This week I did a talk and published a companion website for nonprofit fundraisers on how to get started with AI, including tips, pitfalls, and some of the tools I recommend. It went well, and sitting with all of it afterward, I realized I’d never written a general version of that advice for this site. Most of what I write here assumes you’re already using these tools and looking to go deeper. This post is for those who might be a bit earlier in the journey.

The single most useful reframe I know comes from researcher Ethan Mollick, who calls it the “best available human standard.” The question isn’t whether AI is as good as the world’s best expert at whatever you’re trying to do. It usually isn’t. The real question is whether it’s better than the best help you’d actually have access to in a given moment. For most people, that realistic alternative is either no help at all, or a colleague who’s already doing three other jobs. AI isn’t a replacement for genuine expertise, but it’s often better than nothing. Sometimes a lot better.

That framing has been useful to me because it cuts through a lot of the noise. There’s hype that says AI can do everything, and haters who say it’s never good enough. In my experience, neither of those is the right frame.

It’s not like using Google

Most people’s first instinct is to use AI the way they use a search engine: type a question, get an answer, move on. That misses almost everything that makes it valuable. These tools can draft, push back, help you think through a stuck problem, or role-play a difficult conversation before you have it for real. If you’ve only used a Large Language Model (LLM, like ChatGPT, Claude, or Gemini) to look things up, you’ve barely started.

The mode I find most valuable, and the one many people never try, is thinking out loud. Bring it something half-formed. Describe a problem before you know what solution you’re after. Ask it to help you figure out what you actually think. That exploratory, conversational mode is often where these tools do their best work.

What you put in determines what you get out

A single thin prompt, like “write me a cover letter” or “give me ideas for this project,” gets you something thin in return. The first output is rarely the useful one. What actually works is going back and forth: explain the context, say what you don’t want, push back on what it gives you, ask it to try again differently. You’re having a conversation, not making a request.

The AI knows nothing about you or your situation. It has no idea what your organization does, what your voice sounds like, what’s already been tried, or what constraints you’re working under. None of that is in the model. You have to bring it. I’ve found voice transcription genuinely useful here. Talking for a few minutes about what I’m working on is a lot faster than typing out context, and it often helps me think through the problem before the AI has said anything at all. Handy is the tool I use for this.

The move most people don’t try

Most people use AI to generate things: draft this, brainstorm that. What fewer people try is turning it around and asking it to push back… Find the problems with this plan. Argue against my proposal. Tell me what I’m missing. Read this as a skeptical reviewer and tell me where it falls down.

You don’t have to take every critique. But you’ll almost always know more than you did before you asked. I’ve come to think of this as one of the most underused capabilities these tools have.

Find your jagged edge

AI is very good at some things and surprisingly bad at others, and the edges aren’t where you’d expect. One person finds it invaluable for drafting; another finds it completely flat for their particular writing style. One task clicks immediately; another never quite works. The only way to find this out is to actually use the tools and push them past what feels comfortable. That map is personal, and you can only build it through experience. (This framing also comes from Mollick, he’s that good.)

Keep yourself at the center

AI has no stake in what you’re doing. It has no judgment, no taste, and no sense of what actually matters in your specific situation. The sense that something came from a real person who cares can only come from you. Use AI for the work where it helps, so you have more capacity for the work where it doesn’t.

Don’t assume the computer knows best. It can give you options and surface ideas you wouldn’t have found on your own, but it’s not an authority. It still gets things wrong, and it won’t tell you when it does — it sounds equally confident either way. If it gives you a recommendation, ask for the reasoning. Use your own judgment and stay in charge.

A few tools worth knowing

Beyond the major chat interfaces (Claude, ChatGPT, Gemini — all have free tiers, try them and see what fits):

  • Handy — voice transcription for talking to AI instead of typing, especially useful for providing lots of context quickly
  • LLMFeeder — a browser extension that strips a webpage down to clean text so you can paste it directly into a conversation
  • liteparse — converts PDFs to text so you can work with them in AI tools

For going deeper, a good place to start is AI for the People, a free six-week email course from The Guardian that walks through how to think about AI in daily life. Worth starting with whether you’re brand new or already comfortable with these tools — I learned things from it, and I’ve been at this for years.

After that, Ethan Mollick’s Co-Intelligence and his newsletter One Useful Thing are the most grounded writing I know on how AI actually works in practice. And if you want to see some of what’s possible when you’ve been at this for a while, Moving Beyond the Basics is a good next step from here.