Institute of Product Leadership
Search
Close this search box.

Every Company will be an AI Company

By Prateek Gupte – Co-Founder at Caimera

Imagine you’re managing a music streaming service. You want to keep users engaged by recommending songs they’ll love, but how do you determine what music to suggest? This is where artificial intelligence (AI) comes into play. By understanding and applying AI, you can transform your product into a personalized experience that keeps users coming back.

Artificial intelligence is changing the way products are designed, built, and managed. For product managers, mastering AI can unlock new opportunities, improve processes, and deliver exceptional value to users and stakeholders. This blog explores five key areas of AI integration: understanding AI terminology, finding the right machine learning (ML) opportunities, defining business outcomes, designing user experiences, and managing expectations.

Key Takeaways:

  • Differentiate between AI, machine learning, and deep learning to apply the right techniques for various business problems.
  • Categorize ML problems into ranking, recommendation, classification, regression, clustering, and anomaly detection to find suitable applications.
  • Focus on the broader business impact and establish baseline metrics before implementing ML features.
  • Ensure transparency, set expectations, and incorporate human oversight to enhance user experience and manage AI limitations.
  • Communicate AI’s probabilistic nature, avoid overpromising, comply with data privacy laws, and focus on delivering practical value.
In this article
    Add a header to begin generating the table of contents

    Understanding the Nomenclature of Artificial Intelligence

    Artificial Intelligence (AI) is a broad term that encompasses a wide range of techniques enabling machines to mimic human behavior. This vast field includes multiple disciplines such as Robotic Process Automation (RPA), robotics, and other specialized areas.

    Breaking Down AI

    1. Artificial Intelligence (AI): A broad umbrella term for techniques that enable machines to mimic human behavior. This includes everything from robotics to RPA and other disciplines.
    2. Machine Learning (ML): A subset of AI focused on enabling machines to make decisions based on data. This means codifying decision-making processes dynamically, rather than statically defining them. Essentially, ML allows machines to learn from data and improve their decision-making over time.
    3. Deep Learning: A further subset of ML, deep learning involves using multi-layer neural networks to make decisions. It’s a more complex and nuanced approach within machine learning, relying on layered structures to process data and make informed decisions.

     

    Choosing Between Machine Learning and Deep Learning

    The decision to use a deep learning model versus a simpler machine learning model is more technical. It’s typically left to the discretion of your engineering team or counterparts, who will decide based on the specific requirements and constraints of the task at hand.

    Opportunities for Machine Learning: A Problem-Solving Perspective

    When considering opportunities for machine learning (ML), it’s crucial to approach problem statements from a product outcome or problem perspective rather than a technology perspective. This mindset shift allows you to identify and solve real-world issues effectively using ML techniques. Here, we’ll explore six main categories of problem statements in ML and provide examples to illustrate each one.

    1. Ranking

    Ranking is a common problem in ML, where the goal is to order a set of results based on relevance to a user’s query. The most familiar example is search engines like Google or Bing, which rank search results according to relevance.

    Example: When a user inputs a search term, the engine returns an ordered list of results, with the most relevant at the top. E-commerce sites also use ranking systems to order products based on user preferences and past behaviors, often combined with recommendation systems.

    2. Recommendation Systems

    Recommendation systems aim to suggest items that a user is likely to be interested in, based on their past behavior and preferences.

    Example: Spotify’s Discover Weekly playlist curates songs for users based on their listening history and other input. Similarly, Netflix and Twitter recommend movies or friends to follow based on user behavior and interests.

    3. Classification

    Classification involves sorting items into predefined categories. This is often seen in applications like email spam filters and image recognition.

    Example: Google’s spam filter classifies emails as spam or not spam. Image detection systems classify images as containing specific objects or not, such as recognizing whether a photo includes a specific person or not.

    4. Regression

    Regression is used to predict a continuous outcome based on input data. This is common in pricing models and forecasting.

    Example: Kayak uses regression to predict future flight and hotel prices based on historical data. The model forecasts whether prices will rise or fall, helping users decide when to book.

    5. Clustering

    Clustering is an unsupervised learning technique used to group similar items together without predefined labels.

