Generative AI models use machine learning techniques to generate text, images, audio and video. These models are trained on vast datasets, learning patterns and structures within the data to produce outputs that mimic human decision-making.
In marketing applications, generative AI is often used in tandem with traditional AI to drive efficiency. To take a simple example, generative AI might create advertising copy and imagery, while machine learning determines which customers receive a particular creative asset.
While OpenAI’s GPT-4 and Dall-E remain some of the most publicly recognized models, increasingly leading-edge organizations are creating customized or semicustomized generative AI solutions trained on brand-specific or task-specific datasets. For example, IBM’s granite library of foundation models are trained on enterprise data from the legal, academic and financial sectors to best suit business applications.
Using enterprise-oriented models such as these, an organization can layer its own data—for instance, historical information about customer interactions—over a foundation model. This process creates a more specific and effective series of AI tools. As these technologies “learn” over time, purpose-built AI models trained to complete specific tasks can continually improve and develop more capacity for specific tasks.
Marketing departments are well-positioned to take advantage of this technology, as customer communication and advertising generate vast amounts of data. Generative AI is particularly adept at analyzing unstructured data such as social media posts or chat communications.
Organizations might choose to integrate these tools in assorted ways, with varying degrees of human interaction and business-wide impact. While in recent years, prebuilt generative AI solutions have become nearly ubiquitous in marketing departments large and small, organizations are increasingly embracing custom models and large-scale digital transformations driven by AI. According to a recent report from the IBM® Institute for Business Value, more than half of CMOs say they are planning to build foundation models based on their company’s proprietary data.
Broadly, the degree to which a business adopts AI can be divided into three categories:
