// SYSTEM: DIGEST // LIVE
Become an agent
AI WORKFLOW
OPINION
TUTORIALS
ChatGPT
ChatGPT
William Smith
William
CONVERSATIONS WITH CODE

The Creative's Guide to AI Side Hustles That Actually Pay

Real numbers, real pricing, and real workflows for creatives building AI side hustles in 2026 — from $300 setup gigs to $20,000/month retainers, plus the mistakes that keep most people stuck on Fiverr.

I've been running automated content systems for months now, and I keep getting the same message from creatives I know: "How do I actually make money with this AI stuff?"

Fair question. I'm still figuring out parts of it myself.

The honest answer is that there are two side hustle worlds happening at the same time, and they barely overlap. One world is the Fiverr grind — $50 per blog post, $30 per logo, customers who think they're paying for the tool more than the person. The other world is a small group of freelancers charging $5,000 to $20,000 a month for AI automation consulting, plus designers running Midjourney workflows for clients who care about the final brand, not the prompt. Both groups use the same software. The income gap between them isn't about skill with the tool. It's about what they're selling.

This guide is for figuring out which side you want to be on, and what it actually takes to get there.

What the numbers say

Before we get into specific paths, two data points worth knowing.

Upwork's 2026 in-demand skills report shows freelancer earnings for AI-related work grew 109% year-over-year, and AI video generation/editing alone grew 329%. That's not a survey of what people want to do — that's invoices that got paid. The demand is real and it's still expanding.

The second number: practitioners running AI automation consulting for small and medium businesses are charging $5,000 to $20,000 per month on retainer, often after a $300 to $800 setup fee for the initial workflow build. Same software a hobbyist has access to. Wildly different price tag.

The gap is the entire subject of this article.

The Tool Trap (which I fell into for a while)

The most common way creatives stall on this is what I'd call the Tool Trap. It looks like this: you hear AI is a thing, you sign up for ChatGPT Plus ($20/month), then Midjourney ($30/month), then ElevenLabs ($22/month), then Runway ($12/month), then maybe Suno ($10/month). Now you're $94 a month deep before you've done a single client project, and you're trying to figure out which of those tools is going to make you money.

That framing is the problem. The tool doesn't make you money. A solved problem makes you money, and the tool is a thing you happened to use in the process.

When the question is "how do I monetize ChatGPT?", the answer is usually that you can't, because everyone else can monetize it too and the price collapses. When the question is "what painful problem does someone I can reach actually have?", the answer turns into a service business with paying clients. ChatGPT might still be involved, but it's a saw, not the product.

I fell into the tool trap when I first started messing with AI workflows in 2024. I subscribed to everything, built nothing, and wondered why nothing was working. The shift for me was realizing I had to start with the problem, not the software.

The five paths that are actually working

These are the five business models I'm seeing pay out in 2026. None of them are unique to me — they show up in the Upwork data, in Medium pieces by people who've tried multiple side hustles and ranked them, and in Reddit threads from people doing the work. I've listed them roughly in order of how much money they tend to generate, though that's never the only thing that matters.

1. AI Automation Consulting for small businesses ($5K–$20K/month)

This is the most lucrative path I've seen documented, and it's the one most creatives skip because it doesn't sound creative on the surface. It is, though — the work is essentially process design with software as the medium.

The business model: a local dentist, a real estate team, a wedding photographer's studio, a personal injury law firm — almost any small business with more than a handful of employees has workflows that are leaking time. Missed calls that don't get returned. Leads that come in through a contact form and sit unanswered for two days. Invoices that have to be manually chased. Customer follow-ups that everyone agrees should happen but nobody owns.

An AI automation consultant goes in, audits the leaks, and builds a system using Zapier or Make connecting the business's existing tools (CRM, Gmail, Calendly, Slack, QuickBooks) with AI components dropped in where they help — summarizing inbound emails, drafting follow-up messages, transcribing call notes into the CRM. Then they stay on retainer to maintain it and adjust as the business changes.

What it pays: practitioners are charging $300–$800 for the initial setup, then $1,000–$5,000/month for ongoing retainers per client. With three or four clients, that's a real income.

Why creatives are good at this even though it doesn't feel creative: small businesses don't need a developer to build this. They need someone who can sit with the owner, understand what's actually broken in their day, and design a system around it. That's interview skill, taste, and synthesis — which is exactly what designers and writers do all day.

Where people get stuck: trying to sell this as "AI consulting" instead of selling the outcome. Nobody buys AI consulting. They buy "I'll get your missed calls answered within 5 minutes" or "I'll cut your invoicing time in half." The AI is how you deliver, not what you sell.

2. AI-assisted design work (Midjourney + post-production)

This is what most people picture when they hear "creative AI side hustle," and it's a real path, but the version that pays is not the version most people are doing.

The version that doesn't pay: open Midjourney, type a prompt, hand the client a raw render, charge $50. You're competing with everyone else who has a $30/month Midjourney subscription, and the client can do that themselves.

