What I Learned Running My Business for 30 Days With an AI Chief of Staff
I run a small software business. 4 people, not profitable yet, growing. For 30 days I ran every strategic decision through Alex, our AI chief of staff, before acting. Here's what I learned.
Day 1: The intake is the whole product
Four questions. What the business is, what the goal is, what the bottleneck is, how much time I have per week. Took seven minutes. By the end I already had a clearer picture of the quarter than I'd had in six weeks of running it.
The intake isn't onboarding. It's the product.
Day 3: Alex pushed back for the first time
I told Alex my bottleneck was lead flow. Alex asked: what's your MQL → SQL conversion rate? I didn't know. Alex said "you probably don't have a lead flow problem. You have a qualification problem. Go measure."
I went and measured. Alex was right. I'd have spent the next three weeks buying ads for a problem that wasn't real.
Score: 1 consultant, saved.
Day 8: The weekly review is the magic
Friday afternoon. I opened a session. Alex asked what I'd committed to on Monday. I'd committed to three things. I shipped two. Alex asked about the third.
"Capacity or commitment?" it asked.
"Commitment. I was avoiding it."
"Okay. What's the smaller version that you can ship Monday by 10am?"
I shipped it Monday at 9:47am.
This is the thing a human chief of staff does — the calm, non-judgmental accountability pass that you can't do to yourself. Alex does it too. Every Friday.
Day 14: I stopped opening ChatGPT for business questions
Not because ChatGPT got worse. Because opening Alex meant I didn't have to re-paste context. My goal, my bottleneck, my team, my runway — all already loaded. I'd ask the question and get an answer tuned to my actual situation, not a universal framework.
ChatGPT still wins for drafting emails and learning new concepts. Alex won for "what do I do on Monday."
Day 21: The hard conversation
I had a key hire that wasn't working. I'd been avoiding the conversation for a month. I asked Alex what to do.
Alex didn't say "you should fire them" or "give them a chance." Alex said:
"Two things. One: run one honest 30-minute conversation where you tell them exactly what's not working. Measure whether anything changes in two weeks. Two: prepare the counterfactual — what does the team look like without them in six weeks, including backfill timeline. Then decide."
I ran the conversation. Nothing changed in two weeks. I had the backfill plan ready. I made the decision in 90 minutes instead of 90 days.
Day 30: Five things that changed
- I stopped making important decisions on the fly. Everything goes through Alex first. Five minutes of conversation saves a month of being wrong.
- I stopped confusing "urgent" with "important." Alex's weekly triage forced me to name the one thing that mattered. Everything else got deferred or killed.
- I pushed back on my own plans. Alex asked hard questions I'd been avoiding asking myself.
- I stopped skipping the weekly review. Because Alex remembers and asks, I couldn't skip it.
- I stopped paying a consultant for questions Alex could answer. Saved ~$2,000/month.
The honest limits
- Novel problems. Alex is trained on the web; it doesn't know the inside-baseball patterns of my specific vertical the way a deep domain expert does.
- People. Alex can't read a room. Hard people decisions still need a human.
- Soul. Alex won't tell me whether I should keep running this business. That's still on me.
Would I keep it?
I paid $49 for the month. I made at least one decision — the hard conversation — that would have been worth 100x that. The weekly review alone is worth the cost.
Yeah. I'm keeping it.