    Example: At Haptech, we built a clustering algorithm to identify groups of similar queries that a chatbot did not understand, such as various types of policy-related questions. The algorithm groups these similar queries together to help improve the chatbot’s responses.

    6. Anomaly Detection

    Anomaly detection identifies unusual patterns that deviate from the norm. This is widely used in finance and social media.

    Example: In finance, anomaly detection can spot irregular trading activities. On Twitter, it identifies trending topics by detecting unusual spikes in conversation rates, highlighting topics growing at an unnatural rate compared to others.

    executive mba in product management

    Identifying Business Outcomes for Machine Learning Projects

    To effectively leverage machine learning (ML) in your projects, it’s crucial to clearly define the business outcomes. These outcomes differ from use cases and focus on the broader impact on your business. Here’s a step-by-step guide to help you navigate this process.

    Define the Business Outcomes

    Business outcomes are the end goals you want to achieve through your ML initiatives. They are not merely the technical use cases but the broader objectives that drive value for your business.

    Example:

      • Spotify’s Discover Weekly: The business outcome was to increase user engagement and retention by recommending music users would like, thereby encouraging them to return to the app frequently.
      • Hotel Booking Example: The desired outcome was to boost booking rates by providing users with timely information on price trends, giving them confidence in their purchasing decisions.

    Quantify the Business Outcome

    Before implementing an ML feature, it’s essential to establish a baseline for your key metrics. This helps you measure the impact of the new feature accurately.

    Steps:

      1. Establish Baseline Metrics: Determine your current performance metrics. For instance, if your conversion rate is 2-3%, know this before deploying your ML algorithm.
      2. Measure Post-Implementation: After deploying the ML feature, track changes in these metrics to see if the desired outcomes are achieved.

    Example of Quantifying Outcomes:

    When Kayak wants to predict flight prices, they start by establishing the current booking rates. After deploying an ML model to predict price trends, they measure if the booking rates increase, indicating user confidence in making purchases.

    Model Performance vs. Business Outcomes

    A high-performing ML model does not guarantee the desired business outcomes. Sometimes, even a well-functioning model can miss the mark if it doesn’t align with business objectives.

    Example:

      • Netflix’s Star Rating System: Netflix used an ML model to predict star ratings for movies. While the model worked well, it didn’t improve user recommendations, leading to a decline in viewership. The business outcome wasn’t met because the model didn’t align with user behavior.

    Start Small and Iterate

    Begin with smaller, manageable ML projects that can deliver quick wins and iterative improvements.

    Example: Instead of automating the entire stock trading system, start with small recommendations or human-in-the-loop systems. This allows for gradual scaling and refinement based on real-world feedback.

    Collecting and Managing Data

    Data collection is a critical aspect of ML projects and requires collaboration between product managers, engineering teams, and data scientists.

    Steps:

      1. Identify Data Sources: Determine where and how you will collect the necessary data. In B2B environments, data may be readily available from customers, while B2C environments might require more effort to gather initial data.
      2. Legal and Ownership Considerations: Ensure you have the right legal frameworks and understand data ownership, especially when dealing with multiple clients or users.
      3. Data Quality and Cleaning: High-quality data is essential for accurate ML models. Implement processes to clean and label data, whether through in-house teams or external agencies.

    Practical Data Collection Tips

      1. Buy or Source Data: Use available data sets from vendors, especially in domains like finance, NLP, or image recognition.
      2. Collect Your Own Data: Establish systems to gather data, keeping in mind the costs and efforts involved.
      3. Data Management and Storage: Develop robust methods for storing and managing large volumes of data.
      4. Data Cleaning: Implement rigorous cleaning processes to ensure data integrity, filtering out irrelevant or erroneous data.

    Designing User Experience for AI Products: Key Considerations and Boundaries

    After defining your use case, identifying business outcomes, and gathering data for your model, the next crucial step is to focus on user experience (UX) design for AI products. Often overlooked, this step is vital for addressing the inherent limitations of AI systems, which are inherently probabilistic and not foolproof. By emphasizing good UX design, you can mitigate these limitations and enhance user satisfaction.

