The short version: AI can handle 60 to 70 percent of the mechanical work in social media (writing captions, scheduling posts, pulling analytics), but the human call on strategy, tone and what matters to your audience is non-negotiable. I have stopped pretending otherwise.
Why I stopped hand-writing captions in 2024 and haven't looked back
I used to sit at my desk on Sunday nights for two hours, writing captions for the week ahead. Real captions. Long ones. Thoughtful ones. The kind where I'd delete a line five times because it didn't land right. I was 52, I had a business to run, and I was spending 104 hours a year on caption writing alone.
In April 2024, I told myself: this stops. Not because I don't care about captions (I care deeply) but because AI can write a decent first draft in 90 seconds, and I can edit it in 2 minutes. That's 94 minutes of my time versus the original 120. Over a year, that's 26 hours back.
Here's how the actual workflow goes now: I write a bullet-point brief for the post. One or two sentences. What it's about. Who needs to hear it. What I want them to do next. I paste that into Claude or ChatGPT with the instruction "Write this in my voice, British, blunt, specific, warm, no filler." The AI gives me 2 or 3 options. I pick one, edit the opening line and the call-to-action by hand, and post it. Total time: 4 minutes per caption instead of 12.
The honest part nobody mentions: the first five times you do this, it feels wrong. You feel like you're cheating. You aren't. You're using a tool the same way a photographer uses Lightroom instead of a darkroom.
The scheduling piece is where most people get lazy
Scheduling is the easiest part of AI-powered social media, and that's why most people do it badly. They throw content at the wall four times a day and hope something sticks.
Here's what I do: I use a spreadsheet (Google Sheets, nothing fancy) to map out which platforms get which content and when. Monday through Friday, one post per platform. Wednesday and Friday get something tied to an offer or a free resource because that's when my audience engages with actionable stuff. Saturday and Sunday get something personal or behind-the-scenes because attention is lower but context is warmer.
Then I batch-write a week of captions at once (Monday morning, 30 minutes) using the method I described above. I paste them into my scheduling tool alongside the graphics, set the time zones (crucial if you work across continents, which I do), and I'm done. No thinking about it for seven days.
The thing AI doesn't handle well: cross-platform formatting. A caption that works on LinkedIn is too long for X. A hook that works on Instagram is too casual for Facebook. I used to use a tool that claimed to auto-resize captions, and it turned my voice into beige soup. Now I write one caption and manually edit it for platform. Three minutes. Non-negotiable.
Analytics is where AI earns its money
This is the part that changed my business. Not just my social media posting, but my actual strategy.
Every Friday at 2pm, I pull my analytics from LinkedIn, X, Instagram and YouTube into a single Google Sheet using simple formulas (no AI needed for this part, just basic =IMPORTRANGE functions). Then I paste the numbers into Claude with this prompt: "Which three posts drove the most engagement this week? What do they have in common? What should I write about next week based on what worked?"
Last month, the AI spotted that my posts about "rebuilding after failure" got 3.2x more shares than my how-to content. I had not noticed this because I was too busy writing. Once I saw the pattern, I restructured my content plan. In four weeks, my LinkedIn engagement went from an average of 14 interactions per post to 31. That's a 121 percent increase off one insight.
Here's the step-by-step I use now:
- Export analytics from each platform (LinkedIn: native analytics dashboard, export as CSV; X: use Twitter Analytics; Instagram: Meta Business Suite)
- Combine all data in one Google Sheet with columns for date, platform, post content, impressions, clicks, saves, shares
- Flag posts that scored in the top 25 percent for that week
- Paste the data and the flagged content into an AI tool with a specific question (what worked, what's the pattern, what should I test next)
- Read the output, disagree with half of it, and decide what you want to test
- Move forward only on the insights that match your business goals, not the algorithm's goals
The last point is critical. AI will tell you to post more cat videos if your cat video got 500 views. You know whether that's a good use of your platform. Trust yourself here.
Ideation is not where AI shines
I want to be blunt: AI is bad at original ideas. It's excellent at remixing patterns it has seen before.
I spent three weeks in 2024 asking ChatGPT to "generate 20 social media post ideas for a marketing consultant" and then trying to use them. They were fine. They were also exactly like 10,000 other marketing consultant posts. I had zero ideas that felt like mine.
What changed: I use AI for execution, not invention. I open a document every Sunday and I brain-dump 5 to 7 ideas that came to me during the week. Things I noticed. Mistakes I made. Questions clients asked. Books I read. This takes 15 minutes. It's the only time I'm not using AI.
