How I Used Dumb AI to Become a Smarter Voter

After 43 years of voting, I finally felt like I understood every choice on my ballot. And I have an unreliable AI apprentice to thank for it.

I’ve been voting since I was 18. That’s more than four decades of elections, candidates, propositions, judges, and political promises.

And I’ll admit something.

For most of those years, I didn’t feel nearly as informed as I wanted to be.

I voted. I cared. I tried to do my homework. But there was always more information than I had the time or the patience to dig through, and I knew it.

Especially here in Los Angeles.

This year’s ballot was nine pages long.

Nine.

Statewide propositions. Local measures. Candidates. Judges. Offices most people probably couldn’t name if their lives depended on it.

And somehow we’re supposed to understand all of it well enough to make intelligent decisions.

How?

The Problem With Being an Informed Voter

I don’t trust political advertising.

There’s too much money in it. Too many special interests. Too many organizations trying to shape what we believe, what we fear, and ultimately how we vote.

When a political ad comes on, I usually turn it off. I’m not interested in somebody with a multimillion-dollar ad budget telling me what to think.

But that creates a different problem.

Where do I get reliable information?

I can read the official voter information pamphlets. I can visit candidate websites. I can research ballot measures and read recommendations from organizations I trust.

But every source has its own perspective, and sorting through all of it is real work. The official guide is thick. The endorsements contradict each other. The arguments are written by people who want a particular outcome, even when they’re written honestly.

The Part That Bothers Me Most

Here’s what I keep coming back to.

I’m retired. I have more control over my time than I’ve had in my adult life.

If I want to spend three hours on a Tuesday reading about a water district measure, nobody’s stopping me.

And even with that advantage, the ballot was overwhelming.

So think about the person who doesn’t have that.

Probably you.

Somebody working fifty hours a week.

Somebody with kids in school and a commute and a second job or a parent who needs care.

Somebody doing shift work and getting home at nine at night with just enough left in the tank to eat something and go to bed.

Where in that week does a fair reading of nine pages of ballot measures go?

It’s difficult to imagine where they’d find the time.

We’ve built a system that asks voters to do a level of research most people don’t have the hours for, and then we act surprised when people rely on a mailer or a thirty-second political advertisement.

This year I wanted to do better.

Not because I suddenly had more time, but because I had a different tool.

Meet My Dumb AI Apprentice

I’ve been using artificial intelligence consistently for a little over a year.

In the beginning, it was mostly a toy. I’d ask questions, poke at it, occasionally be impressed by what it could do. Then I started to see it as something closer to leverage.

But working with AI was frustrating as hell.

I’m a retired union electrician. I spent 35 years on large commercial and industrial projects, and I know exactly what it’s like to work with apprentices.

A first-year apprentice needs constant supervision. You hand him a task and you can almost guarantee he’ll find some creative way to do something you didn’t ask for.

Maybe he’ll do it wrong.

Maybe he’ll do it right in a way that breaks something else.

Either way, you’re not leaving him alone with anything expensive.

That’s what working with AI felt like in the beginning.

I’d give it clear instructions and it would ignore them, half-follow them, forget them, or confidently tell me something that turned out to be wrong.

And the last one was the real problem.

An apprentice who says “I don’t know” is easy to manage.

An apprentice who tells you the wrong thing with total confidence is how you end up redoing a whole afternoon.

So I stopped treating it like a reliable tool and started treating it like a trainee.

That meant a few things.

Giving it something to work from. I started sharing my Obsidian notes, reference materials, and the frameworks I already use. I built a Google Drive knowledge library so it could pull from sources I’d chosen instead of guessing.

Showing it where to look. I stopped asking it open questions and started pointing it at specific documents, specific files, specific things I wanted it to compare.

Making it show its work. I learned to insist on knowing where a claim came from. If it can’t point to a source I can go read, I don’t want the claim.

Staying in the loop on anything that mattered. Delegation, it turns out, is not the same thing as abdication. An apprentice can do the grunt work, but somebody still has to check the work.

Over time, the relationship got better.

Today I’d call it a third-year apprentice in a five-year program.

And I'm still on the fence if it will ever achieve journeyman status.

Maybe 80% of the time, it does what I need. The other 20%? Still a mess.

That’s an impression, by the way.

I’m not claiming it’s a measured number. It’s just what the working relationship feels like after a year of paying attention.

And that’s the whole point.

AI doesn’t replace my judgment. It helps me apply it more effectively.

I don’t need it to think for me. I need it to help me do the work.

What It Actually Looked Like

This election, I decided I wasn’t going to rush through the ballot or lean on endorsements from organizations whose agendas I didn’t fully agree with.

So I sat down with my AI and worked through it one item at a time.

That’s a boring sentence for what was actually a slow, sometimes irritating, occasionally surprising process.

First, I gave it voting guides from progressive, moderate, and conservative organizations.

Not because I wanted any of them to tell me how to vote. Because I wanted to see how different people looked at the same issue.

And this is where I want to be careful about something.

Those guides were useful, but useful isn’t the same as authoritative. They’re interpretations.

They’re built by organizations with priorities, and the way they frame a measure is itself a piece of information.

Two honest people can read the same statute and describe it in ways that would sway you in opposite directions, and neither of them has to be lying.

So I wasn’t collecting conclusions. I was collecting arguments.

The value of having three perspectives wasn’t that averaging them would make me neutral. It was that reading all three showed me which parts of the story each side left out.

Then we started working through the ballot.

For each measure, candidate, or race, I asked the same set of questions.

