Both Hands Full

What Creatives Actually Need to Know About AI

I didn’t ask for this.

None of us did.

One day you’re a photographer, designer, writer, filmmaker building a career on skills you spent years developing. The next day, someone in Silicon Valley has scraped your work into a dataset, trained a model on it, and is selling the ability to generate something “like yours” to anyone with twenty bucks a month.

My work is in there. I released thousands of photos on Flickr under Creative Commons over the past two decades, and when I checked a dataset search tool, I found roughly 1,800 of my images in a major training corpus. That doesn’t mean every AI model used them, trai

ning pipelines vary, and the tool shows evidence, not certainty… but it means the consent story around AI is exactly as casual as critics claim.

I’m sympathetic to the people who want to resist. I understand the impulse to refuse, to opt out, to fight back. But I’ve also spent the last two years working with hundreds of creative professionals who are trying to figure out what to actually do in this moment. And what I’ve learned is that neither pure resistance nor uncritical adoption gets you anywhere useful.

So I’m asking you to walk forward with both hands full.

---

The Non-Consensual Moment

Let’s name what’s real.

The consent problem is real.

AI systems were trained on the creative output of humanity… photographs, writing, code, music, art… largely without permission. Some of us released work under licenses that technically allowed it. Many didn’t. The fact that companies are now settling billion-dollar lawsuits over training data tells you everything you need to know about how legitimate those practices were.

The junior pipeline problem is real. The entry-level work that used to train senior creatives is disappearing fast. You’d get hired at a studio, spend six months converting approved assets into different formats, and in the process learn the systems, the clients, the back-and-forth. That’s how juniors became seniors. When AI handles that work, we lose the pathway—and we don’t yet know what replaces it.

The dependency question is real. If you don’t use a muscle, you lose it. What happens when we stop doing the first stages of creative work ourselves? When we outsource idea generation, research synthesis, early drafts? We’re all participants in an experiment we didn’t sign up for, and the results aren’t in.

These aren’t hypothetical concerns. They’re happening now. Anyone who tells you they’ve got it all figured out is either selling something or not paying attention.

And yet.

I’ve spent two years rebuilding my creative practice around these tools, and I’ve never felt more capable. Not because the tools are magic, they fail constantly, in ways I’ll get to… but because learning to work with them has forced me to get clearer about what I actually bring to the work. The parts that are mine.

Both things are true. The system that created these tools violated consent on a massive scale, and working with them has made me a better creative professional. I’m asking you to hold that contradiction and keep walking.

---

Agency Via Workflows

Here’s where most AI advice goes wrong: it’s all about tools.

“Use ChatGPT for this, Midjourney for that, here’s my stack of twelve subscriptions, connect everything with APIs.” That approach creates two problems. First, it overwhelms people who just want to know how to do their work. Second, it misses the point. Tools change constantly. The one I’m using today might not exist next year. What matters is the workflow the process that gets you from input to output, with checkpoints along the way.

Let me share three workflows that have actually changed how I work. I’ll mention tools only in passing, because the specific tools matter less than the pattern.

Workflow 1: Research ? Synthesis ? Briefing

The old way: Spend days reading documents, taking notes, trying to synthesize a complex topic. The new way: Drop the source materials into an AI tool that can work with documents, have it generate a synthesis, then interrogate that synthesis for gaps and errors.

I used this recently with a 485-page federal budget. I needed to understand the AI policy implications. Reading it cover-to-cover would have taken weeks. Instead, I dropped it into a document AI, asked targeted questions, had it generate an interactive summary I could query further.

Where it fails: The AI will confidently synthesize things that aren’t quite right. It’ll miss nuance, flatten complexity, occasionally hallucinate connections that don’t exist. You can’t use the output without verification. I treat it as a first draft that needs checking, not a finished product.

The guardrail: Always go back to source documents for anything that matters. Use the AI to find where in 485 pages the relevant information lives, then read those sections yourself.

Workflow 2: Capture ? Transcribe ? Knowledge Base ? Retrieval

Everything I say in meetings, talks, and conversations gets recorded, transcribed, and stored in a searchable knowledge base. When I need to write something, I’m not starting from blank I’m retrieving my own past thinking on the topic.

The talk I gave last night? Parts of it will become this article. Not because an AI is writing it for me, but because I have access to my own words, my own examples, my own way of explaining things. The AI helps me find and recombine my own material.

