Full Steam 02: Claude Cowork explained
Full Steam is an ongoing conversation about the craft of B2B marketing. In each episode, experts from Iron Horse explore a single topic in depth, giving you practical takeaways and a fresh perspective on the challenges you're navigating every day.
Most of us have used AI as a smarter search engine. But now marketers are using it to build tools that automate Monday morning tasks, connect marketing agencies’ 20+ systems, and make their daily workflows addictive to improve.
In this episode of Full Steam, Iron Horse senior content strategist Alex Jonathan Brown sits down with Iron Horse chief product and innovation officer Samir Mehta to talk honestly about agentic AI.
They discuss how they personally use Claude Cowork to build ADHD-friendly coding lessons, automatically block off time for focused work, and synthesize more data than their brains could handle alone. Samir also shares that the best way to learn where to use agentic AI … is to ask your favorite AI.
Featuring
- AI eliminates the data science request. Connecting the dots across multiple platforms and data sets used to be a slog. Now you can aggregate and interrogate data in seconds, not weeks.
- The goal is to just start building. Identify a personally exciting goal, or even an area you want to be more lazy. The cost of building is cheap, and you’ll fuel more ideas just by getting started.
- With agentic AI, persistence pays off. AI is learning from you at the same time you’re testing it out. The more you use it, the more it adapts to your needs, and the better results you’ll see.
How are B2B marketers actually using AI agents?
B2B marketers are using tools like Claude Cowork to connect data across systems they've never been able to cross-reference before.
Samir pointed out that creating a report used to require filing a data science request, waiting two to four weeks, and hoping the question was still relevant by the time you heard back. Now marketers can simply analyze the data themselves.
On the automation side, the Iron Horse team is using agentic AI in surprisingly practical ways. For instance, they’re scraping competitor ebooks automatically so they don't have to fill out forms, and building interactive, 15-minute ADHD-friendly lessons to teach themselves how to code.
What’s a good first project to try with Claude Cowork that actually saves time?
Samir advised that before building anything, try asking Claude where you should start. If you've been using any AI assistant for a few months, it already knows what you do, what you repeat, and what's been driving you nuts. Let it generate the list, then decide on a project that would actually change your day.
Alex shared an example of how he built a tool to simplify his morning planning. He created a skill that checks his task list, calculates remaining time per project, cross-references his calendar, and blocks focus time automatically.
The first version wasn't perfect: It scheduled focus time for hours that had already passed. But it got him into the process of catching the error, taking it back to Claude Cowork, and improving things from there.
What's the difference between Claude Code and Claude Cowork for non-technical people?
As Samir explained it, what Claude Code did for software engineers, Claude Cowork is doing for knowledge workers. Code requires you to think like a programmer: working in a terminal, writing instructions, and building software. Cowork meets you where you already work: in documents, calendars, task lists, and email.
Samir and Alex also discussed what it’s like hitting a wall with Claude Code. As a marketer, you might be curious enough to want to use it, but then struggle to get over the hump of actually building a tool. Cowork removes that wall, letting you create automations, workflows, analyses, and apps without having to think like an engineer.