Most people build their first AI agent as a morning briefing. This is about why that one rarely earns its keep, and the question that tells a business owner whether the next one will.
The first agent I built sent me a morning briefing. Halifax weather, my calendar, three priorities pulled out of Asana, some AI news, anything in my inbox touching work already in motion. It showed up on time every day and did exactly what I'd asked it to do.
I wrote about it in June, mostly because it spent two days making the news up and I didn't catch it.
What I didn't write about is that I stopped reading it. Not slowed down, stopped. It would land every morning and I'd leave it sitting there. Eventually I deleted it. I'd also run a business for years without anyone briefing me about anything.
I felt bad about it for months. Guilty. I'd built the thing myself, it worked, it was sitting there every morning doing its job, and I wasn't holding up my side of the deal. It was screaming look at me and I ignored it. That's how it felt.
I've listened to Successful Idiots since the first episode. It's my fave AI podcast, I recommend it to friends and other business owners, and I often listen to episodes more than once so I can get as much out of it as I can.
Joe Downs is one of the hosts. He runs self-storage companies and on the show he's the one who goes and tries things. Peter Swain is the AI side of it.
In June, Joe told everyone listening that they needed an agent. It wasn't that difficult, it would change their life, and he'd gone and built one. I built mine the same month (we're both in Peter Swain's AI mastermind).
Ten weeks later he's back on, and the message he'd had to type into his own agent was stop running everything except the morning briefing. A few days after that he told it to stop doing that too.
So it wasn't just me, which was reassuring.
How and why did we start with morning briefings? Simply. The main Hermes tutorial sells itself as build a free AI morning briefing bot that emails you daily news, and its getting started walkthrough is a daily briefing on WhatsApp. Everything sitting around it is some flavour of chief of staff. One runs somebody's inbox. One is somebody's day one.
Nobody actually tells you to build a briefing first. It's just the easy one to picture. You can explain it in a sentence and people get it straight away.
Whether it's worth building is a whole other question. The tutorial isn't going to ask it.
The guilt was misplaced. I'd built the wrong thing.
The agent that I'm still using is boring. Yup. Boring. It pulls in data from Meta, CRMs and Google Analytics every day and puts them beside each other. It doesn't tell me what the data means or what to do about it. Yup. I said it was boring.
And yes, I know. That's a dashboard.
Mostly it is. Same sources, same numbers, and a decent dashboard would lay them out better and cost less. It's true.
How it's different: it literally comes to me rather than me having to go and get it. That sounds like the difference and it isn't, because scheduled reports have been doing that for years. I can reply and ask why something moved, which a chart can't. Real, but small.
I could make it smarter. Tell it what to flag, what to ignore, what a bad week looks like. I won't.
So mine stays boring. It fetches. I decide.
The question was never whether it's a dashboard. It's whether it replaced work or just showed me things.
The briefing showed me things. If I want to know the weather, I just look out the window. My calendar was already open. The news I'd have run into anyway.
The boring one saves me an hour of opening tabs and copying numbers into a sheet. Every day.
None of that is really about AI.
So that's the only question I ask before building anything now. What does this take off me?
If the honest answer is that it shows me things, it's a dashboard, and I should go and build a dashboard. They're cheaper, they're better at it, and nobody ever feels guilty about not looking at one.