how to use AI agents to run your marketing team (the playbook)
文章目录
i haven’t built a million-dollar app yet. but i’ve worked with many of the top AI companies over the last six months, and one thing is obvious: they move so fucking fast.
here’s what that speed can look like.
an idea comes up in a meeting. an agent catches it, finds examples, and sends a different angle to each person who could work on it.
someone replies “yes” in Slack. that triggers Devin to build the first version: a demo, a page, or a piece of content. the team reviews it and ships.
an idea that would have sat in meeting notes is already being built. that’s the speed i’m talking about.
and i think one growth engineer with good taste can set up this whole system.
here’s the playbook.
start with the questions customers already ask
>before making content, find out what people are confused about.
>look through sales calls, support messages, onboarding notes, and the questions people ask before they buy. keep their exact words. “how long does setup take?” is a better starting point than a broad topic like “productivity.”
>put each question in a simple table: what they asked, who asked, where it came from, your current answer, and whether that answer helped.
once a week, look for repeats.
this is a good one btw, check it out :
make content a system the whole team can use
give the writing agent your product docs, real demos, customer questions, founder opinions, and examples of posts your team likes.
then make a few custom skills for the formats you repeat.
one turns a use case into a demo script. one answers a customer question as an article. one turns a product change into a short post.
now make the system useful to everyone.
each morning, send every teammate one idea in Slack based on what they actually know.
the engineer gets a question about a setup step they fixed. the founder gets an opinion from a meeting that deserves a fuller explanation. the person talking to customers gets a common objection to answer.
they can write it, change it, or skip it. if everyone keeps skipping, improve the ideas.
one good answer can become a post, an article, a help page, a demo, and a short video. start with the answer, then change the format.
also design is easier and better.
ship faster
this is where most teams get stuck.
>connect Devin to the marketing channel. you can call Devin in a Slack thread, and its automations can respond to channel activity. start with an explicit tag or reaction so every passing thought does not become a task.
>it can gather the context, draft the first version, and report progress in the thread. the product owner checks whether the flow is real. the growth engineer decides whether the story works. then the team publishes it.
the measure is simple: how long does it take a good idea to become approved work?
4. turn one good demo into several videos
>make a short clip that shows the result first
>@Remotion can help make repeated visual elements easier to produce. >@InVideoofficial can help explore ad concepts. check every edit against the actual product.
you don’t need that step to start. a clear recording of a real result will do more than a complicated editing pipeline around a weak demo.
also here is few articles on making cool demos using claude and devin :
https://x.com/dabit3/status/2104576471515779520?s=20 https://x.com/0xMovez/status/2104216919033192746?s=20
follow these ppl they post cool stuff about ai and marketing :
@fuckgrowth @alexcooldev @mikenevermiss @denohawari
