.svg)
Everybody knows AI will transform the world, from delivering scientific breakthroughs to driving cars and flying planes. Yet the path upward won’t be a straight line, and in the meantime, lots of businesses are struggling with more mundane applications. Namely, how to put LLMs and related applications to productive use.
To get specific, if AI is so revolutionary, why is it difficult for LLMs to deliver a high-quality draft of a press release or a strong marketing plan?
The simple reason is many AI tools are designed to dazzle with wide-ranging capabilities. Few do just one thing really well. Instead, platforms tend to offer full stack solutions, including LLMs, data analyses, content generation, etc.
Remember, these software companies are usually venture capital funded—often at valuations that demand creating massive new markets to return investors’ capital, let alone exit at large multiples.
The result: business users’ specific day-to-day needs aren’t being directly addressed.
The AI-Marcomms Paradox
It’s especially true in marketing and communications. Which is odd. After all, LLMs are content generators, and written and visual material is the fuel that runs successful campaigns. Yet our clients and partners consistently tell us that only between 10 and 25 percent of the output is useful. That’s a huge miss. What’s going on?
The answer is these applications are trying too hard to be smart.
By ingesting everything you can find on the open web, LLMs’ generative work mimics average content—not great work. It’s a softer version of the old Garbage In, Garbage Out (GIGO) problem, and it makes it hard for the LLMs to learn and deliver. You are what you eat.
The actual problems that need to be solved for marketing and comms pros are much simpler:
Content Management: house all your past announcements, reports, blog posts, etc., so you can find what you need when you want it.
Generating Custom Drafts: save time for lean in-house teams via first passes that use the right language in the right way.
Tracking Social Media Engagement: eliminate the drudgery of manually looking through comments, likes, followers, and gather all the relevant information about your audiences fast.
Managing First Party Data: sort your brand’s proprietary data to build lists, customer segments, etc.
You Don’t Need a Smarty-Pants LLM
Each one of these tasks doesn’t require an LLM to ingest and analyze everything on the open web. Rather, automation should focus solely on a business, its workstreams, and its content.
You don’t need a smarty-pants LLM that knows everything. You just need a tool that knows what matters to you. So when you evaluate a potential platform, don’t ask about everything an LLM or an AI platform can do. Instead, look to see if the AI is dumb enough to solve your problem.
