APPLIED AI

Newsletter

I had AI put together two great reports and thought I'd figured it out — then the third came back garbage. The Harvard study that explains why 19% of AI users perform worse than people who never opened the tool. Plus a free tool to map which parts of your job are most exposed, the reason the same prompt gives different answers, and what Gallup found when it asked 22,000 workers about their daily AI habits.

A customer loves our AI receptionist so much she reads its conversations to her husband at dinner. Murphy Tiggelaar on customer obsession, the five questions that surface what customers actually need, and a prompt that makes AI argue against your own business.

There's a new word for something you've probably already felt. AI output that looks polished but means nothing — no real thinking, no decision, just a shell that leaves you to do the actual work. Brian Holmes on how to spot it, stop producing it, and what to do when it lands on your desk.

The AI never stops encouraging you. Every idea you float, it tells you is brilliant. Jake Hodges on why that's the most dangerous thing it can say — and what happened when blank pages hit production. Plus two Applied AI sessions and four stories on brains, bots, and Lloyd's of London.

Almost every AI tool you use is 'closed' — you rent it. Open source is the hedge. Plus two Applied AI sessions on responding to reviews and starting with open source models, what Congress is doing about AI on Main Street, and a paid AI job that requires no degree.

Pick one task, run it ten times without touching it, then move to the next. Plus two Applied AI sessions on email/calendar and responding to reviews, and what 81,000 Anthropic users said about where AI actually pays off.

AI hype is cooling, and that's not a bad thing. Plus two Applied AI sessions on email/calendar and responding to reviews, why your team is probably pasting customer data into AI, and what PwC found about who's actually winning with AI.

I checked. The latest AI model is genuinely impressive. It also still won't sound like your front desk unless you tell it who you are. One example, two responses, and a lesson that matters more than any benchmark.

Stop re-explaining yourself to AI. Set your context once and every answer comes back sharper, faster, and in your voice. Plus: the AI 101 monthly session and a tool to show what slow follow-up is costing you.