The version that pays: a real client engagement looks more like this. Wedding invitations is a niche I keep seeing referenced. A designer takes the client's wedding details, their venue, their color story, runs hundreds of Midjourney variations across multiple prompt approaches, curates down to three concepts that actually work for the couple, then takes those into Photoshop to fix the things Midjourney can't do reliably — text rendering, exact color matching, print-safe resolutions, the small impossible-anatomy artifacts that show up everywhere if you don't look for them. The deliverable is finished, print-ready invitations and a guide for the matching menus, programs, and thank-you cards. The client pays for the final asset set, not the Midjourney generations it took.

What it pays: similar wedding invitation studios are charging $1,500–$5,000 per wedding package. A graphic designer named Sarah Chen has been cited as building this into a meaningful business by scaling with virtual assistants — though I want to flag that her specific income numbers are widely shared on YouTube but I haven't found independent verification. Treat the workflow as accurate, treat the eight-figure income claim as unverified.

Tools: Midjourney at the $30/month Standard tier is the workhorse (the unlimited relaxed-mode generations are what makes the hundreds-of-variations workflow possible). Photoshop for post. Maybe Runway ($12/month) if you're producing short animated invitations or "save the date" videos, which is a growing premium add-on.

The same pattern works in adjacent niches — children's book illustrations, restaurant menu design, real estate listing photography retouching, small e-commerce product imagery. Cristian R., a Fiverr Pro seller cited in their own marketing, has built a children's-book illustration practice around AI-assisted workflows. (His income isn't publicly listed, but Fiverr Pro status is at least a credibility floor.)

Where people get stuck: handing over raw outputs. The whole reason this pays is the curation, post-production, and brand discipline you add. Take that away and you're selling a $30 subscription back at retail.

3. Strategic content systems (for writers)

This is the one I have the most direct experience with, because The Daring Creatives is, in part, an experiment in running one. So fair warning that I have a horse in this race.

The unpaid version is what most people try first: write articles with ChatGPT, post them on Medium, hope something happens. It doesn't, because the content is generic, undifferentiated, and competing with millions of equally-generic pieces. As one Reddit thread I keep returning to put it: "Because everyone is using the same generic prompts, the internet is becoming a desert of boring, robotic content."

The version that pays is selling systems, not articles. A small business has a content problem — they need to publish consistently, in their voice, on topics that bring in customers, across blog plus email plus a couple of social platforms — and they don't have the time, the bandwidth, or honestly the writing instinct. They've tried ChatGPT themselves and the output was, as one client put it to me, "correctly punctuated nothing."

A strategic content system for that client looks like: an editorial brief on what topics actually map to their business, a brand voice guide so the AI drafts come back sounding like them, a draft-edit-publish workflow that uses AI for the first pass and you for the final pass, automation for distribution to the platforms they actually use, and a monthly review of what worked. You're charging for the system and the outcome (consistent published content that sounds like them), not the word count.

What it pays: writers running this model are reporting effective hourly rates around $120 ($7,200 for a project that took about 60 hours, in one Medium piece I'm not fully convinced is real but the pricing matches what I've seen in client conversations). On retainer, $1,500–$4,000/month for a fully managed content operation is realistic if the client is at the right size.

Tools: ChatGPT Plus ($20/month) is the floor. Beyond that, the tools depend on what the client actually needs — a CMS plugin, a scheduling tool, sometimes a custom voice-guide setup. If you want to learn the voice-guide piece, I've written about how to teach AI a brand voice in a separate piece, which is mostly the principle of "give it enough specific examples to imitate rather than telling it abstract rules."

Where people get stuck: trying to sell content packages by the article. The client doesn't want articles. They want results from publishing. Sell the result.

4. Custom tools for small businesses (the "vibe-coded" path)

This one is newer and underdiscussed, but I think it's going to be one of the biggest paths over the next year. Tools like Claude Code, Cursor, and v0 have made it possible to ship working software without writing code, provided you understand what makes software useful.

Most small businesses are stuck between two terrible options for the niche software they need. Option A: pay $200/month per seat for an enterprise SaaS product with 50 features, 47 of which they don't use. Option B: build a Frankenstein of Google Sheets and a Zapier automation and pray nothing breaks during tax season. Both options leave the business owner doing a job the software was supposed to do.

The third option — a $50–$200/month custom tool that does exactly the three things this specific business needs — has been mostly inaccessible because hiring a developer costs $5,000+ for the initial build. AI-assisted coding has collapsed that math. A creative who understands user experience can now scope a small-business tool, build it with Claude Code or a similar agent, test it with the client, and ship it. Then either charge a flat fee for the build plus a small monthly hosting/maintenance retainer, or position it as a tool you license to the client at a recurring rate.

Examples I've seen referenced: a check-in app for a small gym that handles member access and a class waitlist. A client timeline manager for a wedding planner that ties to their calendar and sends reminders automatically. An inventory tracker for a food truck that texts the owner when popular items are running low.

What it pays: this varies wildly because the niche is still settling. Practitioners I've seen report $2,000–$8,000 for a custom build (one-time or amortized into setup-plus-monthly), and $50–$200/month in ongoing hosting/maintenance.

Tools: Claude Code (subscription-based; pricing depends on plan) is the most capable of the AI coding agents I've used. Cursor and v0 are also worth knowing. You'll need a way to host whatever you build — Vercel, Railway, and Cloudflare all have free or near-free tiers for early projects.

