AMA for AI Tools: What To Use, When, and How

What's Your Relationship With AI?

Before we talk about tools, we need to talk about how you're using AI right now. There are three very different relationships people have with it: Are you using AI to help you think? Are you running AI to do work faster? Or are you using it to actually execute tasks for you? Those aren't just different use cases. They're different identities and knowing which one you are right now determines which tools matter, which advice applies, and what your next step should be.

Three AI Identities

Where are you today, and where are you growing toward? Prompter, You're using AI conversationally for your own work. Drafting, research, brainstorming. You drive every interaction. This is where most people start, and it's a perfectly good place to be. The key is being intentional about it. Builder, You're connecting tools, creating reusable workflows, building custom GPTs or skill libraries. You're not just using AI, you're designing systems around it. This is where individual productivity starts compounding. Director, You're orchestrating agents, writing specs, managing AI the way you'd manage a team. You define outcomes and let the systems execute. This is where AI becomes infrastructure, not just a tool. Most people in the room are Prompters moving toward Builder. Some are already building. A few are directing. All three are valid, what matters is knowing where you are so you can grow deliberately.

Don't Chase Tools. Match Tools To Jobs.

The right tool in the wrong context wastes more time than no tool at all. There's a new AI tool launching every day. The instinct is to try everything. Resist that. Instead, think about the job you need done and match the tool to it. Here's the map.

The AI Tool Map

• ChatGPT, General-purpose conversational AI. Consumer-oriented. Increasingly monetizing with ads. It's the one everyone knows, and it's fine for general tasks. Just know that the business model is shifting toward consumer attention, not professional depth.
• Claude, Long-form writing, analysis, and nuanced reasoning. This is where you go when you need AI to think carefully, not just respond quickly. $20/month for Pro, or $100/$200 for the Max plans that unlock higher usage and capability.
• Perplexity, Purpose-built for research. It gives you sourced answers, not generated text. When you need to find something and verify it, not just get a plausible-sounding response, this is the tool.
• Notebook LM, Built for large data sets. Huge context windows. Think of it as your data analysis home base, upload large documents, financial reports, or research sets and work with them directly.
• Claude CoWork, File-based workflows. Spreadsheets, docs, project files. This is Builder-level tooling, it operates on your actual files rather than just chatting about them.
• Agentic Platforms, Platforms like OpenClaw where you create named agents with identities that execute tasks autonomously. This is Director-level infrastructure. You define the agent, give it tools and rules, and it runs.

Your Most Important Tool: You!

Every time you run a prompt that works, ask AI to codify it as a markdown skill file. Build a library. Claude has a standard format, the skill creator, that other platforms accept. Start with one or two per week. It doesn't have to be complicated. After a good session, just say: "Summarize what we did here as step-by-step instructions I can reuse." Save that file. That's a skill. As you progress from Prompter to Builder to Director, those skills become the building blocks for your agents and workflows. What starts as a saved prompt becomes a reusable system component. Your skill library may become the most important professional asset you own outside your network. "When you're done chatting, ask AI to codify what you just did."

What's Coming In the Next Year-ish

Not because you need to act on it tomorrow. But so you recognize it when it arrives. The field is shifting fast, and the most useful thing isn't predicting exact products, it's understanding the eras we're moving through and what each one demands from you.

Three Eras of AI

Most people are still in Era 1. The game has moved.

• Era 1: Prompt Engineering (20222024), Individual, synchronous, session-based. You craft instructions and iterate output. The value is personal, your skill, your results. The hard ceiling is that it doesn't scale or transfer. When you close the chat, the work disappears.
• Era 2: Context Engineering (20242025), The magic isn't the prompt, it's what you feed AI alongside it. Persona documents, brand voice guides, product data, customer research. This is what closes the gap between "generic AI output" and "sounds like our brand." Context engineering is where Builders live.
• Era 3: Intent Engineering (2026Now), Context tells agents what to know. Intent tells agents what to want. This is about encoding organizational purpose into infrastructure, not prose, but actionable parameters. When you write a spec that defines inputs, constraints, and desired outcomes, you're doing intent engineering. This is where Directors operate. The question isn't which era is "better." It's which era you're operating in, and whether you're building the skills to move to the next one.

Better Specs. Better Outcomes.

AI doesn't need you to manage it. It needs better specs and clearer goals. The rest it'll figure out. Here's a data point worth sitting with: senior engineers who wrote code alongside AI only improved their productivity by about 20%. But when they were removed from the code entirely, limited to writing specs and QA'ing the output, productivity exploded. The lesson applies far beyond engineering. It's a software paradigm that works everywhere: define the inputs, state the desired outcome, get out of the way. "When you stop telling AI the steps and instead give it the right context and a clear goal, everything changes."

What To Do This Week

Based on where you are right now.

• Prompters, Pick one tool, use it daily for one thing. After a good session, ask AI to summarize what you did as reusable instructions. That's your first skill. Commit 30 minutes per week to AI literacy, reading, experimenting, learning what's possible.
• Builders, Organize your prompts into markdown skill files. Learn about context windows, how they work, why they matter, and what happens when they fill up. Use Claude CoWork or Notebook LM to hold client context separate from your instructions.
• Directors, Write specs, not prompts. Define inputs, constraints, and outcomes. Ask yourself: "If I had to hand this to someone with zero context, what would they need?" That question is the foundation of intent engineering.

You Don't Need to Master AI

You need to get intentional about how you use it. One tool, one skill at a time.