How to Use AI in Your Freelance Business

This blog summarizes learnings from The Future is Freelance Forums. These forums are where stakeholders from all levels of the freelance ecosystem come together to surface and solve the biggest problems facing independent work. These summaries serve as a place to house our collective wisdom on these topics. As such, I want to credit everyone who attended this forum for these insights.

When I asked people registering for this forum how they use AI in their business, I got answers that ran the full spectrum. Some people said things like "every way I can: strategy development, timelines, data analysis, thought partnership, and I want more." Others wrote "figuring it out," "just starting," or "not at all yet but I see possibilities I don't yet fully understand how to execute." One person wrote "carefully." Another wrote "wrong... that's how I use it..."

The range from power user to tentative observer to honest-to-goodness skeptic is exactly what showed up in the room, and it made for one of the most honest conversations about using AI in a freelance business. Nelly Yusupova, a fractional CTO who has spent over 20 years helping non-technical founders understand and use technology, and Anna Burgess Yang, a freelance content marketer and solopreneur operations strategist who has built AI and automation into the backbone of her business, helped me lead the conversation. Here's what we learned together.

What a human should and shouldn’t be doing

This was the throughline of the whole conversation, and it surfaced in almost every breakout room. Nelly framed it early: AI is your junior partner. Not your replacement, not your co-founder, your junior. It can do the grunt work, it can scan a million books in a second, it can start the process. What it can't do is bring the expertise, the judgment, or the deep domain knowledge that you've spent years building. "Everything that I am good at should stay mine," Nelly said, "and everything that AI is good at helps me start the process." That's a useful frame, because it means you're always in the driver's seat. You're not handing over the wheel; you're handing over the copying and pasting.

Some decisions are easy. Anna had a client send her a Google Sheet with 27 line items that needed to be reformatted into a Google Doc. "There's no joy in copying and pasting from a spreadsheet," she said. That one's a no-brainer. A human should not be doing that. 

But the more interesting version of this is when the output requires something of you first. Anna also talked about how she built a brand voice document, not from a template, but from her own actual writing, so that when she asks Claude to help draft social media posts, they actually sound like her. Without that input, she said, it just sounds generic. The AI didn't do the hard part. The years of writing, the self-awareness to name what makes her voice hers, that's what makes it work.

Nelly made the same point from the systems side: you can't automate chaos. Before you can build an automation, you have to actually understand the process. You have to be able to articulate what you do, in what order, and why.

That’s the gray area: drafting content, helping you think through a process, co-writing a proposal. Those aren't off-limits, but they require something from you first. If you haven't done the work of understanding your brand voice, your process, your client, the output is going to be generic at best and wrong at worst. The litmus test I keep coming back to is asking yourself: have I given it enough of me to get something useful back?

This connects directly to something Elliott said. He'd been using AI and kept running into friction; a vague sense of discomfort about how he was using it, a feeling that he was being more reactive than intentional. "What I use AI for has also showed me what I actually don't like," he said. "What you want to delegate to your junior person is work you don't really enjoy. It showed me where I'm in the most flow versus where I'm just doing grunt work."

That's an underrated diagnostic. If you notice yourself consistently offloading something to AI, that's useful information; either it's a no-brainer automation (Anna's spreadsheet), or it's a signal that maybe you shouldn't be doing that thing at all.

Everyone is working out where the lines are

The word "policy" came up early and kept resurfacing, though Nelly pushed back gently on the framing. "Policies are like values we prescribe to, and then everything else is an experiment," she said. I think what people are actually reaching for isn't a formal policy document, it's a gut check. Something a human should not have to do is one of the better tests I've heard. Elliott framed his desire for a policy as wanting to put his values into writing so he had guardrails, not rules, just agreed-upon ground to stand on that can shift as things evolve.

One thing I keep coming back to: those guardrails need to come from you. AI can help you think through your processes, but if you don't actually know your process, the output is going to be crap. I use Claude to help me think through my automations, map out what I should be doing, sanity-check my systems. But I have to know what I'm doing first. That's not something you can delegate.

Nataliia raised something really important: the risk of feeling like you're already behind. She'd been resistant to switching from ChatGPT to Claude even when people she respected were pushing her toward it. "For some period of my life I was more into networking with others who were resistant as me," she said. She eventually made the switch and found it less scary than expected, but her point about surrounding yourself with curious rather than resistant people landed for me.

Skepticism isn't the same as resistance. Some of the best observations in this room came from people who were cautious, who were asking real questions about privacy, about what it means for their craft, about where the lines should be. One registrant who works in biotech wrote in ahead of time about being "unclear and afraid" around privacy. The skeptics tend to ask better questions, and this room was better for having them.

This is what democratization actually looks like

Anna referenced her background helping small banks implement automation, specifically arguing that a tiny bank without endless resources must automate back-office processes to compete on relationships. "I can do more of the work that matters if something else is doing the work that's repetitive," she said.

That's exactly the position we're in as solo business owners, except the stakes are higher. Big companies can choose whether to adopt AI. They have humans they can throw at problems, budgets to absorb inefficiency, and teams to cover gaps. We don't have that luxury. For solo business owners, getting tight and efficient isn't a strategic option; it's a survival requirement. We have to innovate, or we fall behind. 

What's changed is that the tools are finally catching up to that reality. I built project management systems that used to require a team and a budget. Anna built a custom dashboard that pulls from four data sources, something that simply wasn't possible for a solo person before. Nelly pointed out that this creates a real opportunity for people with deep domain expertise: if you know your process and can document it, you can productize that knowledge in ways that weren't viable before. The playing field isn't level, but it's closer than it's ever been.

That's the part of this conversation I find genuinely exciting. Not the hype, not the fear, the fact that we now have the power to create tools that fit the shape of how solo business owners actually work.

The Future is Freelance Forums are collaborative working sessions where freelancers, coaches, community leaders, and platform founders come together to solve real problems. Click here to join the next conversation.