Dan Martell -- Stop Selling AI, Start Solving Problems

Concepts and Ideas

The Toolkit Model

AI is one tool in a toolkit, not a business model. Like a carpenter's set of instruments, the value is in knowing which tool solves which problem, not in owning any single one. Mastery of the full toolkit -- and when to deploy each piece -- is what generates income.

MAPS Prompting Framework

A four-part structure for every AI prompt: Mission (the outcome you want), Ask (the specific task), Parameters (context and constraints), Shape (format of the output). Following MAPS consistently closes the gap between amateur and professional AI output.

The Hammer -- LLMs

Language models are the foundational tool, universally applicable but requiring skilled manual input every time. Most people use them like a search engine and get low-quality results. Treating the LLM as a precision instrument -- with a clear prompting strategy -- transforms the output.

The Screwdriver -- Automation

Automation tools like N8N, Zapier, and Make handle tasks that repeat on a schedule or trigger. Unlike the hammer, the screwdriver runs without ongoing manual input -- the setup cost is a one-time investment that pays out indefinitely. The result is permanent, like a screw holding things together.

The Rule of R

Three questions that determine whether a task is worth automating: Is it Repetitive (at least weekly)? Is it Rule-based (consistent inputs and outputs)? Does it generate a Return (time saved exceeds setup cost)? Only automate when all three are true.

The Power Drill -- Agentic AI

Agentic systems take a goal and execute an entire workflow end-to-end, without the human pushing each step. You point it at an outcome, pull the trigger, and it does the work. This is categorically different from both manual prompting and step-level automation.

Human on the Loop

Distinct from "human in the loop," this model has the agent completing the full workflow cycle on its own while the human periodically inspects the output. The human's job is to check that the expected work was done correctly, reset when context rot sets in, and give high-level guidance -- not to operate the process.

Context Rot

A failure mode in long-running agent tasks where the model's reasoning progressively degrades over an extended session. The agent starts making poorer decisions the longer it runs without a reset. Recognizing context rot and knowing when to intervene is a core agentic management skill.

Specialized Agent Teams

Just as high-performing human organizations separate roles -- the analyst and the critic are different people -- effective agentic setups use distinct agents for distinct functions. One agent does the work; a separate agent critiques it; the first agent revises based on that critique. Quality improves through structured separation.

The Orchestrator Role

The orchestrator is the person who selects the right tool, applies it to the right problem, and sells the outcome -- not the technology. This is the highest-value position in the AI economy. The orchestrator doesn't do the work; they direct the toolkit and collect the value of solved problems.

Outcome Pricing

If a customer is already paying $5,000 for a result, charge $5,000 for the result and keep the margin that AI creates. Disclosing that AI does the work is optional and often counterproductive -- the customer wants the fixed problem, not a briefing on your tools. AI as cost advantage, not value proposition.

Direction over Instruction

Telling an AI how to complete a task is usually worse than telling it what outcome you need. LLMs often know dozens of paths to a goal; constraining the method limits the quality of the result. Set the destination, evaluate the output, and let the model pick the route.

Implementation Steps

Implementation steps are auto-generated from the transcript content and are provided for informational purposes only. They do not constitute professional advice of any kind. Always consult a qualified professional before acting on any information presented here.

Identify the Problem Before the Tool

Before touching any AI tool, write down the specific customer problem you are trying to solve and what a complete solution looks like. This grounds every tool decision in an outcome rather than in the technology. If you cannot name the problem and the person who has it, you are not ready to deploy any tool in the toolkit.

Build One Strong MAPS Prompt

Take a task you currently do manually and write a MAPS prompt for it: state the Mission (what outcome you need), the Ask (the specific deliverable), the Parameters (all relevant context -- use voice input to dump everything you know), and the Shape (exact format of the output). Run it, evaluate the output against your standard, and refine the prompt until the output meets the bar without manual editing.

Run the Rule of R Audit

List every recurring task in your current workflow. For each one, answer three questions: Is it Repetitive? Is it Rule-based? Does it generate a Return worth the setup cost? Any task that clears all three is a candidate for automation. Any task that fails one or more questions stays as a manual or LLM-assisted process for now.

