Artificial intelligence and machine learning have emerged in the marketing industry as a pathway to competitive advantage. The best marketers are identifying, evaluating and testing AI-driven applications to make better sense of their data, create personalized customer experiences and accelerate revenue growth. In fact, 84% of marketing organizations either implemented or expanded AI and machine learning experiments and implementations in 2018.
While it’s no doubt that artificial intelligence has helped marketing teams improve their productivity, with most brands spending between 25 and 43% of their marketing budget on content, it’s important to understand how AI can impact this specific department.
Driving Content Marketing at Scale
The truth is, artificial intelligence has actually had an active presence in the content marketing industry for years. Various martech platforms have leveraged AI elements to help content-focused organizations build content plans, plan keywords, analyze competitors, create video transcripts and personalize content at scale. Some organizations have even used AI to write simple, fact-based articles. The Washington Post, for example, uses its AI-bot, Heliograf, to provide the public with sports reporting and congressional race updates. Heliograf even reported on the Rio Olympics back in 2016.
Organizations across multiple industries, like technology, consumer goods, manufacturing and healthcare, have made significant investments into their content strategies, building powerful content production engines. For many companies, the volume of created content, videos, visuals and resources can be measured in terabytes.
Managing large volumes of content at scale, across multiple divisions and company locations can become increasingly complex. One application of AI that’s emerging as a solution is auto-tagging, making content easily searchable by keywords, themes or other attributes. This is a process that’s been handled manually for years, taking hours of administration work to tag created content within the organizations content or digital asset management system.
Now, thanks to machine learning, AI models are able to auto-tag using general terms, like color, common objects and text recognition, and can be trained to recognize, tag or index content elements that are unique to specific business cases.
Getting Started with AI for Content Marketing
Leading organizations like Microsoft, Google and Amazon now offer Machine-Learning-as-a-Service (MLaaS), which allows users to enable the auto-tagging of their digital media, saving countless hours. With machine learning, organizations are able to recognize and automatically tag:
- Facial attributes, like gender, age and emotion,
- Objects, scenes, and activities,
- Text within images,
- Company logos,
- And more.
Levels of AI Customization
There are 3 primary levels of AI customization to choose from. Selecting which one is right for you depends on your use-case, machine learning expertise and budget.
Packaged Machine-Learning-as-a-Service is the lowest level of MLaaS customizations. In this level, the provider equips you with a packaged, ready-to-use model with a pre-defined set of artificial intelligence options, like facial recognition, object detection and text extraction. This is the easiest undertaking, as the AI-vendor is 100% responsible for providing the data, training, testing and deployment, so you don’t need any machine learning skills to use it. Once connected with your library, you’re able to start using the model immediately to tag your visual content. Packaged MLaaS is used when an organization has a relatively basic AI use-case, where more generic auto-tagging terms are acceptable for their business.
Example: A tourism company is uploading thousands of vacation photos to their media library every week. They don’t have the time or manpower to manually tag the photos themselves, so they connect a Packaged MLaaS model to their library to auto-tag them. Now, when they need an image for a honeymoon webpage, they can search for “Couple”, “Beach” and “Sunset”, and the library will return photos that have been automatically tagged by the Packaged MLaaS model.
Packaged MLaaS Options: Microsoft Azure Computer Vision, Google Cloud Vision, Amazon Rekognition
Guided Machine-Learning-as-a-Service is more complex and is used to address specific business use-cases. With this customization level, the machine learning vendor provides you with a space where you can classify your own media, adding images and tagging them appropriately to teach the model until you’re satisfied with its confidence level. As you’ll be providing the training and testing, you need to understand basic statistics and precision levels, so you can evaluate if the machine is at the right level of confidence. It does take some time to train the model – Microsoft, for example, suggests about 30 images per tag to get an acceptable level of confidence. This form of MLaaS is used if there’s a business case that can’t be satisfied using generic auto-tags.
Example: A telecommunications company has a library to store visuals of the phones they offer. As they sell a variety of brands and generations of phones, the auto-tag “Cellphone” is too generic to make their visuals easily searchable. They decide to implement a Guided MLaaS model, so they can train the model to differentiate between the various brands and generations (such as an iPhone 8 or a Samsung Galaxy S9). After training the model to a 99% confidence level, they connect it with their library so it can start auto-tagging newly uploaded visuals with the correct phone tags.
Guided MLaaS Options: Microsoft Azure Custom Vision, Google AutoML, IBM Watson Visual Recognition
Specialized Machine-Learning-as-a-Service is the model that provides the most flexibility across tooling, platform and infrastructure, but also requires a full understanding of machine learning models and system integrations. With Specialized MLaaS, the vendor provides you with a virtual machine that’s prepackaged with standard machine learning software and add-ons, so that an internal IT team or 3rd party can code, train, package and deploy the model. In this case, you need to have a data scientist and a developer or technical integrator to provide the data, training, testing and deployment. Specialized MLaaS is only used for very specific use-cases and requires the greatest resource investment of all the MLaaS models.
Example: A pharmaceutical company is performing a scientific experiment that requires the identification and tagging of different strains of bacteria. As the differences between the bacteria are so slight and require advanced knowledge to identify, the company’s use-case is too complex to be satisfied using Guided MLaaS. They hire a data specialist and programmer to build a fully-customized AI model using Azure Machine Learning Service. After developing, training and testing the model, they can begin auto-tagging each specific strain of bacteria.
Specialized MLaaS Examples: Azure Machine Learning Service, Google Cloud ML Engine, Amazon Sagemaker
The Future of Content Marketing
Regardless of the MLaaS level that’s right for your organization, enabling your team with artificial intelligence and auto-tagging will enable them to prepare for the future of content marketing. These 6 best practices for implementing AI in your media library can help you get started.
About the author
In partnership with Technology Partner Microsoft Canada, this blog was published by Carlie Hill, Content Marketing Manager at MediaValet; located in Vancouver, BC.