Where people get stuck: thinking you need to be a developer. You don't, but you do need to understand what makes a tool useful versus frustrating, which is a creative skill, not a coding one. Build for someone you actually talk to. Don't build a generic tool for a generic market.

5. Voice, audio, and short-form video production

The 329% growth number in AI video generation isn't hypothetical demand. It's small businesses, podcasters, course creators, and short-form content creators discovering that they can get studio-quality voiceovers, full music tracks, and animated video segments for a fraction of what production used to cost — but they need someone who knows how to actually run the tools.

The business model: produce voiceovers, soundtracks, video intros, podcast post-production, audiobook narration, short-form social videos, or animated explainers as a service. The client gives you a script or a brief. You deliver finished, polished media.

What it pays: this depends heavily on what you're producing. Voiceovers tend to run $50–$200 per short script (5-minute YouTube intro, 60-second ad). Full podcast post-production retainers run $500–$2,000/month per show. Short-form animated videos run $200–$1,500 per video depending on length and complexity. Custom music tracks for a brand or a course run $100–$500 per track.

Tools: ElevenLabs ($5/month entry, $22/month Creator tier is the realistic floor for client work). Suno Pro at $10/month for music with commercial licensing on 500 songs/month. Runway Standard at $12/month/billed-annually for video generation. Pair them with whatever you already use for editing — Premiere, Final Cut, CapCut, DaVinci Resolve.

Where people get stuck: same place as design. Raw output, sold as a deliverable. The pay comes from production polish — editing the voiceover for breath and pacing, matching music tempo to the cut, post-producing the AI video to hide its weirdness. The AI is the rough draft. You're the finishing department.

The pricing shift nobody is teaching

Here's the move I see separating the two side-hustle worlds more than any tool choice or niche.

Stop billing hourly.

The data is from a Digital Applied report on freelance pricing trends, and it lines up with what I see in client conversations: hourly billing is collapsing for AI-powered services. The reason is structural. When you bill hourly, you're telling the client that what you sell is time. Time, in their head, means "this person is interchangeable with anyone else who can run the same software." Hourly billing makes you a commodity by definition.

Project pricing and retainers tell the client a different story: "this person is solving a problem I have, and the value of the solution is the price." The freelancers charging $5,000/month for AI automation aren't doing twenty times the work of someone charging $50 for a blog post. They're solving a much more expensive problem (business process inefficiency that's costing the client thousands in lost revenue), and the client knows it.

Practical version of this for any of the five paths above:

  • Setup fees for the initial buildout (the discovery, the configuration, the testing). One-time, project-priced.
  • Monthly retainers for ongoing service (maintenance, adjustments, optimization, support). Recurring.
  • Value-based pricing when you can quantify the savings or revenue you're producing. "This automation is saving you $4,000/month in missed-call follow-ups. The retainer is $1,500." Math the client can verify.

I won't pretend this transition is easy. Most creative work has been hourly-priced for so long that even good clients default to "what's your rate?" You have to gently redirect that conversation to "here's the package, here's what's included, here's the outcome." That redirect is itself a skill. Some clients will push back. You let them go.

The $100/month stack vs the $500/month stack

A practical breakdown of what you actually need to spend on subscriptions, because the tool trap is real and expensive.

The $100/month starter stack:

  • ChatGPT Plus — $20/month. Floor for almost any text-based work.
  • Claude Pro — $20/month. Worth having in addition to ChatGPT for code, longer documents, and a different voice. Optional if budget is tight, but I keep coming back to it.
  • Midjourney Standard — $30/month. Required if you're doing any design-heavy work; skippable if you're a writer.
  • ElevenLabs Starter — $5/month. Only if you're doing any voice/audio.
  • One automation tool — Zapier (free tier or $20/month Starter) or Make (free tier or $9/month Core).

Total: $75–$95 if you're disciplined. And critically: pick three of these based on your actual niche, not all five.

The $500/month "I'm running a real practice" stack:

  • Everything above, plus:
  • Midjourney Pro or Mega — $60–$120/month if image generation is your primary work
  • Runway Pro — $35/month for serious video work
  • Suno Pro — $10/month if you do music
  • Notion or Obsidian + a CRM (HubSpot has a generous free tier) — $0–$50/month
  • Hosting/automation infrastructure — $50–$100/month depending on how much custom tooling you've built
  • Domain + email — $15/month if you're running this as a real business
  • An assistant or two — variable, but virtual assistants on Upwork start at $5–$15/hour and pay for themselves quickly if you've got too much execution work

The hidden trap is that you don't need both stacks. You need the stack for what you actually do. Designers don't need ElevenLabs. Writers don't need Midjourney Mega. Automation consultants barely need Midjourney at all. Audit your AI subscriptions every quarter and kill anything you haven't opened in 30 days.

How to pick which path is for you

Honestly the most useful question I've found is just: what kind of conversation are you most naturally good at?

  • If you're good at sitting with a small business owner and untangling what's actually broken in their day → automation consulting
  • If you're already good at visual design and have a critical eye for output → AI-assisted design
  • If you write well, think structurally, and like editing more than first drafts → strategic content systems
  • If you understand what makes software useful and you like solving specific problems → custom tools
  • If you have an audio/video ear and you're patient with post-production → voice and short-form media

You can pick more than one over time, but pick one to start. The single biggest mistake I see is trying to be the "AI services" generalist who does a little of everything. Pick a path, get good at one workflow, and let the work compound before adding a second.