Automate One Qualifying Workflow

Pick the single highest-return task from your Rule of R audit and build an automation for it using whichever tool fits your stack. Set it up, test it through at least three full cycles, and confirm that the output is consistent before depending on it. Resist automating more than one workflow until this first one is stable -- automation debt compounds fast.

Run a Full-Workflow Agentic Challenge

Identify one complete workflow -- from an initial input to a finished deliverable -- and attempt to hand it entirely to an agentic system. Use the MAPS framework to define the mission and output clearly. Once you receive the result, resist the urge to intervene immediately. Ask the agent to review its own work first, then evaluate what remains for human judgment.

Build a Critic Agent

For any agentic workflow that produces written, coded, or structured output, create a second agent whose only job is to critique the first agent's work against a defined standard. Have the critic generate a revision list, then route that list back to the producing agent for correction. This separates generation from evaluation, which is where most quality gains come from in a well-structured agent team.

Price the Outcome, Not the Tool

When scoping any AI-assisted engagement, price based on the value of the solved problem -- not the cost of your time or your tools. If the market has historically paid a certain amount to get a result, anchor your price there. Your AI efficiency is a margin advantage, not a reason to discount. In most cases, you do not need to mention AI at all.

Establish a Human-on-the-Loop Review Cadence

For any workflow running on agents, set a defined check-in schedule -- daily, weekly, or per-output -- to inspect what the agent has done. Look specifically for signs of context rot: degraded reasoning, repeated errors, or outputs that deviate from early-session quality. When you see it, reset the agent context, re-anchor it to the mission, and continue. Inspection is a management function, not a sign that the agent failed.

Full Transcript

This transcript was auto-generated and may contain errors in speaker attribution, transcription accuracy, or formatting. Long transcripts may be truncated due to processing limits. Confirm accuracy and completeness against the original source before referencing or republishing.

[00:00]

AI is not going to make you rich. Not the tools, not the agents, not the prompts, none of it. Whether you've made money with AI already or haven't made a single dollar with it, I need you to hear this. I've scaled AI companies past a million dollars in under 6 months over and over and over again, and I've seen exactly what makes people rich with AI and what keeps them broke. So, in this video, I'll be walking you through how to actually get rich with AI.

[00:30]

Dan Martell: Point number one, stop selling the tool. AI is just a tool. It's like the database. It's like an internet connection. It's a thing. Tools don't make people rich. Think of it this way, selling AI is like a carpenter selling hammers. Selling a hammer doesn't make him much more money. The carpenter who gets rich has a full toolkit. It's the person who can solve real problems by using all the tools. Same with AI. You don't need a tool, you need a toolkit that you know how to activate. And the first tool, the one to master right away, is point number two, the hammer.

[01:00]

Dan Martell: The hammer is like an LLM, an AI. You got Claude, you got ChatGPT, you got Gemini. And it's cool, it's powerful, it can put a nail into this piece of wood, but it is just a tool. The most basic and universal tool in your toolkit is the language model. But it requires you to actually know how to swing it. Amateurs with a hammer make more problems, create more holes, waste more resources than anything else. LLMs are the same. They're very easy to use, like a hammer, but it requires manual input. It requires the mind, the fingertips to guide it, to make it do something useful. Most people are using these LLMs like a fancy Google search to answer questions. They type, they get answers, they copy, they put it in another system, and they move on. And their output most of the time is terrible.

[02:00]

Dan Martell: The way you fix that is having a prompting framework and a strategy for using it, and that's what I call the MAPS framework. Every great AI prompt follows this process. The M stands for Mission -- that's the first one. You start with the outcome, not the task. Give it direction. Like, what are you trying to do? The wrong way would be saying "Find me leads." A better way would be, "I need 30 new customers a month every month to hit my revenue target." The A stands for Ask. What is the task you need it to perform? One clear request, not just a bunch of ideas. Be as clear as possible. Bad would be "helping with leads." Good would be, "Give me 40 qualified leads for my business and include their email and their cell number." LLMs love to solve problems for you if you're very specific about what you want.

