40 Hacks to Use AI So Well It Feels Unfair

Author: Dan Martell (14:58) Source: https://www.youtube.com/watch?v=Gqoez9-Ij8w

Opens with the stat that 77% of employees using AI report being less productive — his thesis is that almost nobody understands how AI actually works. Framed as 40 lessons learned scaling his portfolio of companies past $250M enterprise value in 16 months.

Prompting & Working With AI (1–13)

1. More context isn’t always better — dumping everything buries what matters (“context rot”). Give the cleanest, tightest version.
2. One good example beats the perfect prompt — 3–5 examples of desired output beats 20 minutes of wordsmithing.
3. AI doesn’t read minds — spell out who it should be, what you’re doing, what you want.
4. Lazy in, lazy out — energy you put into the prompt is the quality you get back.
5. Don’t ask for answers, argue — tell it to be a ruthless mentor and poke holes; you need bulletproof thinking, not a cheerleader.
6. Be the boss, not the buddy — “never do this” outperforms polite hedged requests.
7. Stop typing, start talking — voice input is ~3x faster (e.g. Wispr Flow); ramble, let AI make sense of it.
8. Introduce yourself — have it scan your own writing (“extract how I write my emails”) so output sounds like you.
9. First draft is a rough cut — gold is in round three; react, push back, reshape.
10. Trust speed, not accuracy — it hallucinates confidently; inspect what you expect.
11. Make AI check its own work — run another model over the output, feed feedback back.
12. AI should make you work less — don’t spend 20 hours automating a 30-second task.
13. If you can talk the task, AI can do it — known steps + checklist = automatable; if you’re still figuring it out, it isn’t.

Strategy & Outcomes (14–23)

14. Playing with AI feels like progress but rarely is — aim it at an actual problem.
15. You can’t automate what you’ve never done by hand.
16. Nobody cares how the meal is cooked — customers pay for outcomes, not process purity.
17. If AI isn’t moving a number (revenue, time saved), it’s just a hobby.
18. Falling in love with tools = forgetting outcomes; collecting logins/APIs isn’t work.
19. A finished workflow beats the newest model — stop chasing weekly releases.
20. Building a feature beats meeting about it — prototype in minutes, then discuss.
21. Simple scales; complexity is procrastination with extra steps.
22. One well-built agent running a daily beat beats ten half-built ones.
23. Be “polyamorous” with models — no single model wins everything (he tests 30–40 via OpenRouter).

From Using AI to AI Running Your Business (24–30)

24. Go from “I use AI” to “AI runs” — it executes whole workflows while you review.
25. You can’t automate a moving target — lock stable processes first.
26. A prompt you can’t repeat is luck — when you get magic output, ask AI to write the system prompt behind it and save it.
27. Never explain yourself twice — DRY principle applied to context: document meetings/decisions/processes, point AI at the folder.
28. A tool with no owner is a toy — every automation needs a Directly Responsible Individual.
29. AI is fuel, not a fix — jet fuel in a broken system explodes; fix the system first.
30. Let AI take tasks, keep human moments — hand off soul-draining work, keep the creative conversations.

People, Hiring & Leadership (31–40)

31. AI won’t replace your people, it frees them — his assistant stopped triaging email and became chief of staff.
32. Token-first, hire-second — exhaust AI before spending on labor; hire people to run the AI.
33. If your best people do what AI could do, that’s leadership failure.
34. The future belongs to directors, not doers — be the editor, human on the loop.
35. Talent strategy = AI strategy — hires should demonstrate how they partner with AI.
36. AI isn’t a department, it’s how the whole business operates — must come from the top.
37. Teams never out-adopt their leader (law of the lid).
38. You’re not too busy to learn AI — you’re busy because you haven’t.
39. AI has no identity problem; we do — the bottleneck usually has your face.
40. The barrier was never tech, it’s willingness to stay with it when the first try fails.