As we mentioned in our last post, Large Language Models are rapidly opening new possibilities for content generation that were unimaginable a few years ago. However, the hype around ChatGPT and similar models like Bard, Vicuna and Alpaca have set user expectations so high that people are quickly realizing they have some serious limitations.
What You Should Expect
LLMs aren’t factually reliable. You can read more about why in our last blog post, The Possibilities & Limitations of Large Language Models. So don’t expect LLMs to be experts in your field. But you can expect LLMs to jump start your content development into the editing & quality assurance phases and help you with the finishing touches.
Brainstorming New Approaches
Staring at a blank page can be intimidating. LLMs can quickly get you started and throw out a bunch of ideas. All you need to do is choose the best ideas of the bunch and refine.
Rough Drafts Based on Previous Content
Instead of developing new content from scratch, you can use LLMs to quickly generate new content based on old content, updated with topics of your choosing. For example, you can use an LLM to quickly develop a new product announcement, using old content as a template, but updated to focus on your new features. The copy won’t be perfect, but it will give you a big push in the right direction. Always update for accuracy and relevance.
Motiva Generator is a good example of this. You can, for example, provide a previously used subject line, and it will generate high quality new variations that you can use in your campaign.
Structural Outlines
Need help with the structural outline of an email or landing page? LLMs have you covered. LLMs can organize content into a wide variety of structures from simple listicles to more complex announcements. Think of outlining as developing a customized template for you to “fill in the blanks.”
Summarization & Scaling
You can use LLMs to develop content for marketing campaigns based on, for example, an existing landing page. LLMs can summarize the landing page to get the main idea and generate a variety of different content using that summary as its guide, including emails, social posts, press releases and webinar outlines. LLMs are very effective at scaling and repurposing existing content. But always remember to check the output before you publish.
Clarity & Accessibility
Complex ideas can be difficult to communicate. If you’re having a hard time with concision, clarity or making your ideas accessible to people outside your field, then you can use LLMs to dramatically increase your productivity. Just make sure the output isn’t changed so much to lose the nuances of your statements.
Targeting with New Stylistic Approaches
Use LLMs to adapt the style or tone of existing content to resonate more with specific audiences. But be careful. LLMs were trained on the internet, which reflects human biases, assumptions and stereotypes. LLMs can very easily bring those problems into its output. Always check generative AI content to ensure it truly represents your brand.
Simplistic Data Enrichment
Data enrichment starts to hit the limit of what LLMs are capable of doing. Most enrichment tasks rely on access to timely, up-to-date sources, which standard LLMs don’t have. Don’t use LLMs to fetch email addresses, websites, LinkedIn profiles, or to somehow know what industry a company belongs in. It’s not there yet.
But with the right mix of existing data, LLMs can extrapolate simplistic meta tags that could be useful in certain use cases. This gives you the potential to view your existing data in new ways to uncover valuable strategic insights.
