Using AI as a Social Media Manager: the three real barriers and how to get past them

«Yes, but not really.» That's what most Social Media Managers said in the Large Brands study by Marketing Paradise when asked: "Are you using AI in your social media work?"

2026.
«Yes, but not really.»
Why?

Maybe fear that AI will take their job. Maybe the mountain of work that leaves no room to innovate. Maybe plain distrust of the technology.

I've been through all three.

So maybe, by sharing my own experience, we can find some answers.

Why I didn't trust AI at first — and what changed my mind

I get it. After my first output from AI (ChatGPT), my reaction was: «Is this supposed to be the revolutionary change everyone's talking about?». I gave up.

A few days later, I tried again after a very brief look into prompting. «Okay, I get the hype now» was the next thought. Followed by a sincere «but I do it better — it missed so many things».

I came back a week later. I treated it like a new team member (and wrote an article about it on this very blog), its answers got better. I started to like it.

Next step: personalize the experience. I started building custom GPTs almost as a game (here's a fun story about that) and jumped on the wave. My thinking shifted to the now slightly tired «AI first» — better translated as «this is taking forever, isn't there an AI that does it faster?»

⚠️ Disclaimer.
Thinking is not «this is taking forever, isn't there an AI that does it faster?»
Making strategic decisions is not «this is taking forever, isn't there an AI that does it faster?»
Just putting that out there.

Moving on.
I tested other AI tools: Claude, Gemini, NotebookLM, Veed, ElevenLabs, Perplexity, and a bunch of mini AI apps I can no longer remember the names of.

I dropped some for being too expensive. Kept others that, despite the cost, save me more than 50% of the time on a given task.

I'm very analytical by nature. So yes, I time myself, calculate, and compare. If you want to know whether an AI tool is actually worth it — I'd recommend doing the same.

My initial distrust in the technology became a trust that is anything but blind — conscious and earned. The only way to get there, I think, is obsessive iteration. Which brings me to the next point.

How I found time to experiment with AI without burning out

We've all been there. That LinkedIn «I built this impressive thing with AI in 3 hours» post — and me wondering where on earth they found that time in their workday (or personal life).

Which immediately collides with «but I also want to know if I can build impressive things with AI. Can't I do it in 10 minutes?». No. I'm sorry.

But here's what I did: every Thursday, 1 hour. From 8:30 to 9:30, after the daily task checklist but before the inbox and Slack explode.

I opened my Notion «Learning AI». And whatever I tried, I wrote down. What worked and what didn't. Which tool, what use case, what goal, what result (in a table, obviously).

New ideas came naturally, and I'd note those too. Everything that worked got developed the following Tuesday morning — one more hour, then I'd stop.

The time I saved by automating things with AI, I reinvested in new experiments, until the whole process became part of my daily workflow. I no longer block calendar time for trial and error — AI is just part of how I plan my tasks.

Not going to lie: sometimes innovation slows and the experiments happen less often. But the actual use of AI never stops.

There isn't a day I don't use Gemini to generate a brand-style image for social media; or Claude to help me with article SEO (yes, including the headings on this one); or Welov's Insights AI to qualitatively review posts; or my custom GPT to brainstorm; or Veed to add subtitles to videos; or save a post about AI to check later.

And with all this — aren't we replacing part of our «working self» with a machine? Let's get to the last point.

Will AI replace the Social Media Manager?

«Someone still has to write the prompts», they'll say. And you'll think: there are already AIs that specialize in prompting.

«Someone still has to develop and fix it», they'll say. And you'll think: Claude Code is already writing and improving code.

What I'd say is: if you don't know what you need, a prompt won't save you. And if you don't know anything about code, the output will be a mess.

Human intelligence is still necessary. So let's not sideline it.

To prove the point — do you think everything I've described in this article could have happened without human intelligence? An intelligence that knows how to discern, decide, and frame new challenges — whether the solution ends up involving AI or not.