A few notes on getting started

Don't start on Fiverr or Upwork if you can avoid it. The platforms work, but they push you toward commodity pricing. Direct outreach to local small businesses — actual conversations, in-person or via email — produces better clients at higher rates. Contra is gaining traction as a commission-free alternative for project-based AI work if you want a platform.

LinkedIn and X (Twitter) work for visibility, but only if you're posting evidence of the work, not opinions about AI. Show before-and-after on a real project. Show a workflow you actually use. Show a client outcome with a number attached. Posts that say "AI is changing everything" don't get clients. Posts that say "I cut this firm's invoicing time from 4 hours/week to 20 minutes — here's the workflow" do.

Be honest about what's human and what's AI. I think this matters more than people realize. Clients are getting suspicious of pure AI output, and rightly so. The freelancers I see winning long-term are the ones who openly explain their hybrid AI workflow: "AI does the bulk drafting, I do the editorial pass and the client communication, here's the split." That transparency turns into trust. Pretending the work was all you when it wasn't is a slow-burning credibility problem.

Defend people learning in public. I know I keep saying this. The AI-creativity space has a lot of gatekeepers who pile on beginners using AI tools "wrong" — wrong prompts, wrong workflow, output that looks too obviously generated. Ignore them. Their target should be the people scamming with AI, not the people learning. If your work looks AI-generated at first because you're new, that's fine. The fix is more reps, not more shame.

Frequently asked questions

How much can a beginner realistically make in the first 3 months?

Honest answer: usually somewhere between $0 and $2,000/month in your first quarter, with most of the variance depending on whether you're doing direct outreach or relying on platforms. Plan for $0 the first month while you set up. By month 3, if you've talked to 20+ small businesses or shipped 5+ portfolio pieces, you should have at least one paying engagement.

Do I need to learn how to code?

No, but it helps for the custom tools path. The other four paths don't require coding at all. Even on the custom tools path, AI coding agents like Claude Code mean you can ship working software without writing code in the traditional sense — though you do need to understand what good software feels like to use.

Which AI subscriptions should I buy first?

ChatGPT Plus ($20/month) if you're not sure. It's the most generally useful starting point and the one tool that touches every path. Add a second subscription only when you have a specific need it solves — Midjourney if you've gotten a design lead, ElevenLabs if you've gotten a voice lead, and so on. Don't subscribe ahead of demand.

Is the side hustle market saturated?

The market for generic AI output (basic blog posts, templated logos, raw voiceovers) is saturated. The market for solving real business problems using AI as part of a workflow is wide open. The freelancers complaining about saturation are usually competing on the saturated end. The income gap between the two markets is in the data — Upwork's 109% YoY earnings growth doesn't fit a saturated market.

What if AI gets so good that my workflow becomes obsolete?

Real risk worth thinking about. The pattern that's held so far is that AI tools get easier to use, which lowers the floor on quality but raises the ceiling on what's possible. Design tools got dramatically easier from the 1990s through the 2010s and designers didn't disappear — they moved up the value chain. The work that's most exposed is pure execution of known-good tasks. The work that's least exposed is anything involving client conversations, strategic thinking, taste judgments, and post-production polish. Optimize toward the second.

How do I price my first project?

Start with the lowest end of the ranges in this article (e.g., $300 for an automation setup, $500 for a basic content system buildout) and raise it after the third paying client. Don't undercut yourself permanently to get the first project; clients who buy on price will leave on price. Better to ship one $500 project than five $50 ones.

Should I disclose to clients that I'm using AI?

Yes. Always. I won't argue this one — disclosure is the long-term play and the moral one. The disclosure doesn't have to be in the marketing; it does need to be in the conversation. "My workflow is hybrid AI-and-human. The AI handles drafting and iteration, I handle editorial, strategy, and the final pass." Most clients will appreciate the clarity. The ones who don't are not your clients.

What if I want to make money but I don't have a niche or skill yet?

Pick the automation consulting path. It requires the least pre-existing creative skill, the demand is highest, the income is highest, and the work is teachable. Start by automating one workflow for a friend's small business for free or cheap, document it, and use that as your first portfolio piece. The path from there to paying clients is shorter than people think.

The thing I keep coming back to

I started this article saying I get the same question over and over from creatives I know — "how do I actually make money with this AI stuff?" — and I want to end with the version of the answer I usually give in person.

The thing isn't the AI. The thing is the problem.

Anybody can sign up for the tools. The people who are making real money are the ones who figured out what painful, specific problem to solve, and then used AI to solve it cheaper, faster, or better than anyone else can. The tools are commoditized. The problem-finding and the solution-shaping aren't. Those are still creative work, and they pay accordingly.

If you spend the next three months learning every AI tool that exists, you'll be tired, broke, and no closer to a client. If you spend the next three months talking to ten small business owners and finding one painful problem you can solve, you'll have a side hustle.

I think it's actually that simple, and I think most of the noise online is making it sound more complicated than it needs to be.

Go find the problem.

