Stop Being the Glue: How to Connect Isolated AI Workflows in Cowork
Your AI automations don't talk to each other. Here's how to connect AI workflows in Cowork with a 5-step framework and a live demo stitching scheduled tasks together.
One of my readers, Marc, is an advanced AI user who’s built a bunch of different automations and workflows for his business. He sent me a message a couple of weeks ago that’s been on my mind ever since:
“The challenge isn’t any single piece. It’s stitching it all together into something that runs without me babysitting every step.”
And that’s the problem, isn’t it?
We’re getting better at building individual automations. A Cowork scheduled task here, an n8n workflow there, a Claude Code routine you set up last week.
But your monthly report doesn't know your expense tracker just finished categorizing last week's receipts. Your lead magnet process doesn't know a new article is ready to be turned into a downloadable guide. Your client onboarding sequence doesn’t know you just signed a deal.
You’re the one who finishes step A and remembers to go start step B. The glue between your own automations. The human cron job stitching everything together because nothing connects them for you.
When people talk about AI productivity and building a “personal operating system”, they’re almost always thinking about one of two things:
On one side, task management. An AI that summarizes your day, tells you what to focus on, reminds you about deadlines. The "AI chief of staff" crowd.
On the other, deep technical orchestration. Multi-agent architectures, custom pipelines, code-heavy systems that most non-technical people will never build.
There’s a whole middle ground that nobody’s covering: the part where you already have working workflows, and you just need them to talk to each other without you being the middleman at every handoff.
Because if your workflows don’t talk to each other, you don’t have a personal operating system. You have islands.
That’s what I’m trying to figure out: how do I connect the things I’ve already built so work flows through them instead of getting stuck at every junction, waiting for me?
This article is the beginning of that. I believe there are multiple ways to stitch your workflows together, and the right one depends on where your work and automations already live. That’s why I want to show you more variations.
So today we’ll start with Cowork and how to set up scheduled tasks that talk to each other, so the output of one becomes the input of the next.
Since then I've built two of them, and which one fits depends on where your work already lives: if it's mostly in Cowork, a Cowork operating system that gives Claude a real memory of your whole working life; if you've moved into Claude Code, an agent that runs an entire workflow chain on its own, in the cloud, while I sleep. This piece is where it all started.
In today’s article:
Why your AI workflows need to work as a system: Why a pile of disconnected automations isn't a system. It's islands. What that's costing you, and the mental model that makes the mess visible.
How to build and connect your AI workflows: a 5-step framework 🔒: A framework for going from scattered AI automations to a connected system that runs without you. How to see your workflows clearly, spot the handoffs where you're just the glue, and figure out what still needs you.
How to stitch automated workflows in Cowork 🔒: The practical part. Two real Cowork scheduled tasks stitched together, with a shared database as the bridge. Exact prompts, connectors, the full setup. No code.
Cowork Scheduled Task Builder — a new Amplifiers prompt 🔒: Don't know where to start? This prompt walks you through designing your own chain of scheduled tasks, whether you're connecting existing ones or building from scratch.
Why your AI workflows need to work as a system
A system is a collection of repeatable processes that produce consistent results without you starting from scratch each time, all working toward a specific goal.
Think about a company’s customer support. A customer has a problem, reaches out, and gets it solved. A week later, same problem. Different person picks it up, gives a completely different solution, and this time it doesn’t get solved.
Same company, same problem, two different outcomes. That’s what it looks like without a system.
Now think about your AI workflows. If the output changes every time depending on which tool you opened, which chat you found, or whether you remembered to copy something between apps, you’re in the same spot. You don’t have a system. You have disconnected steps that sometimes produce something.
And when the steps needed to do only live in your head, three things happen:
You can’t step away. You’re the only one who knows how things connect. Everything stalls without you.
Every task takes more energy than it should. You’re not just doing the work. You’re deciding how the work gets done, every single time.
You can’t delegate. Not to a person, not to AI. There’s nothing repeatable that someone or something else could pick up and run.
And yet, it’s not that simple. A system that runs your business is made of multiple subsystems, and each subsystem is made of individual processes. It’s layers inside layers, and that’s what makes it feel overwhelming.
Systems, subsystems, and processes: how to map your AI work
At the top, you have your big systems. These are the major areas of your business: content, products, finance, client work.
Each system is made of subsystems. These are the specific functions inside each area. Inside your content system, for example, you might have newsletter production, a brand visual system, social media repurposing, and lead magnet creation as separate subsystems.
And each subsystem is made of processes. These are the individual steps that get the work done. Inside newsletter production, the processes might be: research and outline, write the draft, create visuals, final polish, publish and distribute.
Every business and individual runs on a few big systems like this and most of us are already doing all of this work, every day.
We just haven’t stopped to map it out, to see where we’re just passing things from one step to the next and where we’re the ones making the decisions.
In the next part of this article, I’ll walk you through how to think about your work as systems, a step-by-step framework for building them, and a full demo where I stitch separated workflows together in Cowork so they communicate with each other and produce work while you’re not there.
The goal: remove yourself from the middle. Upgrade to read the rest.