Last week was a big week of conversations about AI in communications, and now everyone seems to be talking about needing AI agents on their comms team. Personally, the AI FOMO feels overwhelming, so I did some research that resulted in feelings of more confusion and then some clarity…. if that makes any sense.
But before we dive in, here’s a quick VERY BASIC introduction to LLMs and agents.
ChatGPT (from OpenAI) and Claude (from Anthropic) are both LLMs (Large Language Models) and are the underlying AI that understands and generates text. They are also the most commonly used type of AI by comms people for writing press releases, summarizing documents, etc. LLMs learn by reading millions of books, websites, and articles. Think of it as a highly sophisticated text processor that can answer questions, generate content, analyze information, and have conversations by predicting what the next word logically makes sense based on all those millions of books, websites, and articles it has read.
But it's essentially stateless and passive. It responds to what you give it at that moment. Each time you start a new chat, it's basically starting fresh (unless it has memory features turned on). I’m going to do a deeper dive on this and how it relates to comms at another time, so don’t come for me with my simplistic overview here.
An agent is an LLM that's been given the ability to take actions autonomously. It can:
Use tools (search the web, access databases, run code, create files)
Make decisions about which tools to use and when
Break down complex tasks into steps
Execute multi-step workflows without you micromanaging each step
So in practice, an Agent could be told to "build a narrative and messaging launch framework for your Series B announcement," and it will research competitive positioning, interview you about key messages, create the narrative doc, draft supporting executive quotes, AND create a content calendar with blog topics.
OK so, people are claiming that agents will be the game-changers in our industry because they don't just create content, they think strategically about the full workflow.
I was initially excited about the idea of automating my work too. But here’s where the aforementioned confusion comes in: after I sat with it a while, I started to wonder how this type of technology fits into our industry 1-2 years from now.
Mainly, what are the ramifications of automating the type of work that defines categories, shifts markets, builds communities, and facilitates understanding of culture?
I kept coming back to this question of ‘in an industry that needs to stay human in order to deliver impact, why are we so focused on building something that takes away our humanity and human judgment?’
I’m sorry guys, I couldn’t help myself.
Now, I'm not saying we don't use AI. It's amazing for scenario planning, message testing, measurement, and data analysis. AI won't improve the way we do comms, but it does give us an opportunity to be more measurable, and therefore improving the understanding of what we do so we can protect our industry from being first on the lay-off chopping block (like we always seem to be).
I guess what I’m trying to say is: just because we all have a hammer doesn't mean we should all build a house.
So, this is where the clarity comes in.
Here are some deep issues that I have with Agents on Comms Teams
1. Agents don't "think strategically." They pattern-match at scale, which is actively dangerous
Calling an agent's output "strategic thinking" is fundamentally dishonest. Agents process patterns from training data—they're sophisticated autocomplete, not strategists. When you ask an agent to "research competitive positioning and build a narrative framework," it's stitching together common patterns it's seen before.
This results in every Series B announcement starting to sound the same. Every thought leadership piece hits the same tired beats. You think you're getting strategic differentiation, but you're actually getting the statistical average of what worked before, which means your narrative is inherently derivative and commodified.
Real strategic communications requires understanding market dynamics that haven't been documented yet, reading subtext in executive conversations, and making intuitive leaps based on years of pattern recognition in human behavior. An agent gave you a "narrative framework" based on what thousands of other Series B companies said and as a result you’ve just automated yourself into the middle of the pack.
2. Workflow automation creates catastrophic skill atrophy
What actually happens when comms pros start relying on agents for "the full workflow" is that they stop developing the judgment that separates good communicators from bad ones.
A junior comms person who uses an agent to "research competitors, draft narratives, and create content calendars" never learns how to identify which competitive intel actually matters versus noise, recognize when a narrative angle will backfire in six months, understand why message sequencing matters for audience psychology, or catch the subtle brand voice inconsistencies that erode trust.
You're not getting efficiency, you're creating a generation of comms professionals who can prompt AI but can't actually think about communications strategy. When the agent gets it wrong (and it will), they won't have the expertise to catch it.
There’s already evidence proving this point in content marketing. Companies using AI content tools at scale are producing more content that performs worse because no one on the team can identify quality anymore.
3. Agents optimize for volume and mediocrity, which directly undermines what makes communications valuable
The entire value proposition of communications is breaking through noise with something that matters. Agents do the opposite by industrializing the creation of more noise.
When an agent can generate a thought leadership strategy, 50 blog topics, and a 6-month content calendar in 20 minutes, what do companies do? They execute all of it. More content, more campaigns, and more "strategic narratives" that are all optimized to sound good but say nothing memorable.
The real problem is that communications programs work when it is hard. When crafting the right message requires deep thinking, iteration, and judgment. Agents don't make that process better, they make it disappear. You get 10X the output at 1/10th the strategic value, and you've just convinced yourself you're being more effective because you shipped more content.
Your CEO's "thought leadership" is now indistinguishable from every other CEO's because you all used agents trained on the same corpus of mediocre business writing. Congratulations, you've automated away the only thing that made your work valuable: genuine human insight.
Breaking it all down
I hate saying this out loud, but comms kinda runs on vibes, and we can't make a vibe agentic.
The magic of what we do isn't in the workflows. It's in the thousands of micro-decisions we make based on context, intuition, and experience. It's in reading the room. It's in knowing when to push and when to pull back. It's in understanding that the perfect message delivered at the wrong moment is worthless.
You can't automate judgment. You can't systematize taste. And you definitely can't agent-ify the ability to make executives feel heard while also steering them toward better thinking.
So no, you probably don't need an AI agent for your comms team. You need comms professionals who understand how to use AI as a tool while maintaining the strategic judgment and human insight that makes communications work in the first place.
I believe that the future of communications isn't about who builds the best agent. It's about who maintains their humanity in an increasingly automated world. And that's the hardest work of all.