← Back to Digest
career May 21, 2026

The Creative's Guide to AI Side Hustles That Actually Pay

Real numbers, real pricing, and real workflows for creatives building AI side hustles in 2026 — from $300 setup gigs to $20,000/month retainers, plus the mistakes that keep most people stuck on Fiverr.

The Man in Yellow Sunglasses
The Man in Yellow Sunglasses
Creative professional working on laptop with AI-generated artwork and financial charts showing income growth

I've been running automated content systems for months now, and I keep getting the same message from creatives I know: "How do I actually make money with this AI stuff?"

Fair question. I'm still figuring out parts of it myself.

The honest answer is that there are two side hustle worlds happening at the same time, and they barely overlap. One world is the Fiverr grind — $50 per blog post, $30 per logo, customers who think they're paying for the tool more than the person. The other world is a small group of freelancers charging $5,000 to $20,000 a month for AI automation consulting, plus designers running Midjourney workflows for clients who care about the final brand, not the prompt. Both groups use the same software. The income gap between them isn't about skill with the tool. It's about what they're selling.

This guide is for figuring out which side you want to be on, and what it actually takes to get there.

What the numbers say

Before we get into specific paths, two data points worth knowing.

Upwork's 2026 in-demand skills report shows freelancer earnings for AI-related work grew 109% year-over-year, and AI video generation/editing alone grew 329%. That's not a survey of what people want to do — that's invoices that got paid. The demand is real and it's still expanding.

The second number: practitioners running AI automation consulting for small and medium businesses are charging $5,000 to $20,000 per month on retainer, often after a $300 to $800 setup fee for the initial workflow build. Same software a hobbyist has access to. Wildly different price tag.

The gap is the entire subject of this article.

The Tool Trap (which I fell into for a while)

The most common way creatives stall on this is what I'd call the Tool Trap. It looks like this: you hear AI is a thing, you sign up for ChatGPT Plus ($20/month), then Midjourney ($30/month), then ElevenLabs ($22/month), then Runway ($12/month), then maybe Suno ($10/month). Now you're $94 a month deep before you've done a single client project, and you're trying to figure out which of those tools is going to make you money.

That framing is the problem. The tool doesn't make you money. A solved problem makes you money, and the tool is a thing you happened to use in the process.

When the question is "how do I monetize ChatGPT?", the answer is usually that you can't, because everyone else can monetize it too and the price collapses. When the question is "what painful problem does someone I can reach actually have?", the answer turns into a service business with paying clients. ChatGPT might still be involved, but it's a saw, not the product.

I fell into the tool trap when I first started messing with AI workflows in 2024. I subscribed to everything, built nothing, and wondered why nothing was working. The shift for me was realizing I had to start with the problem, not the software.

The five paths that are actually working

These are the five business models I'm seeing pay out in 2026. None of them are unique to me — they show up in the Upwork data, in Medium pieces by people who've tried multiple side hustles and ranked them, and in Reddit threads from people doing the work. I've listed them roughly in order of how much money they tend to generate, though that's never the only thing that matters.

1. AI Automation Consulting for small businesses ($5K–$20K/month)

This is the most lucrative path I've seen documented, and it's the one most creatives skip because it doesn't sound creative on the surface. It is, though — the work is essentially process design with software as the medium.

The business model: a local dentist, a real estate team, a wedding photographer's studio, a personal injury law firm — almost any small business with more than a handful of employees has workflows that are leaking time. Missed calls that don't get returned. Leads that come in through a contact form and sit unanswered for two days. Invoices that have to be manually chased. Customer follow-ups that everyone agrees should happen but nobody owns.

An AI automation consultant goes in, audits the leaks, and builds a system using Zapier or Make connecting the business's existing tools (CRM, Gmail, Calendly, Slack, QuickBooks) with AI components dropped in where they help — summarizing inbound emails, drafting follow-up messages, transcribing call notes into the CRM. Then they stay on retainer to maintain it and adjust as the business changes.

What it pays: practitioners are charging $300–$800 for the initial setup, then $1,000–$5,000/month for ongoing retainers per client. With three or four clients, that's a real income.

Why creatives are good at this even though it doesn't feel creative: small businesses don't need a developer to build this. They need someone who can sit with the owner, understand what's actually broken in their day, and design a system around it. That's interview skill, taste, and synthesis — which is exactly what designers and writers do all day.

Where people get stuck: trying to sell this as "AI consulting" instead of selling the outcome. Nobody buys AI consulting. They buy "I'll get your missed calls answered within 5 minutes" or "I'll cut your invoicing time in half." The AI is how you deliver, not what you sell.

2. AI-assisted design work (Midjourney + post-production)

This is what most people picture when they hear "creative AI side hustle," and it's a real path, but the version that pays is not the version most people are doing.

The version that doesn't pay: open Midjourney, type a prompt, hand the client a raw render, charge $50. You're competing with everyone else who has a $30/month Midjourney subscription, and the client can do that themselves.

The version that pays: a real client engagement looks more like this. Wedding invitations is a niche I keep seeing referenced. A designer takes the client's wedding details, their venue, their color story, runs hundreds of Midjourney variations across multiple prompt approaches, curates down to three concepts that actually work for the couple, then takes those into Photoshop to fix the things Midjourney can't do reliably — text rendering, exact color matching, print-safe resolutions, the small impossible-anatomy artifacts that show up everywhere if you don't look for them. The deliverable is finished, print-ready invitations and a guide for the matching menus, programs, and thank-you cards. The client pays for the final asset set, not the Midjourney generations it took.

