Across all industries and verticals, our clients are starting to ask us this question more often. As a digital asset management (DAM) company, we've always considered artificial intelligence (AI) an opportunity, but also a hurdle to overcome. While AI has some major limitations that often aren’t apparent during a shiny demo or an initial test, it also offers some serious power and efficiency that can be leveraged by a DAM program under the right circumstances.
In this blog post, we'll discuss what AI can and cannot do for DAM, and why human oversight and decision-making are still critical in this area.
Where AI Excels
Is there a pattern? Are there faces, numbers, or text? AI often excels in pattern-based tasks and environments. It has the potential to significantly simplify the day-to-day tasks required to manage a DAM program by automating manual tasks, improving asset organization, and providing valuable insights into asset usage and performance. Below are some core tasks AI can help support:
1. Image recognition and tagging: AI algorithms can analyze digital assets such as images, videos, and audio files, and automatically tag them with relevant keywords from a predefined metadata taxonomy, making it easier to search and categorize the assets.
2. Search and retrieval: AI algorithms can enhance search functionality within digital asset management systems, allowing users to quickly find the assets they need based on keywords, image content, or other metadata.
3. Automated categorization: AI algorithms can help categorize digital assets based on features such as color, shape, and object recognition, making it easier for users to find and organize assets.
4. Predictive analysis: AI algorithms can analyze historical data and usage patterns to predict which assets are likely to be in demand in the future, helping organizations prioritize their resources and improve their asset management processes.
AI and Enriching/Tagging Assets
At Stacks, we're often asked about the potential impact of AI during the asset enrichment phase, whether for large file migrations or daily or weekly asset ingestion. While AI algorithms for enriching assets have improved significantly in recent years, they still have significant limitations.
When it comes to enriching assets, we've found that leveraging AI’s strengths, knowing its weaknesses, and combining them with our human capital creates an almost magical, cost-effective combination.
When enriching assets, AI struggles with:
1. Lack of context: AI algorithms have difficulty understanding the context of a photo, which can result in incorrect and/or inappropriate tags.
2. Insufficient training data: AI algorithms rely on the data they're trained on. If the training data is limited, the accuracy of the tags will be too. Due to the volume of assets, this can be a critical deficiency when dealing with asset archives.
3. Ambiguity and subjectivity: Tagging is often subjective, which means two people may tag the same photo differently. This can be difficult for AI algorithms to resolve.
4. Diversity and variations: Photos come in a wide variety of styles and subjects; AI algorithms may struggle to identify and tag rare or unique elements.
5. Creative judgment: One of the biggest limitations of AI is that it can't replace human creativity and discretion when it comes to design or visual content. While AI algorithms can identify patterns and categorize data, they can't make creative decisions about color, composition, or overall visual impact. This is a key factor to consider when setting up approval and quality control workflows.
When enriching assets, humans struggle with:
1. Consistency: Humans may use inconsistent terminology when tagging digital assets, which can make it difficult to search for and retrieve assets later on.
2. Speed: Humans can be slow when tagging large volumes of digital assets, which can be time-consuming and impractical for large organizations.
3. Objectivity: Humans may be influenced by their personal biases and perspectives. This can lead to inconsistent or subjective tagging of digital assets.
4. Repetition: Humans may get tired of repetitive tasks, such as tagging many similar images. This can lead to decreased attention to detail and accuracy.
5. Scalability: The human capacity for tagging digital assets is limited, making it difficult to handle the large volumes of assets generated by organizations.
When enriching assets, humans are great at:
1. Contextual understanding: Humans have a deep understanding of the context in which photos and other digital assets are taken, which can be difficult for AI algorithms to replicate.
2. Subjectivity and creativity: Humans are capable of assigning tags based on their own subjective interpretation of an asset. This can be important for creative or expressive uses of digital assets.
3. Complex reasoning and judgment: Humans are capable of making complex judgments about digital assets based on their content, context, and meaning, which can be difficult for AI algorithms to replicate.
4. Collaboration and consensus building: Humans can work together to agree on tags for a particular digital asset, which can be important in large organizations where different departments may have different perspectives on the use and meaning of assets.
5. Quality control: Humans are capable of performing quality control on tags generated by AI algorithms, ensuring that they're accurate and relevant.
When thinking about how AI can help your DAM program, it's important to remember that it's not a replacement for human governance and oversight. Whether it's in the area of judgment, contextual understanding, complex reasoning, or ethics and governance, human expertise and decision-making will continue to play a critical role in the world of digital asset management.
As we’ve discovered, when you connect a DAM team with the right AI tools you create a mutually beneficial relationship and a powerful and efficient combination. It's important to understand the strengths and weaknesses of both human and artificial intelligence in order to benefit the most from this approach.
To show that we here at Stacks like to put our thoughts into action, this blog post is a result of human oversight filling in the gaps of AI intelligence. ChatGPT helped Robert Boag, our Managing Director, write a blog post. He’s been promising to write one for over two years and it finally happened!