    Importance of UX in AI Products

    AI systems deal with probabilities, and their outcomes can vary. Effective UX design helps manage user expectations and improves the overall interaction with AI-driven features. Understanding and communicating the boundaries of AI systems are crucial in this context.

    Understanding AI Boundaries

    AI systems have inherent limitations, primarily due to their probabilistic nature. These boundaries must be clearly communicated and managed through effective UX design to ensure users have a realistic understanding of what the AI can and cannot do.

    Best Practices and Examples

      1. Transparency and Expectation Setting
      • Example: Always clearly communicate whether users are interacting with a bot or a human. This transparency sets appropriate expectations and avoids misleading users. People are generally more forgiving of errors from a bot than from a human, so clear communication helps manage user expectations.
      1. Contextual Boundaries
      • Example: Virtual assistants should set clear boundaries by specifying what they can help users with, such as providing status updates or information on specific topics. This helps users understand the bot’s capabilities and limitations, reducing frustration and improving the interaction experience.
      1. Communicating Probabilistic Outcomes
      • Example: Features like price prediction should include a confidence level (e.g., 89% confidence that the price will drop). This transparency about the probabilistic nature of predictions helps users make informed decisions and manage their expectations.
      1. Human-in-the-Loop Systems
      • Example: Certain recommendations may require human validation before being fully automated. This approach combines the efficiency of AI with the reliability of human oversight, ensuring higher accuracy and user satisfaction.
      1. Iterative Product Design
      • Example: Start with simpler algorithms, such as k-means clustering, to gather initial feedback and validate the concept. Once proven valuable, you can invest in more complex solutions like deep learning to enhance performance gradually.
      1. Handling Insufficient Data Scenarios
      • Example: For apps generating playlists or recommendations, clearly guide users on how to provide necessary data if initial data is lacking. For instance, inform users that adding a few songs will enable the system to generate personalized playlists.
      1. Collecting User Feedback
      • Example: Implement features that allow users to provide feedback, such as like or dislike buttons for recommendations. This iterative feedback loop is essential for refining AI models and enhancing user satisfaction.
      1. Frequency and Data Management
      • Example: Update features like personalized playlists on a regular schedule that balances computational cost with user expectations. Additionally, allowing users to reset preferences or clear history helps manage long-term biases in recommendation systems.
      1. Adhering to Data Privacy Laws
      • Example: Ensure compliance with data privacy regulations like GDPR, the California Consumer Privacy Act, and India’s personal data protection laws. These laws impact how you collect, store, and use user data, necessitating careful planning and integration into your product design.

    Emerging Roles in AI Product Design

    The role of design in AI products is evolving, leading to the creation of new job titles and skill sets.

    Example: The role of “Conversational Designer” blends AI, UX, and copywriting. These designers focus on crafting content that communicates the bot’s capabilities, handles failures gracefully, and guides users effectively.

    Managing Expectations and Setting Boundaries for AI Products

    After developing an AI product, setting clear expectations with customers and internal stakeholders is crucial. AI is inherently probabilistic, meaning outcomes cannot be guaranteed with absolute certainty. This reality often poses challenges in communicating the capabilities and limitations of AI systems. Here are some key considerations and best practices for managing expectations and setting boundaries for AI products.

    Understanding AI’s Probabilistic Nature

    AI systems, by design, operate on probabilities and cannot offer guaranteed results. This probabilistic nature must be clearly communicated to avoid misunderstandings and unrealistic expectations.

    Example: When selling call center automation software or chatbots, it’s essential to clarify that the bot may work correctly around 80% of the time. Highlighting this helps manage expectations about the system’s reliability and sets the stage for realistic performance metrics.

    Avoiding Accuracy Guarantees

    Promising specific accuracy levels in contracts can lead to complications. AI systems’ real-time accuracy is difficult to measure precisely, as it’s based on dynamic and live data rather than static training data.

    Example: If a customer demands an 80% accuracy guarantee in a contract, it can lead to significant manual efforts to validate and ensure this accuracy. It’s more practical to focus on the system’s value and benefits rather than making hard guarantees on accuracy.

