Data and AI for business growth: Overcoming challenges, and navigating ethical frontiers

Manjeet Dahiya, VP and head, machine learning and data science, Ecom Express speaks to Dataquest on the critical importance of data and AI-driven decision making in achieving business growth objectives. He emphasizes how AI technologies directly impact revenue growth and cost reduction. However, organizations aiming to leverage data and AI face common challenges and risks. When selecting, implementing, and adopting AI technologies, he suggests considering clear target metrics that directly impact revenue or cost.

Read on to know more:

DQ: How critical is “Data and AI” based decision making when it  comes to business growth objectives? And how can they be effectively leveraged?

Manjeet Dahiya: I strongly believe that the integration of AI and data is crucial for achieving our business growth objectives. Our primary goals revolve around increasing revenue and reducing costs, and I have witnessed firsthand how AI technologies have a direct impact on these objectives. For instance, in the realm of logistics, where our business entails moving packages from one location to another, we encounter a complex network of facilities. By reducing the travel time of each package by 20%, we can effectively reduce costs. To achieve this, we have implemented various technologies that optimize the placement and connectivity of our facilities. AI enables us to determine which first mile facility should pick up a specific package and, likewise, which last mile facility the package should be delivered to. These decisions are made swiftly and efficiently by AI, resulting in a direct impact on our key metrics.

DQ: What are the common challenges and risks that organizations face when trying to leverage data and AI to further growth of the organization?

Manjeet Dahiya: From an external perspective, logistics may appear to be a straightforward operational game. However, upon delving deeper, one realizes that most of the challenging AI problems lie within logistics. I personally gained valuable experience in this field immediately after completing my PhD while working with Delhivery. Since then, I have transitioned across multiple organizations and domains. Throughout my diverse experiences, I came to the realization that logistics is an area where AI can make a significant impact. Thus, when the opportunity arose to return to the logistics industry, I eagerly seized it.

Speaking of the challenges faced by organizations in this field, the foremost obstacle is the availability and quality of usable data. Often, data within organizations is disorganized, scattered across various sources, and lacks the necessary quality for utilization in AI models. This particular challenge is pervasive in many organizations. Another hurdle lies in formulating the right problem to tackle. This requires a belief in new thought processes, considering various factors, and identifying the problem that will create the most impact. Selecting a problem that does not directly align with objectives can lead to wasted efforts. Therefore, formulating the appropriate problem itself is a significant challenge.

Once the problem is defined, the next challenge arises—determining whether the right solutions have been identified. Many organizations tend to rely on rule-based techniques, while machine learning-based techniques may offer easier and superior outcomes. This challenge is tied to the necessity of having the right talent within the industry. Given the rapid progress in AI, staying up to date with new technologies becomes challenging. Acquiring and retaining skilled professionals thus becomes another obstacle in this field.

DQ: What challenges do organizations encounter in establishing/operationalizing effective governance framework?

Manjeet Dahiya: I completely agree with your point regarding the significant challenge posed by data. The underlying problem behind the inadequate state of data governance lies in the absence of dedicated means for data management. This is a major issue I have observed in various organizations. However, data governance is undeniably crucial for any organization, especially when planning to implement AI initiatives. It is essential to invest in communities or teams solely focused on managing data, ensuring its accessibility, usability, and maintaining its quality. Once proper data governance is in place, the organization can begin reaping the benefits. Data governance provides security, ensuring that compliance requirements are met when handling customer data. Additionally, finding the right talent for data engineering poses another challenge, given the fast-paced nature of evolving technologies. Determining the appropriate tools and systems to handle the scale of data adds further complexity to the equation.

DQ: What are the key considerations for organizations in selecting, implementing and adopting AI technologies in business decision making?

Manjeet Dahiya: Yes, one of the most crucial considerations is to have a clear understanding of the target metrics. These metrics should directly impact revenue or cost, serving as indicators. It is essential to have a precise understanding of these metrics. Once the metrics are established, modules can be developed to influence them, and the right problems can be identified to drive significant impact. Before deploying any changes, it is vital to have proper telemetry in place. This enables measurement and tracking of the impact when implementing a model. Telemetry is crucial as it ensures that changes are measured, and their effects, both positive and negative, are understood. Launching a model without telemetry is akin to operating in the dark, without visibility into its impact or potential performance degradation.

Another crucial consideration is stakeholder management. It is important to recognize that AI technologies are inherently probabilistic. They may occasionally backfire or produce unexpected results. Therefore, it is essential to communicate clearly with stakeholders, making them aware that these technologies carry a certain level of probability. Building a platform that can tolerate and handle such situations is crucial. It allows for managing and mitigating risks when things do not go as planned.

Regarding the emergence of generative AI in the market, there is indeed a renewed focus on the probability aspect. The unknown nature of generative AI brings forth ethical concerns. As the technology evolves, there is a growing need to address ethical considerations and ensure that AI systems adhere to ethical guidelines. It is vital to foster responsible development and use of generative AI to avoid unintended consequences and potential ethical dilemmas.

DQ: With the advent of Generative AI, there is a renewed focus on Data privacy, data security and ethical & explainable AI. What are the ethical and privacy considerations your organization should address to be able leverage Data & AI to drive business profitability?

Manjeet Dahiya: I agree that generative AI has captured widespread attention and interest, with its potential impact being felt across various fields. However, it is crucial to approach this technology with a clear understanding of ethical considerations. Ensuring that generative AI systems are unbiased and do not inadvertently reveal sensitive information is of utmost importance. For example, when training a generative AI model using personally identifiable information (PII) such as addresses, one must be cautious as the model can output PII as well. If such a model is utilized for tasks like autofilling addresses, it becomes easy for hackers to extract actual addresses from the code. Handling PII and generative models requires a cautious and responsible approach.

Another significant aspect to consider is the presence of biases within the data itself, which can manifest in the generative model. This issue is particularly evident in recommender systems. There is a common question regarding whether generative AI will replace humans. In my view, the goal should be to empower humans rather than replace them. Generative AI can automate tedious tasks, freeing up human resources to focus on designing and contributing to more sophisticated systems. By removing repetitive and mundane work, humans can contribute their creativity and expertise, leading to the development of richer and more meaningful outcomes. This philosophy emphasizes collaboration between humans and AI, where each complements the other's strengths.