AI Agents & The Future Of Work

Every morning at 7 AM, I get a brief on the world of digital marketing from Claude Co-work & what’s changed overnight.

Tools:

• Claude Cowork (built into the Claude Desktop app — Mac or Windows) - here is how to install this

• Claude Dispatch (to trigger and check results from my phone)

• Pro plan ($20/month)

Actions:

1. Create a folder called /digital-marketing-research on your desktop

2. Add an AGENT.md file that defines exactly what the agent researches:

   
   To set this up, go to:
   
   TextEdit on Mac or terminal & paste in the text below, or edit for your vertical:
   
   Research Agent

   
   You are a digital marketing research specialist. Your job is one thing:
   Find what's new, relevant, and actionable in digital marketing on a daily basis:
   
   What you research

• SEO and algorithm changes (Google, Bing)

• Paid media trends (Meta, Google Ads, TikTok)

• Social media platform changes

• Email and content marketing shifts

• Tools and AI entering the marketing space
  
  Output:

  Write your findings to DAILY-INTEL.md in this folder.
  Structure: Date → Top 5 findings → Source links → Why it matters for marketers.
  
  Principles:

• No speculation. Every claim has a source & site link.

• Signal over noise. Skip minor updates, flag only meaningful shifts.

• The sources need to be within the last two weeks

• If uncertain, mark it [UNVERIFIED]

3. Add a DAILY-INTEL.md file (starts empty — this is where findings are written)

4. In Cowork's Global Instructions, tell Claude to always read AGENT.md first and append dated findings to DAILY-INTEL.md

5. Type /schedule in Cowork, set a daily time (I use 7 am), and write the task prompt:

"Read AGENT.md, research the latest digital marketing trends, write a dated summary to DAILY-INTEL.md, then post a formatted bullet-point digest"

6. Add the Slack connector: Settings → Connectors → + → Slack → authenticate

7. Set up Dispatch for mobile access (60-second setup) so you can trigger or check results from your phone

You can then ask Claude to make a summary for social posts & just like that, you have a skeleton post you can share on socials.

Check out my post from earlier in the week.

How I Work

In a follow-up to my post last week.

Another way I work, which I’ve not heard others talk about, is my goal for each workday.

Some folks choose work based on vibes. Others use to-do lists. And a few old heads (the ones who’ve figured out all the productivity stuff is rubbish) focus on doing “the most important thing”.

I don’t do any of that. Instead, every day, my entire focus, for each day, is to make one good decision.

I remind myself of that every morning before I start work, and I journal about my decisions in the evening after I’ve finished work.

If I make a good decision, I’m happy. If not, I’ve failed.

I started doing this a year ago now, and the difference in my impact on our businesses has been profound.

I realised it was time to make this change when the size of my team, our free cash flow from operations and other assets like IP had gotten so large that my marginal gain from a good decision was significantly outweighing my marginal gain from a unit of work.

In other words, my decision leverage exceeded my work leverage.

It’s not that I don’t do work anymore, I still do a heap of it.

It’s just that it comes second to me, thinking deeply about problems and trying to make good decisions to move forward.

I’ve gone from “operator” as business wankers like to call themselves to “conductor”.

My good decision today was buying a set of dumbbells from another gym for 1/5 of the market price. It saves us money and makes a meaningful improvement to our gym.

Nothing special, but long-term, the impact of that decision likely beats any work I’ve done or could have done today.

Other good decisions I’ve made this week are:

- Moving forward with a project for Gravy
- How to hire our next assistant gym manager
- Promoting our gym operations manager
- Setting a new compensation package for both of those roles

I don’t have an exact definition of a good decision; it’s a tad arbitrary. But roughly speaking a good decision is one that employs leverage and compounds for a long-time.

If you’re a one-man band or close to that size, this approach is probably useless, and you’re better off focusing on getting shit done. But it could be worth considering as your leverage increases.