    Data Ownership and Privacy

    Data privacy laws and ownership concerns are critical, especially in B2B and B2C contexts. Ensuring compliance with data privacy regulations and obtaining proper consent for data usage are fundamental.

    Example: When deploying AI solutions, make sure to have clear agreements on data ownership and usage rights, especially when dealing with customer data. This helps in maintaining compliance and building trust with users.

    Marketing and Positioning

    It’s tempting to market AI products by highlighting the use of advanced technologies like deep learning and neural networks. However, focusing on the tangible value and benefits of the product is more effective.

    Example: Instead of saying, “We use cutting-edge deep learning technology,” emphasize, “Our solution reduces your customer support costs by 30%.” This approach highlights the practical value and resonates more with potential customers.

    Selling to the Right Audience

    Understanding who buys your product and their motivations is key. Selling to innovation teams might yield short-term gains, but targeting operational teams ensures long-term value and sustainability.

    Example: Innovation teams may be drawn to the AI hype, but operational teams will focus on the product’s practical benefits and ROI. Selling to the latter ensures that the product delivers real value and meets business objectives.

    Clear Communication of AI Limitations

    Being upfront about what AI cannot do is as important as highlighting its capabilities. This transparency helps build trust and sets realistic expectations.

    Example: If voice technology struggles with recognizing names and addresses accurately, make this clear to users. Explain the current limitations and provide alternative methods to achieve the desired outcomes.

    Incorporating Services Component

    AI products often require ongoing services for data collection, tagging, and model maintenance. This services component should be integrated into the business model.

    Example: Continually maintaining and updating AI models to adapt to changing data and user intents is essential. This ongoing effort should be factored into pricing and product strategy to ensure sustainability.


    Leveraging AI in product management involves a multidisciplinary approach, combining technical knowledge, UX design, and strategic marketing. By understanding AI nomenclature, identifying ML opportunities, defining clear business outcomes, designing effective user experiences, and managing expectations, product managers can harness the power of AI to deliver exceptional value and drive business success. As AI technology continues to evolve, staying informed and adaptable will be key to maximizing its potential.

    About the Author:

    Prateek Gupte – Co-Founder at Caimera

    Frequently Asked Questions

    Machine learning is a subset of AI that enables machines to make decisions based on data, allowing systems to learn and improve over time without being explicitly programmed.

    AI impacts product management by providing tools to analyze data, predict outcomes, automate processes, and personalize user experiences, ultimately leading to better decision-making and enhanced product value.

    Examples of AI in everyday life include personalized recommendations on streaming services like Spotify, voice assistants like Siri, predictive text on smartphones, and customer support chatbots.

    Companies use AI to enhance customer experiences, improve operational efficiency, predict market trends, personalize marketing efforts, and automate routine tasks, thereby driving business growth and innovation.

    Challenges of implementing AI in products include managing user expectations, ensuring data privacy and security, obtaining high-quality data, avoiding overpromising AI capabilities, and integrating AI seamlessly with existing systems.

    Institute of Product Leadership is Asia’s First Business School providing accredited degree programs and certification courses exclusively in Product Management, Strategy, and Leadership.
    Explore Our Programs
    Facebook
    Twitter
    LinkedIn
    Our Popular Product Management Programs
    executive mba in product leadership
    Our Trending Product Management Programs
    Trending Posts
    WordPress admin themes

    About Institute of Product Leadership

    About Us
    Contact Us
    Faculty & Coaches
    Alumni
    Success Stories
    Press & Media
    Corporate Training
    Career Management
    Hire Talent
    Privacy Policy
    Refund Policy
    Student Loans and Financing
    Sitemap
    Careers @ IPL

    Product Management Courses

    University Programs
    Executive MBA in Product Leadership
    Certificate Programs
    Certified Associate Product Manager
    International Certificate in Product Management
    Post Graduate Program in Product Management
    International Certificate in Strategy & Leadership
    Product Management Professional Workshop

    Product Community

    Product Leaders Forum
    IPL Alumni Network

    Free Resources

    Events
    Blogs
    Knowledge Center
    Toolkits & Templates