domeet webmaster

Let’s not be fooled by the concept of “AI” …

Editor’s note: This article is from WeChat public account “PM Bear Uncle “(ID: pmxiongshu) .

Author | Patrick Cao Product Director, PATRICK TSAOGetsafe

Translation | PM Bear Uncle

Starting in March 2019, we are building a product called GET Smart Writing. Due to lack of experience, we have done some seemingly cool but unable to solve the problem. Looking back at these errors, the essence of the problem lies in the failure to match the problems in the scene with the AI ​​capabilities.

Today, I have “forgotten” the concept of AI, because I know that the real user value comes from solving specific problems in the scene, rather than simply creating a cool experience. I remember one time an investor asked me: “What do you think AI can’t do?” I replied firmly: “The last thing AI can do is to replace humans.”

This article translated today also coincides with my point of view. The author suggests that we should not be confused by the technical concept of “AI”. AI is only a tactic to solve problems, not a company’s vision and strategy. In addition, he also provided us with a useful decision framework-the “scenario-decision matrix” for assessing the suitability of AI. When you are ready to use an AI or data technology in your product, check out this article.

Original Title: Ground Rules for Applying AI to Product Management

Published: January 16, 2020

With the hype of the concepts of artificial intelligence (AI) and machine learning, various terminology has also come into our eyes, making this powerful technology more difficult to understand. Below, I have listed some tips that are useful to me personally. I also hope that this article can help product managers reduce noise and understand how to better integrate artificial intelligence into their daily work.

Application of Machine Learning in Vehicle Recognition (Image: Shutterstock)

I. A broad definition helps to solve the problem better

First, let’s forget what artificial intelligence is. Based on my experience with data-driven products, I have found that a broader definition of artificial intelligence will help me focus on the problem to be solved rather than focusing on a certain technology used in the solution .

As product managers, we must have a deep understanding of our problem space so that we can define requirements correctly and let our team solve the right problems. But before we define the problem correctly, we tend to think “how to do it” prematurely, and we unconsciously lean towards a particular solution.

To avoid this, I prefer to define AI as “automated decision making.”

Most of the products we make require “decision making” based on data, but “decision making” methods can be diverse. It can be made by machines or humans, and data can be static or dynamic. When our attention is focused on the word “decision”, we can avoid the complexity of specific methods and the noise of industry terminology in an abstract way. This broader definition can prevent us from getting into the premature consideration of a solution, allowing product managers to focus more on the problem space.

II. AI is a tactic used to solve problems

Product managers need three indispensable cornerstones before they can act:

Vision: our ultimate goal;

Strategy: do the right thing to achieve our vision;

Tactical: execute our strategy correctly;

Regardless of the stage of the product life cycle, product managers need to do these three things. We must align our team with the vision we are working to achieve, while also ensuring that everyone understands our strategy and integrates feasible tactics into the overall plan. We need to note that artificial intelligence is a tactic that can be used to solve specific problems, not a strategy or vision. If we lack the ultimate goal, even the use of artificial intelligence will not bring any value to the end user.

Take an example of a specific online video company, NetFly. Netflix is ​​one of the earliest companies to effectively apply artificial intelligence on a large scale:

Netfei’s vision: to be the best entertainment publishing service provider in the world;

NetFly’s strategy: to enhance member retention through an attractive and personalized experience;

Netfly’s tactics: scoring system, recommendations, heroic shots, usage tracking, etc.;

As you can see, under the strategy of personalized experience, there are many different tactics to help NetFly achieve the goal of improving retention. The use of data and AI varies by tactic, and the description of the vision and strategy does not indicate which technology or algorithm is used.

III. AI empowers humans instead of replacing them

Currently, the topic of automation has led to some interesting discussions on the ethical issues of future work, and with it, people have some confusion about how AI can empower humans. A common example is self-driving cars. Within the automotive industry, there are five levels of automotive automation. Most of the topics focus on how the world will change after the realization of fully autonomous driving (level 5).

Product managers must recognize that the development of AI capabilities requires stages and time, not overnight. Machines are different from humans, so certain decisions are easier to automate. High-performance AI capabilities first require a large training data set. This training data set is not only large, it also needs to have a good data structure and be easy for the machine to read. Ideally, this data set should also define clearly what is success and failure, because past results are predictions of the future.

Here is a framework that I often use when considering how to apply Automated Decisions:

Along the Y axis, Routine Scenarios occur very frequently and their variability of spread is low, while Fine Scenarios occur very infrequently, and may contain difficult to reproduce detail.

Along the X axis, Informational Decisions provide end users with additional contextual information, and Action-oriented Decisions perform actions instead of End Users. Conventional scenarios tend to generate more reliable training datasets, so machines are easier to learn; informational decisions are often less risky than action-oriented decisions.

Combining these two dimensions, you can get four types of automated decisions:

• General information: easy to predict, low risk of error:

  For example: the car estimates the remaining driving distance based on the remaining fuel and driving behavior;

  • Detailed information: difficult to predict, low risk of errors:

    For example: the car prevents the driver from falling asleep based on image recognition and driving behavior;
    

  • Routine action: easy to predict, high risk if wrong:

    For example: Under normal circumstances, the car is driving on high-speed highways;
    
    

    • Delicate action: difficult to predict, high risk if wrong:

      For example: cars are driving autonomously in busy construction areas.
      

      The influence of three different ways of AI

      In the past decade, by studying data-driven products, I have identified three main use-case groups that can help product managers increase their data impact:

      1.AI can optimize or automate business operations

      Products can generate a data set by using reasonable behavior tracking that empowers teams to make informed business operations decisions. For example, you can use data to optimize customer touchpoints and communications to increase conversion rates or reduce customer churn. By predicting topics or outcomes, service requests can be filtered and assigned more efficiently. In a sense, artificial intelligence is an advanced business intelligence tool that can improve team efficiency and productivity.

      2.AI can significantly improve the user experience of the product

      Here are some examples of companies using AI to create delightful experiences for customers:

      In these examples, the form of the product delivered to the end user has not changed (for example, Uber’s mobile terminal), but through application data (for example, matching passengers to nearby Uber drivers), the product experience will better. With this model, product teams can often create unique user experiences and become a company’s long-term competitive advantage.

      3.AI can fundamentally change products

      Perhaps the most famous example is the story behind NetFly’s TV series House of Cards, which uses data to redefine how entertainment is created. This series has not only won many awards, but also loved by Netflix users. This also marks the company entering a new period of significant growth. This shows that AI has the potential to create new product categories and new trends for the entire industry.

      

      V. Conclusion

      To sum up, product managers can refer to these four basic principles when thinking about how to integrate artificial intelligence into products:

      • The broad definition of AI can help us focus on solving problems instead of thinking directly about the final solution;

      • AI is a tactic that helps solve problems, not a strategy or end goal.

      • AI can empower humans but not replace them.

      • Three different ways in which AI works: optimize operations, improve product experience, and create new product categories.