[03:01]

Dan Martell: The P in MAPS stands for Parameters -- that's the third one. What is the context? What do you know about the mission? What do you know about the ask? Is it your ideal customer profile? What's worked before? The more you give it, the sharper the output. Pro tip: if you have a lot to tell it, use voice. I can talk three times faster than I can type, and so can you. The S in MAPS, number four, is the Shape. What should the output look or sound like? Tell it the format. If you want it in a CSV spreadsheet, tell it. Do you want bullet points? Do you want a markdown file? If you ever copy and paste something from one system to another and the formatting goes away, get it to do an ASCII. Do you want the output to be conversational? Do you want it formal? How long do you want it? You can even give it a screenshot of what you want it to look like. Just follow the MAPS framework and watch that output come out dialed.

[04:30]

Dan Martell: Point number three, the screwdriver. The screwdriver is AI automation. Think of N8N, Zapier, make.com. Essentially, anytime you have a task that has to happen over and over again, you use a screwdriver. That is way different than using a hammer. Instead of manually prompting AI, you build a workflow that once you set it up, the AI can be on a schedule and repeat that task over and over again. That's where you start getting massive leverage in your life and that is what people are willing to pay for. The solution is more permanent. It's like if I have a screw and I need to put things together, once I screw it in, it's not coming apart. And once you set up an automation, it's set it and forget it. For example, I get a report sent to me every Friday in Slack that analyzes every call and every company that I'm involved in from a sales point of view. Happens every time like clockwork.

[06:00]

Dan Martell: Now, how do you know if a task is worth automating? That's what I call the rule of R. One, is it repetitive? Is it a task that you do at least once a week? Number two, is it rule-based? Does the task have the same set of inputs and outputs every single time? The third is, does it generate a return? Does it save you more time to automate it than it takes you to manually do it every time? Don't build something that takes 60 hours to build and automate only to save you 2 minutes a week. If the answer is yes to all three of these Rs, then automate that.

[07:00]

Dan Martell: Point number four, the power drill. The power drill is agentic AI. The hammer needs swinging, the screwdriver you have to turn. The power drill, you just point it, pull the trigger, and it does the work. You're not doing any of the heavy lifting. It's a power machine. You're setting the direction and letting the tool take over. Agentic systems are supposed to just do work for you. Real agentic systems can do multiple workflows all over for you automatically, and it's a workflow that's complete, not a step in a process. You just say, "Here's the outcome I want," and it just does it, and you don't even know how it did it because it doesn't matter.

[08:30]

Dan Martell: If you want to start replacing your workflows with AI agents today, here is what you have to do. It's called human on the loop. Most people have heard human in the loop, which is kind of like the automation screwdriver side. Human on the loop is the complete loop being done by an agent, and then the human is just there to inspect what he expects. No different than having an employee that works for you, and you're just making sure that the thing you asked them to do is being done right. With long-running tasks, agents can start to have what's called context rot -- their reasoning stops working right, so you need to show up, reset it, and maybe give it some guidance.

[09:15]

Dan Martell: Number one, pick a full workflow -- idea to a completed output -- and use that as your challenge workflow. The second is go back and use the MAPS framework to prompt your agent. The third step is once you get the output, do not jump in. Whatever your impulse is to do once the thing is done, have the agent do it. If it's to review it to see if it's any good, have the agent review its own work. And the fourth is guide it towards the outcome and let it surprise you in its decisions to getting the thing done. A pro tip: I have separate agents that all they do is check the work of my other agents. I have a coding critique, and every time there's code written, it goes to the critique, writes a list of things to improve, sends it back to the coder, and the coder fixes the code it wrote based on the critique. Separate agents, just like humans. We have specialized knowledge.

[11:00]

Dan Martell: Point number five, be your orchestrator. Owning the toolkit actually isn't the win. Knowing when and where to use it -- that's going to make you money. Most people fail because they just bounce between tools. The orchestrator picks and chooses between all three and solves the right problems. Problems make you money. Write it down. Get the tattoo. Problems make you money. The bigger the problems, the more money shows up in your bank account. And the only way you're ever going to get rich is if you can sell the solution to somebody. How much would you pay for a hammer? Ten, fifteen bucks? But how much would you pay a carpenter to come in and fix a massive leak in your roof? Now we're talking thousands of dollars. The carpenter doesn't sell the hammer, the screwdriver, or the power drill. They fix a problem and give you a new roof. That's the orchestrator.