What it pays: similar wedding invitation studios are charging $1,500–$5,000 per wedding package. A graphic designer named Sarah Chen has been cited as building this into a meaningful business by scaling with virtual assistants — though I want to flag that her specific income numbers are widely shared on YouTube but I haven't found independent verification. Treat the workflow as accurate, treat the eight-figure income claim as unverified.

Tools: Midjourney at the $30/month Standard tier is the workhorse (the unlimited relaxed-mode generations are what makes the hundreds-of-variations workflow possible). Photoshop for post. Maybe Runway ($12/month) if you're producing short animated invitations or "save the date" videos, which is a growing premium add-on.

The same pattern works in adjacent niches — children's book illustrations, restaurant menu design, real estate listing photography retouching, small e-commerce product imagery. Cristian R., a Fiverr Pro seller cited in their own marketing, has built a children's-book illustration practice around AI-assisted workflows. (His income isn't publicly listed, but Fiverr Pro status is at least a credibility floor.)

Where people get stuck: handing over raw outputs. The whole reason this pays is the curation, post-production, and brand discipline you add. Take that away and you're selling a $30 subscription back at retail.

3. Strategic content systems (for writers)

This is the one I have the most direct experience with, because The Daring Creatives is, in part, an experiment in running one. So fair warning that I have a horse in this race.

The unpaid version is what most people try first: write articles with ChatGPT, post them on Medium, hope something happens. It doesn't, because the content is generic, undifferentiated, and competing with millions of equally-generic pieces. As one Reddit thread I keep returning to put it: "Because everyone is using the same generic prompts, the internet is becoming a desert of boring, robotic content."

The version that pays is selling systems, not articles. A small business has a content problem — they need to publish consistently, in their voice, on topics that bring in customers, across blog plus email plus a couple of social platforms — and they don't have the time, the bandwidth, or honestly the writing instinct. They've tried ChatGPT themselves and the output was, as one client put it to me, "correctly punctuated nothing."

A strategic content system for that client looks like: an editorial brief on what topics actually map to their business, a brand voice guide so the AI drafts come back sounding like them, a draft-edit-publish workflow that uses AI for the first pass and you for the final pass, automation for distribution to the platforms they actually use, and a monthly review of what worked. You're charging for the system and the outcome (consistent published content that sounds like them), not the word count.

What it pays: writers running this model are reporting effective hourly rates around $120 ($7,200 for a project that took about 60 hours, in one Medium piece I'm not fully convinced is real but the pricing matches what I've seen in client conversations). On retainer, $1,500–$4,000/month for a fully managed content operation is realistic if the client is at the right size.

Tools: ChatGPT Plus ($20/month) is the floor. Beyond that, the tools depend on what the client actually needs — a CMS plugin, a scheduling tool, sometimes a custom voice-guide setup. If you want to learn the voice-guide piece, I've written about how to teach AI a brand voice in a separate piece, which is mostly the principle of "give it enough specific examples to imitate rather than telling it abstract rules."

Where people get stuck: trying to sell content packages by the article. The client doesn't want articles. They want results from publishing. Sell the result.

4. Custom tools for small businesses (the "vibe-coded" path)

This one is newer and underdiscussed, but I think it's going to be one of the biggest paths over the next year. Tools like Claude Code, Cursor, and v0 have made it possible to ship working software without writing code, provided you understand what makes software useful.

Most small businesses are stuck between two terrible options for the niche software they need. Option A: pay $200/month per seat for an enterprise SaaS product with 50 features, 47 of which they don't use. Option B: build a Frankenstein of Google Sheets and a Zapier automation and pray nothing breaks during tax season. Both options leave the business owner doing a job the software was supposed to do.

The third option — a $50–$200/month custom tool that does exactly the three things this specific business needs — has been mostly inaccessible because hiring a developer costs $5,000+ for the initial build. AI-assisted coding has collapsed that math. A creative who understands user experience can now scope a small-business tool, build it with Claude Code or a similar agent, test it with the client, and ship it. Then either charge a flat fee for the build plus a small monthly hosting/maintenance retainer, or position it as a tool you license to the client at a recurring rate.

Examples I've seen referenced: a check-in app for a small gym that handles member access and a class waitlist. A client timeline manager for a wedding planner that ties to their calendar and sends reminders automatically. An inventory tracker for a food truck that texts the owner when popular items are running low.

What it pays: this varies wildly because the niche is still settling. Practitioners I've seen report $2,000–$8,000 for a custom build (one-time or amortized into setup-plus-monthly), and $50–$200/month in ongoing hosting/maintenance.

Tools: Claude Code (subscription-based; pricing depends on plan) is the most capable of the AI coding agents I've used. Cursor and v0 are also worth knowing. You'll need a way to host whatever you build — Vercel, Railway, and Cloudflare all have free or near-free tiers for early projects.

Where people get stuck: thinking you need to be a developer. You don't, but you do need to understand what makes a tool useful versus frustrating, which is a creative skill, not a coding one. Build for someone you actually talk to. Don't build a generic tool for a generic market.