Then, if I get stuck on how to frame one of those ideas, I ask AI: "Here's my thought: [idea]. How would you write an opening line that's surprising?" That's a different question. That works.
The technology stack I recommend
I've tried dozens of social media tools. I'm drawn to the ones that let you write in your own voice, not the ones that promise to automate everything.
For scheduling: Google Sheets plus a scheduling platform that integrates with multiple networks (Hootsuite, Buffer, Later). I prefer platforms with a good native mobile app because half the time I'm writing captions while on the train. Costs are typically 30 to 50 pounds per month if you're managing 4 to 6 accounts.
For writing: I use Claude through the paid subscription (20 pounds monthly) and ChatGPT Plus (20 dollars monthly). I prefer Claude for long captions and ChatGPT for snappy one-liners. Yes, I pay for both. The 5 to 7 hours I save every month is worth 35 pounds to me.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
For analytics: I test a lot of AI tools for analytics, but honestly, a spreadsheet plus 10 minutes with Claude is faster than learning another dashboard. Your analytics aren't the insight. Your question about the analytics is. Let AI answer the question, not visualize the data.
What I was wrong about
I used to believe that full automation was the goal. Post 20 times a week, no overlap, perfect timing, maximum reach. I automated myself into irrelevance. My engagement dropped. My voice disappeared.
What works: fewer, better posts. Consistent voice. Time to respond to comments by hand (I spend 20 minutes daily on this). A real person in the comments, not an AI bot.
AI handles the repetitive part. You handle the human part. The business impact comes from the human part.
The real cost calculation
Let me give you the actual numbers so you can decide if this is worth it for you.
Time saved per week using AI for social media management: 4 to 5 hours (assuming you're managing 3 to 5 platforms, posting 3 to 5 times per week).
Cost per month: approximately 40 to 80 pounds in AI subscriptions plus a scheduling tool at 30 to 50 pounds. Total: 70 to 130 pounds per month, or 840 to 1,560 pounds per year.
What you do with those 4 to 5 hours instead: client work, strategy, sales calls, rest. At the rates most consultants charge (80 to 200 pounds per hour), that's 320 to 1,000 pounds in reclaimed value per week. Over a year, you're looking at 16,640 to 52,000 pounds in opportunity cost recovered.
It pays for itself in week one if you're going to fill that time with billable work. It doesn't pay for itself if you're going to use it to scroll more social media yourself.
One thing I wish I'd done sooner
I spent 18 months thinking I had to use the same AI tool for everything. Ideation, writing, analytics, scheduling. One integrated platform, I thought, would be cleaner.
It wasn't. It was slower and more frustrating because no single tool was best-in-class at all of those things.
What I do now: best tool for each job. Write in Claude. Analyse in Claude. Schedule in Hootsuite. Track ideas in Google Docs. It's less elegant. It works better. That's the move.
Frequently asked questions
How often should I post if I'm using AI to manage my social media?
Quality over frequency. I post 3 to 4 times per week per platform, and each post is a real thought from my head plus 5 minutes of AI-assisted editing. Posts from my audience are more likely to save or share a post that feels like it came from a real person, not an algorithm. If you're posting 20 times a week and none of them get traction, you're wasting both your time and the AI's.
Will my audience know I'm using AI to write my captions?
They will if your voice disappears. They won't if you edit the output and keep your actual personality in it. AI writes in a baseline professional tone. I edit for British, blunt, one-line punches. That edit is where the human voice comes back. Do the edit.
What's the biggest mistake people make when they start using AI for social media?
They stop thinking. They turn on a tool, set it to post automatically, and check back in three months wondering why nobody engaged with anything. AI is a force multiplier for your strategy, not a replacement for having one. If you don't have a content strategy before you automate, you'll just automate the wrong things faster.
Should I use one AI tool or multiple tools for social media management?
Multiple tools, each specialised at one job. I don't use a single "AI social media platform" because no single platform is best at writing, scheduling, and analytics. It's slower to learn multiple tools for the first week, then faster forever. Specialised beats integrated every time.
For the bigger picture, see my full guide to AI marketing.
Related reading: Careers for Mothers at Home: What Pays and What's Just Noise and What a Fractional CMO Does and When Your Small Business Should Hire One.
Related: how to make money on Instagram with just 1k followers.
Related: work with Lilach on AI strategy and implementation.
