In today’s world, relevant digital content is at the center of customer experience, with consumers expecting to engage with images, videos, podcasts, and more. As photo and creative studios shut down due to COVID-19, organizations that use AI to mine and leverage their existing assets can build new marketing campaigns and customer experiences by using what they already have in new and exciting ways.

In a recent conversation with the Henry Stewart DAM community during the webinar Enhancing Digital Asset Management (DAM) Operations with Artificial Intelligence (AI), we discussed how AI and machine learning (ML) can improve how organizations tag, find, and use rich media assets with increasing opportunities to drive new revenue streams and scale content operations across the organization. Below are some of the ways organizations are thinking about managing their digital assets when it comes to AI and ML in today’s world.

Everything You Need to Know About Digital Asset Management - Read More Guide

AI for the Digital Supply Chain

Avoiding Expensive Costs Recreating Content

One of the most important parts of an enterprise DAM is knowing what digital assets you have - and then being able to find, access, use them. Research shows that when people can’t find an asset they’re looking for, about 60% of the time they end up recreating it - costing organizations as high as \$50,000 to recreate content that already exists.

The need to recreate expensive projects over and over is a real issue for many organizations, one that could be resolved with a DAM that enables teams to find, access and use the existing content.

Accelerating Ideas to Market To Drive Value

The best DAM systems go beyond just a creative studio or marketing function. Everything from materials, to design, photography, packaging development process, marketing campaigns, and all the way to the sales floor need access to content and the data that describes it. It’s important to have a content hub software to store and search for digital assets.

Connecting all of these processes accelerates the digital content supply chain - decreasing time to market, and in turn driving higher revenue.

The old ways of managing content and data are over. Assets are spread out all across an organization, and often aren’t connected to one another. Research shows that an organization works across, on average, 9 different systems - causing teams to spend hours just looking for content. Many times, they end up unable to find it, and have to recreate it.

Enhancing DAM Operations with Artificial Intelligence

Consistency, Speed, and at Scale

Leveraging AI and ML plays a large role in connecting various systems across organizations, working alongside humans to do things fast, consistently, and at scale. AI is being used to determine content types, extract data, and enrich the content metadata to make assets more useful across the organization. With ML, organizations can predict delivery information and what’s important to their customers - noticing patterns and connections. Using AI and ML, it’s helping teams do things consistently, fast, and at scale.

Generic AI vs Business-Specific AI

Generic AI is a good foundation to enrich digital assets, and can be leveraged using a broad set of services widely available - commodity models. It’s really helpful for searchability, but fairly generic, used for things like: transcribing video, extracting text, automating translations, etc.

Building upon generic AI, organizations are using business-specific AI with custom machine learning (ML) models trained on their content and data to enrich assets. Business-specific AI enables organizations to describe their assets in the context of their business which is much more meaningful than generic tags.

Read this to lear more about the difference between generic AI and custom AI

AI for Content: Nuxeo Insight

At Nuxeo, our business-driven AI offering is called Nuxeo Insight, where we can form intelligent predictions based on the information you already have. Classify, predict, and enrich rich media, documents, and other content, without coding or ML expertise.

When it comes to tagging, humans are only accurate about 80% of the time. But AI and ML can produce more accurate and consistent results while freeing up time for creative tasks that human minds excel in.

Organizations are beginning to leverage AI and ML to bring real business value to the organization, focused on:

  • Avoiding expensive costs recreating content
  • Accelerating ideas to market across the digital content supply chain to drive value
  • Reducing manual tagging and mundane activities
  • Generating new insights about their content

AI for autotagging

Whether you’re already using some form of AI or beginning the journey now - it doesn’t take much to get started. You can start training with 200-300 images and then continually train as more images are added to improve accuracy and make assets more findable, therefore more valuable.

Real-Life Examples of AI Used in Digital Asset Management

During the webinar, we discussed how a top clothing manufacturer is using AI and ML to fuel their data - recognizing image types, talent records, costs, seasonality, geographies, etc, and using that data to promote related products online, which leads to cross-sell and up-sell opportunities – in turn driving more revenue.

AI to fuel data in DAM system

Another example we explored was using business-specific AI to enrich an automobile accident photo, with details that make it searchable and contextualized such as, vehicle manufacturer, manufacturer color, model, license plate state, facial recognition, predictive analysis etc.

AI for enriching Digital Asset Management

You can understand here that custom models produce much more relevant data values, true entity extraction enables workflow automation, and that the specificity of the data increases the business value of your chosen AI system.

You can watch the recent webinar Enhancing DAM Operations with AI here:

Some FAQs around the topic of DAM and AI that we answered in the webinar include:

  • Assuming the metrics for success with AI are accurate, is there still an oversight role for humans to manage content and ensure maximum accuracy? Especially as metadata is a living, breathing component of DAM.

Definitely. Humans are an essential part of growing and training machine learning. We need to ensure the results are accurate. Maintenance is not a one and done thing. We’ll need to do audits, and more. Humans will definitely be a part of the DAM process, our roles just may change slightly - freeing up time to be more creative, and do the things that humans do well, while machines do the repetitive tasks.

  • How are these services for Nuxeo integrated as well as paid for?

It’s an add on to the Nuxeo Platform.

  • What is the role of adopting particular metadata standards/specific vocabularies? Can you install these schemas into an existing AI program?

Yes! If you’re starting from scratch, we also have a wizard-based UI that can be used by folks who are not data scientists.

  • Where’s the best place to get started to apply AI to my DAM management process?

Find something that is going to provide real value to the company. Make things more findable, etc. Speed things up for everyone - a project that needs to be done quickly and needs consistency.

  • How long does it take to configure, to train the model?

It depends, but it can be done within a few hours.

  • Can Nuxeo accurately read text on images as well?

Yes - It can do OCR text and read text

  • Have any Nuxeo clients pursued new content creation based on AI-driven insights?

Yes, especially due to the impacts of COVID-19, as commercial photo studios and creative studios are shut down. Organizations are using AI to mine and see what existing assets they have and can use to build new creative campaigns and customer experiences around during this time. People are looking at what they’ve got and how they can be more efficient with things they have already.

  • Are the keywords used in AI tagging taken from the DAM’s controlled vocabulary?

Yes, that’s where you’d likely want to pull the content from and teach it

  • Do you run only one model at a time, or do you run multiple models simultaneously?

You run multiple simultaneously.