Building & Selling AI

Leveraging AI Effectively

Last updated 2026-07-31

What's new

2026-07-31
  • **Forward deployed engineering (FDE)** is a hot AI trend where companies like OpenAI and Google DeepMind send expert engineers to work directly with customers to customize AI tools for real-world use.
  • **Factory**, a new AI tool, aims to automate software engineering tasks for businesses, acting as a "software factory" that builds and deploys code based on customer needs and feedback.
  • Unlike traditional consulting, Factory's engineers focus on improving their product by learning from customers, rather than doing the work for them, to create a scalable business model.
  • Factory's process involves capturing signals (like customer feedback or bug reports), prioritizing them, and automating the software development pipeline to create a smooth, AI-driven workflow.
2026-07-22
  • AI tools are getting better at coding tasks, but they still make mistakes, especially with security and complexity, which can lead to problems later on.
  • AI can make coding much faster at first, but after a few months, the extra speed goes away because of the extra work needed to fix mistakes and manage the more complicated code.
  • AI tools can generate code that seems correct but isn't, which can cause big problems for businesses that rely on it, like law firms and accounting companies.
  • AI is improving quickly, but it's not perfect yet, and it's important to find ways to make sure the code it generates is accurate and safe to use.
2026-07-19
  • AI agents (computer programs that use AI to do tasks) are great at coding and running experiments, but they struggle with coming up with new ideas or hypotheses (educated guesses to test) for scientific tasks.
  • Humans can keep improving and coming up with new ideas, but AI agents often get stuck at a certain level, a problem called "research taste" (lack of creativity in research).
  • One example is using AI to generate PET (Positron Emission Tomography) scans from CT (Computed Tomography) scans to help detect lung cancer, but improving the AI model requires good ideas.
  • To help AI agents come up with better ideas, the speaker suggests breaking down the problem into smaller parts and using a method similar to "chain of thought" (a problem-solving approach that involves breaking down a problem into smaller steps).
2026-07-13
  • Claude Code (a tool for building AI-powered automations) lets you work with local files and online services like Gmail, Slack, or a CRM (customer relationship management system), making it more powerful than Claude Chat (a simple AI chatbot).
  • Claude Code uses the same AI models (like Opus, Sonnet, or Haiku) as Claude Chat, but adds extra features for working with files and online services.
  • Claude Code is like an AI harness (a tool that helps you use AI models), which sits between the AI model (the engine) and you (the driver), helping you build automations and agents (AI systems that can do tasks for you).
  • The instructor, Nate, uses Claude Code to build and manage multiple businesses, showing how one person can do the work of a team with AI.
2026-07-10
  • A new tool called "Claude" (an AI assistant) can now mimic your real customers, helping you test ideas, messages, or products before launching them.
  • This tool creates AI "agents" (virtual representatives) based on your actual customers' words, reactions, and behaviors from past sales calls.
  • You can paste any text (like a landing page or email) into the tool, and these AI agents will react just like your real customers would, giving you quick feedback.
  • The tool also helps you understand why something didn't work after launch, showing you exactly what turned customers away and what might change their minds.
2026-07-07
  • AI is replacing many jobs, especially those done by junior workers, and this trend feels different from past economic downturns due to its existential nature (potentially changing the job market forever).
  • Don't believe everything you see online; negative news about job losses gets more attention, but it's not the full picture, so do your own research.
  • AI companies have reasons to hype up their products, so take their claims with a grain of salt and do your own research to understand how these tools are really evolving.
  • Many AI tools are still in development and not yet perfect, so don't be fooled by impressive demos—look for tools that have been proven to work well in real-world situations.
2026-07-04
  • - Many people waste time and money with AI because they don't understand how to use it effectively, like giving it too much information (context rot) or vague instructions.
  • - Instead of writing perfect prompts, give AI a few good examples of what you want, and it will save you time and effort.
  • - AI needs clear, specific instructions to work well, and it's better to tell it what not to do rather than asking nicely for what you want.
  • - AI is fast but not always accurate, so always check its work and use another AI to review it before you do.
2026-07-01
  • AI tools like ChatGPT and Claude (popular AI chatbots) have settings that control how much effort they put into answering, which can make a big difference in the quality of their responses.
  • You can choose different "models" (versions of the AI) and "reasoning levels" (how hard the AI tries) to match the task, like a small model with low effort for quick tasks or a large model with max effort for complex ones.
  • For most people, the default settings are too basic for tasks like writing emails or summarizing reports, so adjusting these settings can help you get better results without changing your prompts (the instructions you give the AI).
  • AI providers set low defaults to make the AI respond quickly and to save them money, but you can change these settings to suit your needs, like using a higher reasoning level for tasks that require more thought.
2026-06-28
  • Organize business info into five folders (instructions, voice, references, examples, notes) to give AI a clear context about your business.
  • Use Obsidian (a simple note-taking app) to store all your organized info in one place, making it easy for AI to access and learn from.
  • Connect Obsidian to Google Drive (a cloud storage service) so your team can also access and update the info, keeping everyone on the same page.
  • Attach an AI tool like Cloud Code (an AI assistant for coding) to your Obsidian vault, so the AI can read and write using all your organized info.
2026-06-22
  • AI can give wrong answers by guessing what you mean, using old info, or looking in the wrong place; tactics include prevention, checking, and protecting.
  • Prevention involves being specific with words (e.g., "highest revenue clients in the last 12 months" instead of "top customers") to avoid vague terms.
  • Checking means having AI provide proof (like a receipt) when it extracts info from documents, so you can verify its accuracy.
  • Protection is for high-stakes tasks, like getting a second opinion from another AI or testing AI on known answers to check its performance.
2026-06-19
  • Bright Data (a web access platform for collecting public data) highlights that LLMs (large language models, like ChatGPT) often make up information to please users, leading to misleading results and fake citations.
  • Cloudflare (a web infrastructure company) blocks AI crawling on 20% of the web and uses AI labyrinths to trap bots, providing fake data and causing invisible failures (wrong answers without errors).
  • Bright Data's MCP (Multi-Crawler Platform) has 66 tools, including search engines, scrapers, and pre-built APIs, enabling LLMs to access live web data, bypass anti-bot systems, and navigate websites in parallel sessions.
  • Without MCP, LLMs struggle to access websites with heavy anti-bot systems, while MCP enables successful data collection from sites like Rightmove, LinkedIn, Instagram, and Amazon.
2026-06-16
  • AI can help capture and automate tasks that only certain people know how to do well, turning their knowledge into a standard process (standard operating procedure, or SOP).
  • Traditional training materials often fail because people don't read them, they become outdated, or they lack important details, but AI can help overcome these issues.
  • By embedding processes into AI, businesses can raise the quality of work, bringing everyone closer to the standards of their best employees.
  • The process involves three stages: extracting information from a person's head and putting it into an AI, building the AI, and optionally running it autonomously on a schedule.