Where it fails: Garbage in, garbage out. If your recordings are rambling, your retrievals will be rambling. If you haven’t actually thought through a topic, having transcripts of your half-baked ideas doesn’t help.

The guardrail: The knowledge base is only as good as the thinking that went into it. This workflow forces me to actually develop ideas, because vague input produces vague output.

Workflow 3: Rapid Prototyping with Client Feedback

The old consulting process: Meet with client, go away for weeks, come back with a proposal, discover they wanted something different. Repeat until budget exhausted.

The new process: Stay on the call, take their requirements in real-time, generate a working prototype while they’re still talking. Show them three versions of what they asked for before the hour is up. Let them discover what they actually want by seeing options, not by trying to imagine them.

Where it fails: The prototypes are rough. They often have bugs, edge cases that don’t work, assumptions baked in that need questioning. Showing someone a “working” thing that’s actually broken can damage trust.

The guardrail: Frame it correctly. “This is a sketch, not a product. It’s here to help us figure out what you actually need.” Manage expectations about what rapid means and what it doesn’t.

---

The Human Moat

Here’s the part that matters most, and the part that’s hardest to hear: if you don’t know who you are as a creative, AI will expose that.

AI is a mirror. It amplifies whatever you bring to it. Bring confusion, get generic slop. Bring clarity about your perspective, voice, and values, get something that actually sounds like you.

The people I’ve watched thrive with these tools aren’t the most technically sophisticated. They’re the ones who can answer questions like: What do I believe about this work? How do I make decisions? What would I never say? What do I notice that others miss?

This is the human moat. Not “creativity” in some abstract sense, everyone claims that. The moat is judgment. Taste. Discernment. The ability to look at ten options and know which one is right, and be able to articulate why.

I’ve started building documents that capture this:

A writing style guide. Not generic rules about grammar. My patterns. The words I use, the ones I avoid, the rhythm of how I construct arguments. I fed an AI my past writing and had it extract the patterns. Now when I ask it to write something, it has a reference point.

A worldview document. What I believe about creativity, technology, ethics, community. Not a manifesto—a working document that captures my actual positions so I can check AI output against them.

A glossary and anti-glossary. The thousand words I use most. The popular words I never use. The phrases that signal “this isn’t me.”

These documents serve two purposes. First, they make AI output more useful—it has something to match against. Second, and more important, creating them forced me to get clear about things I’d never articulated. The exercise of writing them was as valuable as having them.

The Question That Matters

Close your eyes for a moment. Think about one thing you do in your creative work that you would never want to give up to AI. Not because you couldn’t automate it—maybe you could—but because it’s the part that makes the work yours.

That thing you’re thinking of? That’s your creative DNA. That’s the piece worth protecting, developing, doubling down on. Everything else is negotiable. That part isn’t.

---

The Call

Let me end with something uncomfortable.

The most critical voices on AI—the skeptics, the resistors, the people with genuine concerns about consent and labor and dependency—those are exactly the people we need engaged in this conversation. If the most thoughtful critics opt out, the decisions about how this technology develops get made by the people who are all-in. The governance, the norms, the policies—they’re being shaped right now by whoever shows up.

I’m not asking you to abandon your critique. Keep it. The fears are real. The problems are real. The non-consensual nature of this whole situation is real.

But engage anyway. Learn enough to know what you’re criticizing. Build enough to understand what’s possible and what’s hype. Get in the room, because the room is making decisions whether you’re there or not.

Both hands full. Left hand holds the critique… the stolen work, the collapsing pipelines, the genuine uncertainty about what we’re creating. Right hand holds the curiosity—the workflows that might give you back some agency, the clarity about your own creative identity that the mirror forces you to develop.

Walk forward. The moment didn’t ask your permission. Neither did any of the other moments that reshaped creative work… the printing press, photography, digital tools, the internet. Creatives have always had to navigate new capabilities that threatened to commodify their work. The ones who thrived learned to use the new tools while holding onto what made them irreplaceable.

This moment is no different. Harder, maybe. Faster, certainly. But not different in kind.

Both hands full. Keep walking.

---

Kris Krug is a photographer, educator, and founder of BC + AI Ecosystem Association. He runs TheUpgrade.ai, a certification program for creative professionals, and hosts Vancouver AI, the city’s longest-running AI community meetup.