Demystifying AI

Rob Dunkel
CEO and Founder, 3PM Solutions

AI Mythology

The term artificial intelligence or “AI”- is a marketing and sales term that is actually rarely used by technology and data science teams. Some companies use deceptive and exaggerated sales pitches to define AI that may feel like a Hollywood movie. AI is not a scene from the “Matrix.”  It has become a buzzword among brand protection vendors to create the perception of credibility and competence around certain product offerings. However, AI is not entirely new nor entirely unique; it is no different than other technologies that increase production and efficiency. Don’t get overwhelmed by the term.

Myth 1: AI Gets Smarter the More it is Used.

To be successful in brand protection efforts, it is detrimental to know if the product being listed online is authentic. In most cases, there is only a limited amount of data that can be trusted. When a brand protection vendor says, “our AI continues to learn from your actions, and it gets smarter,” a damaging misunderstanding is being perpetuated. Namely, they are inferring AI can learn on its own. That is not how it works. There is no load it and go. In order for AI to succeed as an effective tool, brand protection professionals and vendors need to work together to approach the challenges with a shared understanding of what are the true possibilities and limitations of AI.

Data is the foundation. To be effective, trusted, reliable and accurate, “clean” data is imperative. How the data is collected, cleaned and stored is critical to success. The team collecting the data needs to care about the data quality as it directly correlates with the success of the analysis. Garbage in creates garbage out results.

Key questions to ask and explore:

  • How does your vendor approach data collection?
  • Are they collecting data on the entire ecosystem, industry or by client?
  • What data are they capturing for your most important websites?
  • Do they have any Service Level Agreement ”SLAs” for collecting data?

If mechanisms for collecting data are too narrow, counterfeit listings will be missed. If data collection is too broad, the number of listings needed to be reviewed will be massive. If vendors are still using humans for this task, this should be a major red flag. To put this in perspective, for one client, over a million listings per hour can be analyzed.

Myth 2: AI Will Take Away Brand Protection Jobs.

AI will change existing jobs by automating some of the mundane work, which will allow people to focus on more important tasks. Don’t take this lightly or give computers too much credit. The goal is to increase the confidence in decisions over time by using meaningful insights and data points. Because of the complexity in brand protection, humans and computers will always need to work together.

 AI: Forms you might already know and use

Data needs to be stored in a way that allows it to be analyzed fast to provide insights. Because e-commerce data is unstructured and not standardized, this is not a trivial task. Knowing this is important, but you don’t have to be a data scientist to ask how a vendor analyzes listing content.

The following forms of AI are valuable to business operations. Some AI you may already utilize:

  • Natural Language Processing (NLP)
    NLP allows users to understand and analyze listing text, customer experience and other insights that contribute to decision-making.
  • Machine Learning (ML)
    Classic methods in machine learning enables users to make decisions about risk levels and decision-making.

    It is used throughout the company to automate tasks and reduce manual processes.
  • Computer Vision
    The importance of images in e-commerce warrants an investment in this area.

    Few Shot Image Classification is cutting edge technology that trains image data faster with fewer images.

    Logo detection helps users identify unbranded listings that try to circumvent counterfeit detection.
  • Pattern Detection
    By analyzing sellers, listings, and storefronts for patterns, bad actors and counterfeit product listings can be identified. By identifying patterns of counterfeits, accounts are identified that are managed by a single person and/or organizations, and listings are removed faster.
  • Deep Listing Analysis
    By collecting and analyzing listings data points, insights are extracted impacting business decisions.

Myth 3: You Don’t Need an AI Strategy.

Companies need an AI strategy or at least a full and thorough review why they don’t need one. It’s about finding use cases where forms of AI technology can increase your team’s efficiency and businesses productivity. As AI myths continue to be debunked, it’s clear that forms of AI are here to stay. Forms of AI should be part of the brand protection team as they are vital to the success of business.

A Parting Truth: What AI Can Do.

AI can help analyze unstructured data faster. It can help companies work more efficiently by removing mundane tasks from their employee’s workload, allowing staff to focus on more strategic opportunities.