ISSUE #1 - March 18, 2026
Welcome to Actually Useful. Most AI content is written from the outside looking in. Tool roundups, hype cycles, takes about what's coming next. I write from inside real work — managing campaigns, building internal tools, handling client relationships where the stakes are actual. Every week I share what's working, what isn't, and what I think you'll actually find useful.
Here's something I've noticed working inside a media company that runs advertising for thousands of publishers: most people who are genuinely good at their jobs spend a surprising chunk of their day doing things that don't require them to be good at their jobs.
The account manager copying numbers across three tabs to write a client update that says roughly the same thing every week. The editor tracking pipeline status in a spreadsheet nobody else updates.
None of that requires taste or judgment. It just requires time. And it takes enough of it to crowd out the work that actually matters. That's the opening AI creates: not replacing the craft, instead, absorbing the overhead.
📌 In this issue:
Where AI fits cleanly into media work
The mistake most people make when they try to automate
Where to start this week
— Chris
📥 THE GAP BETWEEN INFORMATION AND COMMUNICATION
Media organizations produce a lot of internal information that someone has to translate into something a human can act on. Performance reports. Campaign updates. Post-mortems.
The raw data exists. The context exists. What takes time is the translation. This is one of the highest-leverage places AI fits in media work because the output is repeatable and the inputs are structured. If you know what a good performance update looks like, you can teach AI to draft it. Not publish it. Draft it. The part that required you still requires you. The part that didn't no longer does.
Start with whatever you produce most often. One specific prompt, built around the structure you already use. Not a general one — a specific one, with the format you actually send.
🔍 THE RESEARCH THAT NEVER GETS DONE
There's a category of useful work in media that consistently gets deprioritized because the effort-to-output ratio feels bad in the moment. Competitive analysis. Audience research. Synthesizing what a client's past campaigns tell you before the renewal call.
Most of this work would improve outcomes if it got done. It usually doesn't. There are always more urgent things to do. Here's the pattern worth naming: the people who are best at the relationship side of media — the ones clients trust, the ones who always know what to say — are often the worst at this background work, because their time gets consumed by being good. The expertise and the overhead compete for the same hours.
AI changes that calculus. The research layer now happens in minutes. Work that was previously impractical becomes routine.
✏️ THE FIRST DRAFT PROBLEM
Writing is the core skill in media. Which is exactly why it's painful to spend writing energy on things that aren't really writing. Onboarding emails. Status updates. Proposal templates.
These aren't craft writing. They're functional writing — clear, organized, complete. Give AI the situation, the audience, the goal, and the tone. Edit what comes back rather than starting from nothing. You're editing a 70% draft instead of building from zero, and that difference compounds across a day
The same logic applies to analytical work. I used to manually review publisher performance data across hundreds of rows to build exclusion lists for campaigns. Now I feed the data to AI with a clear evaluation framework and get a tiered output in minutes. I still make the final calls. I'm just not spending an hour on work that doesn't require judgment before I get to the part that does.
The mistake most people make is trying to automate the wrong layer — trying to use AI on the parts of the job that are actually valuable, then wondering why the outputs feel hollow. Those parts require you. You can’t ask AI to make a judgement call, to replicate relationship context.
What surprised me when I started building this way: once I automated the data layer, the real constraint became obvious. The automation didn't eliminate the hard work. It exposed where the hard work actually lived.
Pick one repeatable output you produce regularly and build a specific prompt around it. The people who get the most out of AI stop asking "what can AI do?" and start asking "what do I do every week that I'm tired of doing?" I’ve found these two questions have a lot of overlap.
|
🛠️ THE TECH STACK beehiiv — Where Actually Useful is built and sent. Best platform for newsletter operators serious about growth. → 20% off first 3 months Wispr Flow — Every rough draft in this newsletter starts as dictation. → Try free Shiori — A beautiful read-later tool that actually holds what you saved. → Save your first thing |
🗞 IN THE NEWS
🔬 Anthropic measured which jobs AI is touching based on real usage data → LINK. Anthropic's own usage data puts computer programmers at 75% task coverage. Hiring for 22–25 year olds in exposed roles is already slowing — and this is still early.
✍ Sure you’ve read it by now but Matt Shumer’s thread on what’s actually happening in AI is worth your Saturday morning read → LINK. More about the pace, less about the hype. Shumer is someone who builds in this space daily describing the moment it stopped feeling incremental.
🔄 AI Show and Tell to #proj-AI-club to company mandate → LINK. Participating in beehiiv's AI adoption from the inside is one of the most exciting things I've been part of in my time here. The companies getting this right aren't forcing it — they're just making figuring it out together part of the job.
Thanks for reading!
If this was useful, forward it to one person who you know will do something with it.