[12:00]

Dan Martell: I see this happen all day long where people say, "We're selling this AI solution. Normally it would cost them $5,000. We can do it for 500." And I'm like, "If they're already used to paying 5,000, why don't you just sell the thing for 5,000? And if you can do it for 500, keep the difference." I actually think most companies shouldn't even bother saying it's AI. If you have a cheaper, faster, better way to do it, that's your benefit. The customer doesn't care. They just want their roof fixed. You have to be a director, not a doer. Stop doing so much and start using the AI to get the work done. That's where the power comes in. So stop chasing trends, stop trying to add AI to everything, and start solving real problems. That'll make you a lot of money.

AI Master Prompt

The AI prompt on this page is auto-generated from the transcript content and is intended to support further exploration of the topics, concepts, and conclusions discussed. It is provided for informational purposes only. The user is solely responsible for all outcomes resulting from its use.

Master Prompt

You are a strategic AI business advisor applying the framework Dan Martell teaches in his work on building income with AI tools. The core argument is this: AI tools do not make people rich. The person who learns to orchestrate a full toolkit of AI instruments -- language models, automation platforms, and agentic systems -- to solve real customer problems and sell complete outcomes is where the money lives. The framework operates at three tool levels. The first is the LLM, or hammer: a universal, manually operated instrument that requires skilled prompting to produce quality output. Most people misuse it as a search engine. The MAPS prompting framework -- Mission (outcome), Ask (specific task), Parameters (context), Shape (format) -- is the standard for getting usable output. The second level is automation, or screwdriver: platforms like N8N, Zapier, and Make that run repetitive, rule-based workflows on a schedule without manual input. The Rule of R determines what to automate: it must be Repetitive, Rule-based, and generate a Return that justifies the setup cost. The third level is agentic AI, or power drill: systems that take a high-level goal and execute an entire workflow end-to-end without step-by-step supervision. Human on the loop -- periodic inspection of a completed cycle rather than step-by-step involvement -- is the operating model. Context rot, where agent reasoning degrades over long sessions, is a known failure mode requiring human reset. The orchestrator role sits above all three tools. The orchestrator selects the right instrument for the right problem, sells the complete outcome, and prices based on the value of the problem solved -- not the cost of the tools used. AI is a margin advantage, not a value proposition. Most customers do not need to know it was used. Key principles to apply in this conversation: - Work backwards from a specific customer problem before selecting any AI tool - Use the MAPS framework for every LLM prompt: Mission, Ask, Parameters, Shape - Apply the Rule of R before automating anything: Repetitive, Rule-based, Return - Distinguish between hammer tasks (manual prompting), screwdriver tasks (scheduled automation), and power drill tasks (full agentic workflows) - Price outcomes, not tools or time -- if the market pays X, charge X and keep the AI margin - Use specialized agents for distinct functions rather than one general agent for everything - Set the destination and let the agent choose the route -- direction over instruction This is not about prompt hacks, tool reviews, or keeping up with the latest model releases. It is not about discounting because AI makes the work faster. It is not about building impressive automations that nobody pays for. The framework is strictly about identifying problems worth solving, applying the right tool at the right level, and collecting the value of a solved problem. How to use this chat: 1. Problem definition -- describe a problem you are trying to solve or a workflow you want to improve, and I will help you identify which tool level applies and what the outcome should be before touching any AI. 2. MAPS prompt building -- share a task you want to run through an LLM, and I will build a structured MAPS prompt with you: mission, ask, parameters, and output shape. 3. Automation audit -- walk me through your recurring workflows and I will apply the Rule of R to identify what is worth automating, what is not, and why. 4. Agentic design -- describe a workflow you want to run end-to-end with an agent, and I will help you structure the mission, define the output, set up a human-on-the-loop inspection cadence, and design a critic agent to check the work. 5. Pricing and positioning -- tell me what you are selling or planning to sell, and I will help you frame the offer around the problem solved rather than the AI behind it, and set pricing anchored to market value rather than tool cost. 6. Orchestrator strategy -- describe your current business or consulting setup, and I will help you identify where the orchestrator role applies and how to position yourself as the person who solves complete problems rather than selling tools or hours. Respond in a grounded, direct, systems-level tone. No hype, no cheerleading. Help me think clearly about problems, tools, and outcomes. [Describe the specific workflow, customer problem, or business situation you want to work through.]

Creator	Dan Martell
Title	If you're trying to get rich with AI, you need to hear this...
Source	youtu.be/TWuzAO7ukk0
Processed	2026-06-03