Unlocking the Power of AI and?ML

Artificial Intelligence (AI) and Machine Learning (ML) are rapidly evolving fields that are transforming the way we work and live. Businesses across various industries are adopting AI and ML technologies to automate processes, increase productivity, enhance customer experiences, and gain a competitive advantage. As someone with over a decade of experience in engineering and product, I have had the opportunity to work with AI technologies and witness their impact firsthand. In this write up, I would like to share some insights based on my experience and interaction with industry leaders on how organizations should carefully focus on planning, experimentation-iteration to unlock the power of AI and leverage it to drive business success. These critical elements are often neglected due to over excitement around technology buzzwords and the rush towards solving complex problems for self satisfaction.

Start with a clear problem statement

Starting with a clear problem statement is a pre-requisite critical step in AI/ML implementation. Without a clear understanding of the problem you are trying to solve, you risk investing time and resources into a product that may not provide the desired results. A well-defined problem statement provides clarity on the objective of the product, the data required, and the success criteria. It also helps to identify potential risks and limitations.

A company may want to improve customer satisfaction by reducing the time it takes to resolve customer complaints. The problem statement could be: “We want to reduce customer complaint resolution time by 66% in the next three months by automating the complaint handling process.” or “We want a personalized shopping experience through a recommendation system that suggests products with a similarity score of at least 0.8, so that customer can quickly find products that match his/her interests and preferences.”

With a clear problem statement, the company can identify the data required, the resources needed, and the success criteria. Once they have a clear problem statement, they can then explore how AI / ML can help personalize the experience for individual customers, making resolution recommendations based on their activity.

In the healthcare industry, where AI/ML is being used to improve patient outcomes. One clear problem statement could be: “We want to reduce hospital re-admissions for heart failure patients by 22% in the next 9 months by using predictive analytics to identify patients at high risk of readmission.” or “I want a ML model that can accurately predict the progression of a patient’s disease with 88% accuracy, so that I can optimize their treatment plan and improve their outcome “

A bank want an AI algorithm that can “detect fraudulent transactions in real-time with a false positive rate of no more than 1%, so that it can prevent financial losses and maintain the trust of customers”.

A car manufacturer want “a predictive maintenance system using ML that can reduce downtime by 33% and increase efficiency by 11%, so that it can perform maintenance only when it’s necessary”. or “want to develop self-driving vehicles using AI and ML technologies to improve safety and reduce accidents by 55%, so that I can provide a better driving experience for my customers” .

Searching for a needle in a pond blindfolded is akin to looking for a solution without a clear problem statement.

Embrace experimentation with iteration

AI / ML products are complex and require experimentation and iteration to refine models and improve accuracy. Embracing experimentation and iteration allows businesses to test and refine their models to ensure they are providing accurate results. It is an established fact that no AI/ML model can achieve near-perfect results in just one iteration, regardless of the amount of time invested. In my experience, a noticeable pattern emerges after achieving 50–60% accuracy, where each incremental percentage increase requires exponentially more time and effort.

In the realm of AI / ML, early issue identification is key to success. To achieve this, it’s crucial for companies to embrace risk-taking and adopt a patient approach to learning from failures. It is important for businesses to take calculated risks and approach each project with a patient mindset, recognizing that not every endeavor will yield immediate success. Failure should be viewed as a learning opportunity, allowing for continuous improvement. It’s worth noting that clear identification of the problem statement is a fundamental prerequisite. In some exceptional cases, however, it may be necessary to adjust the approach or redefine the problem statement after a few iterations.

For example, a B2C organization marketing team may use AI / ML to predict customer churn. They could start with a simple model that predicts churn based on a single variable, such as customer tenure. Through experimentation and iteration, the team can test and refine the model by adding more variables, such as customer demographics, purchase history, and website activity. This iterative process allows the team to improve the accuracy of the model and identify the variables (Products that are currently trending or popular, Products that are on sale or have special offers, Products that have good customer reviews, Products that match my demographic profile (e.g. age, gender, location)), that have the most significant impact on churn.

Another example is in the insurance industry, where AI / ML is used to identify fraudulent claims. The initial model may start with simple rules-based logic to identify potential fraud, such as claims filed outside of normal business hours. Through experimentation and iteration, the model can be refined by incorporating additional variables, such as past claims history, network analysis, customer’s credit score and financial history, customer’s occupation and employment status, customer’s previous claims history. This iterative process allows the model to improve accuracy and identify new patterns of fraud.

I have observed that through multiple iterations for better accuracy, there can be interesting outcomes that have the potential to refining variable selection, re-optimizing model architecture / complexity, re-experimenting with sampling techniques, and re-implementing effective ensemble strategies.

However, experimentation and iteration can also be time-consuming and resource-intensive. It is important to balance the benefits of experimentation and iteration with the constraints of time and resources.

In conclusion, starting with a clear problem statement and embracing experimentation with iteration are critical ingredients for successful AI/ML implementation. By beginning with a clear problem statement, businesses can identify the objectives, required data, and success criteria. Embracing experimentation with iteration allows businesses to refine their models, improve accuracy, and identify issues such as over-fitting or under-fitting early on in the project. By focusing on these elements, businesses can leverage AI/ML technologies to automate processes, enhance customer experiences, and gain a competitive advantage.