Listen to the podcast episode
Or listen now on Apple Podcasts or Spotify.
Key Takeaways
Measure AI by moved KPIs, not by how busy it keeps you: AI interfaces are fast and almost addictive — you ask, it answers, and it feels like work is happening. Conrad compares it to people who walk around the gym for 50 minutes, burn 17 calories, and go home. Set specific business metrics — bookings, occupancy, revenue — and check whether the AI output actually moves them. If it doesn’t, that’s fake progress.
A thousand thin AI posts will lose to ten good human articles: Clicking a button to generate hundreds of 300-word blog posts is a path to nowhere: you might see an initial traffic spike, but that’s the rise before the fall, and Google’s spam updates are actively rolling through sites that did it. Vacation rental sites are already bloated with listing pages. Answer the real questions your travelers and homeowners have and you’ll show up in search and LLM results.
Point AI at technical audits, where it actually shines: Give Claude a 10,000-page website and it will crawl every page in chunks, flag typos and broken links, and report back — work you used to do by hand, or display in Screaming Frog without any way to fix it. That’s the real win right now: AI as a tireless auditor and drafting assistant, not a strategist making decisions alone.
AI will generate ideas forever — that doesn’t make them good: Ask any LLM how amazon.com could convert better and it will happily oblige, then keep going every time you say ’give me more ideas.’ It never stops, and plenty of those ideas are bad. If you expect it to be a sparring partner that pushes back, you’ll be sorely mistaken — you’ll keep tweaking copy and design while spinning your wheels.
Your context and prompt quality decide whether the plan is any good: Conrad asked a free model for a vacation rental marketing plan and got advice like rewriting listing titles weekly and leading with a steaming-coffee-cup ’emotional anchor’ photo — he agreed with maybe a third of it. The same prompt in Claude, loaded with his book, client transcripts, and prior plans, asked smart clarifying questions and produced a far stronger plan. Generic input, generic output — and roughly 97% of users are on free models.
Fifty email journeys nobody enters is fake progress: Hyper-segmentation only works if people actually reach those journeys. Paul has seen HubSpot builds with elaborate automated pathways where nobody ever entered the sequence — a solution to a problem that doesn’t exist. Personalized email absolutely lifts conversion, but only when you understand the real traveler or homeowner journey and route actual humans into it.
You already got burned by this — it was called Google Ads recommendations: Ads people are more skeptical of AI because they watched algorithmic recommendations and single-keyword ad groups fail years ago. Apply that same skepticism when AI repackages old mistakes with new branding, and never blindly accept auto-written ad copy or one-click optimizations — an account run at 100% of Google’s recommendations would blow your mind in the wrong direction.
Use AI as a draft-writer with you in the loop: The best current use case: AI pre-writes the guest message, email reply, or support response, and a human confirms it before it sends. That collapses latency — a guest getting the Wi-Fi code back in six seconds is good hospitality. But you are still the decision maker in your business. Feed AI your data, take its insights, and keep the final call.
What We Cover In This Episode
Conrad and Paul draw the line between real AI progress and fake progress — the button-clicking, report-generating activity that feels productive but never moves bookings, occupancy, or revenue. They walk through good and bad AI use cases across SEO, paid ads, and email, and stress-test a generic AI-written marketing plan line by line.
