Cómo Usar la IA Para Ser Disciplinado, Hacer Cosas Difíciles y Ahorrar Tiempo

Main Ideas

The video argues that AI can save time and improve productivity—but only when people choose the right problems, provide enough context, and critically evaluate its answers. The presenter draws on a workplace study, management frameworks, and his own experience.

The “Jagged Frontier” of AI

A 2023 study involving researchers from Harvard Business School, Wharton, MIT, and the University of Warwick tested GPT with employees at Boston Consulting Group. The presenter reports that AI users completed 12% more tasks, worked 25% faster, and produced work rated more than 40% higher in quality.

The study also illustrates AI’s “jagged frontier”: its capabilities do not decline predictably as tasks get harder. AI may handle a seemingly difficult task well and fail at a seemingly simple one. Users therefore need enough expertise to judge whether an answer is reliable. In the study, some AI-generated responses to a task outside AI’s strengths were wrong but sounded especially convincing.

The presenter also notes that people who received training in using GPT did better than those given the same tool without training.

Three Common Mistakes

1. Giving AI tasks instead of problems

   • People often use AI for routine activities such as drafting emails, summarizing, translating, or writing posts.
   • Instead, start with an important problem—one that could meaningfully affect results, such as a business bottleneck or a consequential negotiation.
   • Apply the Pareto principle: prioritize the small share of work that produces most of the results.
   • Work with AI on topics where you can judge whether its response is good. If you cannot judge it, use AI to explore ideas rather than make the decision for you.

2. Treating AI like a search engine

   • A short, vague prompt often produces a generic answer because the model fills in missing details with likely patterns.
   • The result can be unhelpful or fabricated while still sounding confident.
   • The strongest users in the study worked with AI iteratively. Some divided tasks between themselves and AI; others intertwined their work with it. Both approaches involved an ongoing exchange rather than a single prompt and answer.
   • AI memory can preserve context across conversations, but users still need to provide relevant information.

3. Confusing a convincing answer with a correct one

   • AI can make errors in a persuasive, coherent way, so fluency should not be mistaken for accuracy.
   • The presenter says less-skilled consultants in the study improved more with AI than more-skilled consultants. However, expertise remains important: it helps people tackle complex problems and recognize when AI is wrong.

Three Skills for Using AI Well

1. Prioritization

   • Ask: “What is the one thing I can do that would make everything else easier or unnecessary?”
   • List the week’s tasks and mark those likely to produce a substantial impact.
   • Choose just one high-impact task to focus on.
   • Check whether you have enough knowledge to evaluate AI’s answer. If not, use it for exploration rather than relying on it for a decision.

2. The CRIT Method: Context, Role, Interview, Task

   • Context: Explain the situation, desired outcome, what you have tried, available resources, constraints, suspected problems, and relevant data. The presenter recommends voice dictation as a way to include details that might be omitted when typing.
   • Role: Specify the kind of expert AI should act as, including relevant experience and specialization.
   • Interview: Ask AI to question you before giving a solution. Its questions can surface missing information, contradictions, and assumptions—and help you think through the problem.
   • Task: After establishing context, role, and interview, request a specific output, such as strategies, a plan, likely causes, or overlooked risks.
   • Over time, repeated conversations can build context and reduce how much background needs to be restated.

3. Judgment

   • Review the answer: State what you like, what you do not like, and what you would change, then refine it.
   • Challenge the answer: Ask AI to argue against its own proposal, identify errors, risks, assumptions, and blind spots, or assess it from another perspective—for example, as a client or competitor.
   • The goal is to lead AI’s work rather than defer to it, and to reach a result you would personally stand behind.

A Six-Step System for Applying AI

1. Identify three areas to improve: List potential problems under:

   • Money: Increase income or prevent major losses.
   • Time: Simplify, document, or automate repeated work.
   • Decisions: Determine what to prioritize, structure, eliminate, or delegate.

2. Choose one area and problem: Avoid trying to optimize everything at once. Select a valuable problem before attempting a complex AI setup.

3. Work through it using CRIT: Provide context, define the role, let AI interview you, and then ask for a proposal.

4. Review and challenge the proposal: Identify what works and what needs changing, then ask AI to test the plan for risks and alternatives.

5. Automate only after the process works: AI agents may carry out longer, multi-step tasks, but first test and refine the process manually. Automating a bad process can make it fail faster, and longer-running agents require supervision.

6. Build on repeated use: Use AI regularly on an important problem so that each conversation builds useful context about your work, constraints, and preferences.

The video closes with a 30-day challenge: once a day, use AI on a genuinely important decision, strategy, opportunity, or long-postponed problem. Choose a high-impact issue, provide context, evaluate and challenge the answer, and improve it. The aim is both to improve the result and to build richer working context over time.

Other Material Featured

The video includes a sponsor segment for Visual Business School’s master’s program in applied AI and production-process optimization. The presenter describes it as a six-month program focused on building AI-supported work systems, assistants, dashboards, and automations.

Speakers and Sources Featured

• Álvaro Hernández Jarque — Presenter and narrator; the only person speaking directly in the subtitles.
• The 2023 workplace study — Attributed to researchers from Harvard Business School, Wharton, MIT, and the University of Warwick; conducted with Boston Consulting Group staff.
• Karim Lakhani — Harvard professor identified as the study’s lead.
• Geoff Woods — Management and AI author whose ideas inform the prioritization and judgment sections; the subtitles render his name as “Goh Woods.”
• Books and frameworks referenced: The ONE Thing and The AI-Driven Leader.
• Visual Business School — Sponsor mentioned in the presentation; the subtitles also name Chema Alonso and David Hurtado in connection with the program.
• AI companies and products mentioned: Anthropic, Google/Gemini, and OpenAI/ChatGPT. The presenter cautions that specific models and products may quickly become outdated.