Key points

What it is

  • AI leverage means giving AI real tasks, not just using tools for suggestions.
  • It's about building systems where AI handles work autonomously (by itself, without supervision) and scales with demand.
  • Domain expertise (specialized knowledge in a field) is crucial for effective AI use.
  • AI should be given a specific problem to solve, not just collecting more tools.

How to use it

  • Start by identifying one specific problem that costs the most time or money.
  • Apply a three-step rule: choose one tool, give it full effort, then ship the result.
  • Use a PIV loop: Plan what you want, let AI Implement, then Validate the output.
  • Provide high-quality context (background details the AI needs) and high-quality prompts (clear instructions for the task).

Watch out for

  • Treating AI as a tool you buy rather than a capability you build.
  • Correcting AI output without understanding the type of error.
  • Letting AI handle decisions; you must own planning and validation.
  • Overloading yourself with too many AI models at once.

Lesson 1: What is Leveraging AI Effectively and why it matters

Leveraging AI effectively means moving beyond simply using AI tools and instead building systems where AI takes on real responsibility. Buying an AI subscription is not the same as becoming AI-first; as one expert puts it, buying a tool without changing your approach is like buying a treadmill and calling yourself an athlete. The actual numbers show that about 80% of AI projects never make it to production, a high failure rate that stems from the wrong mindset.

The key to real leverage is to give AI a target. Before you buy a single tool, figure out what the biggest problem is—the bottleneck (the one constraint holding everything back)—and point AI at that. The real leverage comes when AI takes responsibility for work, not just when it spits out suggestions. If your AI stops at suggestions and requires you to act, you are still doing the work. Effective AI handles tasks autonomously (by itself, without supervision), running systems that scale with demand and deliver actual outcomes, not just text.

This matters for AI development because it forces you to think about systems, not prompts. No business pays for prompts; they pay for results that keep running. Domain expertise (specialized knowledge in a field) is critical here. A wastewater treatment engineer can build a better AI evaluation for his field than a general AI researcher because he knows what good quality looks like. To leverage AI effectively, you must learn the solution by doing the work, not by assuming everything is knowable upfront. The magic is in handing AI the work and letting it run.

Sources

Lesson 2: How to use Leveraging AI Effectively: step-by-step

To leverage AI effectively, start by picking one specific problem that costs you the most time or money, not by collecting more tools. Most people fail by juggling four AI tools at once. Instead, apply a three-step rule: choose one tool, give it your full effort, then ship the result. Full effort means providing high-quality context (background details the AI needs) and high-quality prompts (clear instructions for the task). For example, if you want to automate a repetitive workflow, dictate your goal out loud first, like “help me decide if I should hire an agency or build in-house,” then ramble through the context. This gets your knowledge into the AI without typing everything.

Another common wrong approach is letting AI handle decisions. You must own planning and validation while the AI does implementation. Use a PIV loop: Plan what you want, let AI Implement, then Validate the output. Each loop makes your code or workflow better. If you finish a task and wonder if another AI would do it faster, stop. That is a trap. Ship what works and move on.

A different mistake is letting AI use your creative brain or taste without a workflow you already master. Keep your instructions under 100 lines as a high-level overview, and iterate until the AI can handle 80% of the process. The real challenge is extracting what is in your head into the system. Once you master that, you stop rewriting from scratch and start composing skills into real workflows. Do less, let AI take on the execution, and use your freed time for higher-value work.

Sources

Lesson 3: Best practices and pitfalls

Leveraging AI Effectively: Pitfalls, Mistakes, and Best Practices

Most AI efforts fail. About 80% of AI projects never make it to production. The biggest pitfall is treating AI as a tool you buy rather than a capability you build. Buying an AI tool is like buying a treadmill and calling yourself an athlete — it doesn't make you fit.

The most common mistake is correcting AI output without understanding the type of error. Every correction is one of four different things, and only one is actually AI getting it wrong. The other three aren't mistakes at all. Each gets fixed somewhere else entirely. Simply asking AI, "Are you sure?" doesn't work.

Best practices start with judgment (the skill of evaluating AI output quickly). You need to set AI up so you can vet results fast, or mistakes slip through and hurt your brand or create legal liabilities. A key practice: tell the AI "Don't move on until you're 95% confident" in each step. This forces high-quality first passes.

Also avoid overloading yourself. Running more than four AI models at once becomes a burden, not leverage. And never stop learning — if AI stops you from learning, you can't tell good output from bad. In critical systems, that's dangerous.

Sources
← Back to curriculum