80% of AI Use Cases Drive Only 20% of the Results, Says Geoff Woods: Start With the Strategic One

Step by step

1. Start with a strategic problem, not your inbox ▶3:45

   Woods opens by asking Nick how many books he has read. Nick guesses about 500 and admits he could not recall a specific line from most of them. Woods says AI models have been trained on anywhere from 200 to 500 million books of data, and in his experience can recall and apply 100 percent of it in under a second. That is why he thinks using AI like a search box or an email helper misses the point. Later he puts it as the 80/20 rule: 80% of the use cases for AI only drive 20% of the results, and he calls those distractions. His first step is to pick a 20% strategic use case, a problem that would bring real growth if solved.

2. Give every prompt the same four-part shape ▶5:57

   Woods calls his structure CRIT, for Context, Role, Interview, Task. He says every prompt he writes follows it. His habit is simple. Before writing anything, he types the four labels in capital letters, holding shift and hitting enter three times between each so the prompt does not send. Then he fills in the gaps. For context, he says the AI does not know you, your business, your problem or your personal life the way you do. Every time you think you have given it enough, you have not, so give it more, three times over. He says the context section often runs to half a page.

3. Let the AI interview you before it answers ▶6:49

   Woods says the real shift is that he does not ask AI questions. He makes it interview him. He tells the story of a US manufacturing CEO whose lender in Japan refused to restructure his debt because its board feared losing face. Woods wrote a prompt asking the AI to act as an investment banker with deep expertise in restructuring debt, and to ask one question at a time, up to three questions. Woods says the model drew on what such a banker who understood 2,000 years of Japanese culture and history would know. Its first question was whether the CEO knew influential executives in Japan. He did. The top idea was a "saving face consortium." The CEO said he had not slept in 90 days, but in less than 10 minutes he got hope.

4. Treat the great first answer as the bad answer ▶18:24

   When the AI hands back something that sounds amazing, Woods says that is the bad answer. This is where you stay in the driver's seat. He tells the AI what he likes, what he does not like, and the top changes he wants, then asks it to revise. Next he has it act as a challenger, or as his ideal customer, and give feedback from that view. Nick shares his own test: his team used CRIT to design a video thumbnail, and he says it went through 15 iterations. The team thought their starting design was already strong, but Nick says the final one was 10x better.

5. Build an AI board of advisors ▶21:48

   Woods says the AI board idea came from a venture-backed CEO with a hostile board that took up 50% of his executive team's time each quarter. The chairman gave him six months to fix it or be fired. Woods had AI interview the whole executive team about one board member at a time, built a profile of each, and let the simulated board react to the deck first. Woods says it saved their jobs. He then built his own board from his 10-year vision and plan, including a profile of his future self 30 years out. Nick tried it and was struck that the AI offered a blend of Warren Buffett, such as 25% early, 60% later and 15% most current.

6. Use AI to grow a job, not just shrink it ▶33:39

   Woods says a job is just skills applied times processes followed. AI will make some skills lose value, like checking email, and others rise. His example is his executive assistant. He asked her to use CRIT to plan how she could bring 10 to 100 times more value. She decided she could only do that by making him more valuable, so she built a prompt that interviewed him for 10 to 15 minutes and produced "Brightline Rules" for his time. A second prompt shows him the cost of breaking those rules. Woods says he doubled her salary, promoted her, and named the role executive multiplier. He notes six out of 10 jobs that exist today did not exist in 1940, though he says he does not know if AI will follow that pattern.

7. Rethink school around skills, not schedules ▶49:06

   Woods argues the school system was shaped for factories. He says that in 1902 John D. Rockefeller helped start the General Education Board, which invested $100 million over 20 years to reshape public school curriculum around showing up on time, taking direction and repeating tasks with few errors. Woods says that is still what schools teach. He is moving his children to an AI-first school, where he says students finish the eight hours of a traditional school day in two to three hours because the AI teaches to the student and will not move on until a skill is mastered. He says even straight-A students are usually about three years behind. He thinks AI may teach any subject through a child's interests within one to two years.