Photo by Alexandre Debiève on Unsplash
Back in early 2025, my team and I dove headfirst into AI. Claude 5 had just dropped, GPT-5 was rolling out, and the outputs were... honestly, jaw-dropping. Suddenly, I could get deep strategy docs, smart blog drafts, even campaign outlines that felt eerily close to what a senior strategist might write.
It felt like magic.
And then reality hit.
The Honeymoon with Prompts
At first, prompting felt like the future. I would type a request, and boom — a draft blog post, an outline for a campaign, even some positioning ideas.
Everyone on my team started doing the same. We had Claude windows open, GPT tabs running, Gemini threads scattered across browsers. It felt exciting. It felt fast.
But here is the thing: everyone was doing it differently. One person had clever prompt formulas. Another winged it. Some outputs looked sharp, others missed the mark entirely.
We tried to fix it with a shared Notion page — a little prompt library. Nobody used it. Dead on arrival.
And when you zoomed out, the cracks were obvious:
Skill gaps widened. Some people became prompt magicians, others just gave up.
Outputs lost credibility. I would see drafts with completely different messaging or tone.
Context disappeared. Our ICPs, positioning docs, research — all lived in a forgotten Notion file instead of being baked into outputs.
Models kept changing. One week Claude 5 was best, then GPT-5 caught up, then a new model pulled ahead. We were constantly chasing the right tool.
And meanwhile, the chats piled up. Thousands of them. Scattered across tools. Unsearchable. Impossible to manage.
The promise of speed was there... but the operational cost was enormous.
Related: 8 Atlas Prompts Every Founder Should Try
The Moment It Broke
The turning point came when we needed a press release. Normally, that would have taken a few hours, back-and-forth reviews, and plenty of coffee.
Instead, we tried something new: an early agentic workflow we had built. It pulled in our context, followed the PR structure step by step, and produced a draft in three minutes.
It was not just fast — it was right. Structured, consistent, and 90% production-ready.
That is when it clicked: prompts alone are not enough. You need workflows.
Why Prompts Are Not Marketing
Marketing is not just words on a page. It is systems thinking.
If you are building an SEO strategy, for example, you cannot just say write me a plan. You need to provide:
- Your ICPs
- Your budget
- Competitors
- Positioning and messaging
- Business goals
That is too much for one prompt. You need a flow — a logical sequence where each step builds on the same.
Same with a content strategy. With prompts, it is weeks of messy back-and-forth. With workflows, it is half a day: context in, plan out, review, done.
See also: AI Agents Unleashed: Blockchain’s New Power Players
What I Learned, The Hard Way
Prompts are personal. Everyone uses them differently. That means outputs are inconsistent and unscalable.
Teams need structure. Without workflows, context gets lost and trust breaks down.
AI moves too fast. Models change monthly. Maintaining good outputs without a system is impossible.
Workflows free people. Once we moved to agent flows, my team spent less time prompting and more time thinking, reviewing, and being creative.
Where I Landed
Today, I still use prompts — for quick ideas, a bit of brainstorming, or a one-off draft. But when it comes to serious marketing work, I do not trust prompts alone anymore.
I trust workflows.
Because in the end, marketing is not about the perfect prompt. It is about:
- Repeatable processes
- Reliable outputs
- Giving your team the space to do what only humans can: strategy, creativity, empathy
Prompts gave me speed.
Workflows gave me confidence.
And that is the day I stopped believing in AI prompts.
Thanks for reading! This post is public so feel free to share it.
