s goals. By combining strong data management, useful analytics, employee training, and cybersecurity protections, organizations can make better decisions and remain competitive over time.
References
(article?){alber2025medical, title={Medical large language models are vulnerable to data-poisoning attacks}, author={Alber, Daniel Alexander and Yang, Zihao and Alyakin, Anton and Yang, Eunice and Rai, Sumedha and Valliani, Aly A and Zhang, Jeff and Rosenbaum, Gabriel R and Amend-Thomas, Ashley K and Kurland, David B and others}, journal={Nature Medicine}, volume={31}, number={2}, pages={618–626}, year={2025}, publisher={Nature Publishing Group US New York} } (article?){brynjolfsson2025generative, title={Generative AI at work}, author={Brynjolfsson, Erik and Li, Danielle and Raymond, Lindsey}, journal={The Quarterly Journal of Economics}, volume={140}, number={2}, pages={889–942}, year={2025}, publisher={Oxford University Press} } (article?){fassi2025social, title={Social media use in adolescents with and without mental health conditions}, author={Fassi, Luisa and Ferguson, Amanda M and Przybylski, Andrew K and Ford, Tamsin J and Orben, Amy}, journal={Nature human behaviour}, volume={9}, number={6}, pages={1283–1299}, year={2025}, publisher={Nature Publishing Group UK London} } (article?){papagiannidis2025responsible, title={Responsible artificial intelligence governance: A review and research framework}, author={Papagiannidis, Emmanouil and Mikalef, Patrick and Conboy, Kieran}, journal={The Journal of Strategic Information Systems}, volume={34}, number={2}, pages={101885}, year={2025}, publisher={Elsevier} }