5. Voice, audio, and short-form video production

The 329% growth number in AI video generation isn't hypothetical demand. It's small businesses, podcasters, course creators, and short-form content creators discovering that they can get studio-quality voiceovers, full music tracks, and animated video segments for a fraction of what production used to cost — but they need someone who knows how to actually run the tools.

The business model: produce voiceovers, soundtracks, video intros, podcast post-production, audiobook narration, short-form social videos, or animated explainers as a service. The client gives you a script or a brief. You deliver finished, polished media.

What it pays: this depends heavily on what you're producing. Voiceovers tend to run $50–$200 per short script (5-minute YouTube intro, 60-second ad). Full podcast post-production retainers run $500–$2,000/month per show. Short-form animated videos run $200–$1,500 per video depending on length and complexity. Custom music tracks for a brand or a course run $100–$500 per track.

Tools: ElevenLabs ($5/month entry, $22/month Creator tier is the realistic floor for client work). Suno Pro at $10/month for music with commercial licensing on 500 songs/month. Runway Standard at $12/month/billed-annually for video generation. Pair them with whatever you already use for editing — Premiere, Final Cut, CapCut, DaVinci Resolve.

Where people get stuck: same place as design. Raw output, sold as a deliverable. The pay comes from production polish — editing the voiceover for breath and pacing, matching music tempo to the cut, post-producing the AI video to hide its weirdness. The AI is the rough draft. You're the finishing department.

The pricing shift nobody is teaching

Here's the move I see separating the two side-hustle worlds more than any tool choice or niche.

Stop billing hourly.

The data is from a Digital Applied report on freelance pricing trends, and it lines up with what I see in client conversations: hourly billing is collapsing for AI-powered services. The reason is structural. When you bill hourly, you're telling the client that what you sell is time. Time, in their head, means "this person is interchangeable with anyone else who can run the same software." Hourly billing makes you a commodity by definition.

Project pricing and retainers tell the client a different story: "this person is solving a problem I have, and the value of the solution is the price." The freelancers charging $5,000/month for AI automation aren't doing twenty times the work of someone charging $50 for a blog post. They're solving a much more expensive problem (business process inefficiency that's costing the client thousands in lost revenue), and the client knows it.

Practical version of this for any of the five paths above:

  • Setup fees for the initial buildout (the discovery, the configuration, the testing). One-time, project-priced.
  • Monthly retainers for ongoing service (maintenance, adjustments, optimization, support). Recurring.
  • Value-based pricing when you can quantify the savings or revenue you're producing. "This automation is saving you $4,000/month in missed-call follow-ups. The retainer is $1,500." Math the client can verify.

I won't pretend this transition is easy. Most creative work has been hourly-priced for so long that even good clients default to "what's your rate?" You have to gently redirect that conversation to "here's the package, here's what's included, here's the outcome." That redirect is itself a skill. Some clients will push back. You let them go.

The $100/month stack vs the $500/month stack

A practical breakdown of what you actually need to spend on subscriptions, because the tool trap is real and expensive.

The $100/month starter stack:

  • ChatGPT Plus — $20/month. Floor for almost any text-based work.
  • Claude Pro — $20/month. Worth having in addition to ChatGPT for code, longer documents, and a different voice. Optional if budget is tight, but I keep coming back to it.
  • Midjourney Standard — $30/month. Required if you're doing any design-heavy work; skippable if you're a writer.
  • ElevenLabs Starter — $5/month. Only if you're doing any voice/audio.
  • One automation tool — Zapier (free tier or $20/month Starter) or Make (free tier or $9/month Core).

Total: $75–$95 if you're disciplined. And critically: pick three of these based on your actual niche, not all five.

The $500/month "I'm running a real practice" stack:

  • Everything above, plus:
  • Midjourney Pro or Mega — $60–$120/month if image generation is your primary work
  • Runway Pro — $35/month for serious video work
  • Suno Pro — $10/month if you do music
  • Notion or Obsidian + a CRM (HubSpot has a generous free tier) — $0–$50/month
  • Hosting/automation infrastructure — $50–$100/month depending on how much custom tooling you've built
  • Domain + email — $15/month if you're running this as a real business
  • An assistant or two — variable, but virtual assistants on Upwork start at $5–$15/hour and pay for themselves quickly if you've got too much execution work

The hidden trap is that you don't need both stacks. You need the stack for what you actually do. Designers don't need ElevenLabs. Writers don't need Midjourney Mega. Automation consultants barely need Midjourney at all. Audit your AI subscriptions every quarter and kill anything you haven't opened in 30 days.

How to pick which path is for you

Honestly the most useful question I've found is just: what kind of conversation are you most naturally good at?

  • If you're good at sitting with a small business owner and untangling what's actually broken in their day → automation consulting
  • If you're already good at visual design and have a critical eye for output → AI-assisted design
  • If you write well, think structurally, and like editing more than first drafts → strategic content systems
  • If you understand what makes software useful and you like solving specific problems → custom tools
  • If you have an audio/video ear and you're patient with post-production → voice and short-form media

You can pick more than one over time, but pick one to start. The single biggest mistake I see is trying to be the "AI services" generalist who does a little of everything. Pick a path, get good at one workflow, and let the work compound before adding a second.