• What exactly am I voting on? Explain it in plain English.
• What’s the history behind this issue?
• What problem is it supposed to solve?
• What are the strongest arguments in favor?
• What are the strongest arguments against?
• Who supports it, and why?
• Who opposes it, and why?
• Who’s funding the campaigns?
• What are the potential unintended consequences?
• What information is missing or uncertain?

And then I chased whatever didn’t add up.

This was the actual work, and it’s hard to describe without making it sound too clean. It wasn’t.

It was reading a summary, thinking “that can’t be the whole story,” asking another question, and getting an answer that either closed the gap or opened a bigger one.

Sometimes an argument sounded convincing right up until I read the opposing one.

Sometimes a measure sounded good until I looked at how it would be funded or who’d have to implement it.

Sometimes organizations I generally agree with recommended something I wasn’t sold on.

And sometimes organizations I usually disagree with raised concerns that genuinely made sense.

That’s what I was after. Not people who agreed with me, but a real understanding of what I was voting on.

One issue at a time. One question at a time. By the end, it had stopped feeling like nine overwhelming pages and started feeling like individual decisions I could actually examine.

Even the Judges

Judicial races were always the hardest part for me.

I’d look at the names and wonder how an ordinary voter is supposed to evaluate people running for judge.

Most of us don’t follow judicial careers.

We don’t spend evenings reading court opinions or researching attorneys’ professional histories. The names appear on the ballot with a job title and nothing else that a normal person can use.

This time, I used AI to find information about the candidates: their academic and professional backgrounds, their experience, whether they’d served on the bench or worked as prosecutors, public defenders, or in private practice.

And here’s the thing I had to learn: not every candidate has a judicial record. Some have never sat on a bench.

So the question isn’t “how did they rule,” it’s “what does their actual professional record show, and what does it tell you about how they’d handle the job.”

I wasn’t asking AI which candidate was better. I was using it to gather the evidence, and then deciding for myself which qualifications mattered to me.

An AI summary isn’t proof of anything. Important claims still need to be checked against official evaluations, court records where they exist, and credible reporting.

But the research became manageable.

For the first time, I felt I had a reasonable basis for choices in races I’d always found hard to evaluate.

The Best Thing AI Has Done for Me

I’ve used AI for writing, research, organizing information, building business systems, and handling tedious administrative work.

Some of that has been genuinely useful.

But helping me work through this election may be the most valuable use I’ve found.

Not because AI knew the answers. Not because it was smarter than me. And certainly not because I trusted everything it said.

I don’t.

It was valuable because it helped me ask better questions, weigh competing arguments, and find information I might have missed.

It turned an enormous research project into a series of manageable decisions. One issue, one question, one decision at a time.

After 43 years of voting, I mailed in my ballot feeling more confident about my choices than I ever remember feeling before.

Not because I knew everything.

Because I’d done the work.

And every decision was mine. That last part matters more to me than the confidence does.

NEVER Outsource Your Thinking

Here’s the mistake I see people making with AI.

They treat it like an oracle. Ask a question, get an answer, assume the answer is correct.

That’s dangerous.

• AI can misunderstand the question.
• It can invent information.
• It can rely on outdated sources.
• It can present an argument as if it were objective when it isn’t.
• And it can sound extremely confident while being completely wrong.

That’s why I treat mine like an apprentice.

Useful?

Absolutely.

Capable of a huge amount of research and grunt work?

Definitely.

Qualified to make important decisions for me?

Hell no.

That’s my job. And yours.

The Limit Worth Naming

I want to be honest about something, because it would be easy to oversell this.

Feeling well-informed and being right aren’t the same thing.

I did more due diligence than I ever have.

That doesn’t mean every conclusion I reached was correct.

Some of what I learned may have been incomplete.

Some of my reasoning may have been off.

AI can be confidently wrong, and a confident-sounding summary can still lead you somewhere you wouldn’t have gone if you’d read the original document yourself.

I did more research than I’ve ever done before.

But I also recognize that AI can get things wrong, and no amount of confidence replaces checking the original sources.

That’s something I intend to keep improving.

So here’s the habit I’d recommend to anyone doing this.

For anything that matters, verify it against the official source.

The actual text of the measure. The official voter guide. The candidate’s own filings.

Somewhere the AI’s reading isn’t the only thing between you and the facts.

The AI is a research assistant.

It’s not the authority.

Which means the confidence I have is confidence in my process, not a guarantee about my conclusions. That’s a different thing, and an important one.

⚔ The Dojo Drill: Research Your Ballot

You don’t need to build an elaborate system like mine. You can start with a simple prompt.

Copy this into your preferred AI assistant:

Then give it your official ballot information or tell it which election you’re researching.

Take your time with it. Ask follow-up questions. Push back on its answers. Verify anything important against the original source.

And make the decision yourself.

If you’d like the more detailed prompt and research process I used, just reply to this email and I’ll share it.

Because the point was never to have AI do my thinking for me.

The point is to use AI to become better at thinking for myself.

And if my dumb third-year apprentice can help me do that, imagine what it could do for you.

📚 From the Leader’s Library

​The Trusted Advisor — David H. Maister, Charles H. Green & Robert M. Galford​

Why?

Because trust is the foundation of every meaningful relationship, whether at work, at home, or on the mat.

This book introduces the Trust Equation, a practical framework for understanding what makes someone trustworthy.

More importantly, it challenges you to examine your own credibility, reliability, and ability to put others’ interests alongside your own.

Learn to build relationships where people feel safe enough to be honest, take risks, and grow.

P.S. Know a martial arts gym owner who’s stressed about money or student numbers?

Do them a favor: send them to The Leader's Dōjō, my website where I help owners get more students and keep them longer with simple systems.

One forward from you could change their gym: The Leader's Dōjō​

Chuck