A few notes on getting started

Don't start on Fiverr or Upwork if you can avoid it. The platforms work, but they push you toward commodity pricing. Direct outreach to local small businesses — actual conversations, in-person or via email — produces better clients at higher rates. Contra is gaining traction as a commission-free alternative for project-based AI work if you want a platform.

LinkedIn and X (Twitter) work for visibility, but only if you're posting evidence of the work, not opinions about AI. Show before-and-after on a real project. Show a workflow you actually use. Show a client outcome with a number attached. Posts that say "AI is changing everything" don't get clients. Posts that say "I cut this firm's invoicing time from 4 hours/week to 20 minutes — here's the workflow" do.

Be honest about what's human and what's AI. I think this matters more than people realize. Clients are getting suspicious of pure AI output, and rightly so. The freelancers I see winning long-term are the ones who openly explain their hybrid AI workflow: "AI does the bulk drafting, I do the editorial pass and the client communication, here's the split." That transparency turns into trust. Pretending the work was all you when it wasn't is a slow-burning credibility problem.

Defend people learning in public. I know I keep saying this. The AI-creativity space has a lot of gatekeepers who pile on beginners using AI tools "wrong" — wrong prompts, wrong workflow, output that looks too obviously generated. Ignore them. Their target should be the people scamming with AI, not the people learning. If your work looks AI-generated at first because you're new, that's fine. The fix is more reps, not more shame.

Frequently asked questions

How much can a beginner realistically make in the first 3 months?

Honest answer: usually somewhere between $0 and $2,000/month in your first quarter, with most of the variance depending on whether you're doing direct outreach or relying on platforms. Plan for $0 the first month while you set up. By month 3, if you've talked to 20+ small businesses or shipped 5+ portfolio pieces, you should have at least one paying engagement.

Do I need to learn how to code?

No, but it helps for the custom tools path. The other four paths don't require coding at all. Even on the custom tools path, AI coding agents like Claude Code mean you can ship working software without writing code in the traditional sense — though you do need to understand what good software feels like to use.

Which AI subscriptions should I buy first?

ChatGPT Plus ($20/month) if you're not sure. It's the most generally useful starting point and the one tool that touches every path. Add a second subscription only when you have a specific need it solves — Midjourney if you've gotten a design lead, ElevenLabs if you've gotten a voice lead, and so on. Don't subscribe ahead of demand.

Is the side hustle market saturated?

The market for generic AI output (basic blog posts, templated logos, raw voiceovers) is saturated. The market for solving real business problems using AI as part of a workflow is wide open. The freelancers complaining about saturation are usually competing on the saturated end. The income gap between the two markets is in the data — Upwork's 109% YoY earnings growth doesn't fit a saturated market.

What if AI gets so good that my workflow becomes obsolete?

Real risk worth thinking about. The pattern that's held so far is that AI tools get easier to use, which lowers the floor on quality but raises the ceiling on what's possible. Design tools got dramatically easier from the 1990s through the 2010s and designers didn't disappear — they moved up the value chain. The work that's most exposed is pure execution of known-good tasks. The work that's least exposed is anything involving client conversations, strategic thinking, taste judgments, and post-production polish. Optimize toward the second.

How do I price my first project?

Start with the lowest end of the ranges in this article (e.g., $300 for an automation setup, $500 for a basic content system buildout) and raise it after the third paying client. Don't undercut yourself permanently to get the first project; clients who buy on price will leave on price. Better to ship one $500 project than five $50 ones.

Should I disclose to clients that I'm using AI?

Yes. Always. I won't argue this one — disclosure is the long-term play and the moral one. The disclosure doesn't have to be in the marketing; it does need to be in the conversation. "My workflow is hybrid AI-and-human. The AI handles drafting and iteration, I handle editorial, strategy, and the final pass." Most clients will appreciate the clarity. The ones who don't are not your clients.

What if I want to make money but I don't have a niche or skill yet?

Pick the automation consulting path. It requires the least pre-existing creative skill, the demand is highest, the income is highest, and the work is teachable. Start by automating one workflow for a friend's small business for free or cheap, document it, and use that as your first portfolio piece. The path from there to paying clients is shorter than people think.

The thing I keep coming back to

I started this article saying I get the same question over and over from creatives I know — "how do I actually make money with this AI stuff?" — and I want to end with the version of the answer I usually give in person.

The thing isn't the AI. The thing is the problem.

Anybody can sign up for the tools. The people who are making real money are the ones who figured out what painful, specific problem to solve, and then used AI to solve it cheaper, faster, or better than anyone else can. The tools are commoditized. The problem-finding and the solution-shaping aren't. Those are still creative work, and they pay accordingly.

If you spend the next three months learning every AI tool that exists, you'll be tired, broke, and no closer to a client. If you spend the next three months talking to ten small business owners and finding one painful problem you can solve, you'll have a side hustle.

I think it's actually that simple, and I think most of the noise online is making it sound more complicated than it needs to be.

Go find the problem.

// LEXICON_CITY_DISPATCH_REQ
// STATUS: CONNECTION_STABLE
// SOURCE: CENTRAL_DISPATCH_HQ

SHERMAN UPLINK: "I'm at HQ holding down Central Dispatch. Enter your query below to pull relevant data records and I'll see what data cards we've recovered!"