The AI Tools That Actually Run My One-Person Agency

Every week I see another "Top 25 AI Tools for Freelancers" listicle. They all look the same. Beautifully formatted tables with logos, pricing tiers, and feature checkmarks that tell you absolutely nothing about whether the tool actually works in the real world.

I tried most of them. Over six years of running a one-person web development and SEO agency, I've signed up for, paid for, and abandoned more AI tools than I can count. The ones that survived the first month? Even fewer. The ones I still use daily? One.

Yeah. One.

This isn't a listicle. This is what my actual workflow looks like, right now, in 2026. No affiliate links, no sponsored sections, no "use code KEMAL for 20% off." Just the honest version.

Everybody Has a List. Nobody Shows Their Actual Workflow.

Here's the thing about AI tool recommendations: they're almost always written by people who tried the tool for a week, or worse, never used it at all. They read the landing page, watched the demo, and wrote 2,000 words about features they never tested on a real project.

I get it. Listicles rank well. They're easy to produce. And 77% of freelancers have adopted AI tools by 2026, so the audience is massive. But there's a difference between adopting a tool and actually depending on it. Between trying something for a Twitter thread and building your entire business on it.

My constraint is simple: I'm one person. No team to delegate to. No junior dev to hand off the boring parts. No VA to manage the admin. Every tool I add needs to earn its place every single day, or it's gone. That filter is brutal, and it should be.

The Tool That Changed Everything: Claude Code

I only use Claude. That's it. No ChatGPT, no Copilot, no Cursor, no Gemini. Just Claude with a Max plan.

I know that sounds extreme. It felt extreme at first. But here's what happened: I stopped spreading my attention across five different AI tools that each did one thing okay, and instead went deep with one tool that does everything I need.

The way I work now: I've created multiple local project folders on my machine, each with custom skills built for a specific part of my business. One folder handles bookkeeping and talks directly to the Lexware API. Another handles content ideation, SERP research, writing, publishing, and even Google indexing notifications. Another is for client development work. Each folder is its own self-contained AI workspace with custom tools and automations.

Think of it like this. Claude is like Amazon Basics. Amazon sees the products that third-party sellers sell that actually work, finds the supplier, and sells them under their own brand. Claude does the same thing with AI tooling. They see what people build on top of the platform, what workflows actually stick, and then implement the best patterns natively. MCP servers, custom skills, tool use, agentic coding. All things that used to require stitching together five different products.

That's why one tool is enough. Not because I'm lazy. Because one deep tool beats five shallow ones.

Is it perfect? No. I've written about how Claude is great at building software and also great at breaking it. It hallucinates. It sometimes refactors code you didn't ask it to touch. It confidently suggests solutions that don't work. Spec-driven development is what made AI coding actually reliable for me, giving Claude a clear spec to work against instead of letting it freestyle.

But the productivity gain is real. I once let Claude Code cook for 20 minutes on a weekend project redesign and got what would have taken me a full day. Freelancers using AI report cutting development time by 60-70%. That tracks with my experience.

The Boring Stuff That Actually Matters: Automation Nobody Sees

Here's the part nobody puts in their AI tools article: the most valuable thing AI does for my agency is the stuff nobody will ever see.

My bookkeeping folder talks directly to the Lexware API. I can pull invoices, check payment statuses, and reconcile accounts without opening a separate app. My content folder handles the entire pipeline: generate ideas from Reddit and SERP data, research the topic, write the draft, publish to Directus CMS, import images, create the German translation, and submit the URL to Google's indexing API. All through Claude.

No Zapier. No Make.com. No fancy SaaS automation platform with a $50/month subscription and a visual workflow builder that breaks every time an API changes.

Just Claude plus direct API calls. That's the entire automation stack.

The unsexy truth is that 91% of SMBs using AI report revenue increases, but those gains don't come from flashy AI demos or chatbots on your website. They come from operations. From the boring, repetitive stuff that used to eat three hours of your week and now takes three minutes. Like the tailor who can't sew his own pants, I spent years building fancy automations for clients while running my own operations manually. AI fixed that.

SEO Without a Team (and Without Losing Your Mind)

Solo SEO is brutal. Keyword research, content strategy, technical audits, link analysis, competitor monitoring. Each of those is a full-time job at an agency. I'm supposed to do all five plus the actual web development work.

AI compresses the research phase dramatically. What used to be a full afternoon of pulling keyword data, analyzing SERPs, and reading competitor content now happens in minutes. I can have Claude analyze a SERP, identify content gaps, check People Also Ask questions, and draft an outline before I've finished my coffee.

But here's my honest take: AI-generated SEO content is a race to the bottom. Everyone has access to the same tools. Everyone can generate the same 2,000-word "Ultimate Guide to Whatever" in ten minutes. The content all sounds the same because it is the same.

The value isn't in letting AI write your content. It's in letting AI do the research and structure, then writing with real experience that no AI can fake. Marketing agencies lead AI adoption at 63%, which means the bar for generic AI content is about to drop to zero. The only thing that'll rank is stuff that's genuinely useful, genuinely personal, and genuinely different.

What I Tried and Dropped (Everything Except Claude)

Let me be transparent about what didn't work.

I tried dedicated AI writing assistants. They produced content that sounded like it was written by a committee of people who had never actually done the thing they were writing about. Generic, safe, forgettable.

I tried AI project management tools. They added complexity instead of removing it. Another dashboard to check, another inbox to monitor, another thing sending me notifications about notifications.

I tried AI image generators for client work. The results were either unusable or required so much manual correction that I might as well have just sourced the image traditionally.

I tried AI chatbots for client sites. None of them were good enough to put in front of a client's customers without embarrassing both of us.

Dropped them all. Every single one.

The lesson: every AI tool you add is another thing to maintain, another subscription to track, another workflow to remember, another potential point of failure. 82% of very small firms believe AI isn't applicable to them. I think the real problem is that most AI tools aren't applicable to how small firms actually work. One good tool used deeply beats five mediocre tools used shallowly.

The Real Competitive Advantage Isn't the Tools

There's a stat floating around from an Indie Hackers survey that solopreneurs using AI agents saw a 340% average revenue increase. I'd take that with a massive grain of salt. Survivorship bias is real, and the people answering surveys about AI success are probably not the ones who tried it and failed.

But the underlying point is valid: AI is a multiplier. It multiplies what you already have.

If you know your niche deeply, AI makes you faster at delivering in that niche. If you have strong client relationships, AI gives you more time to invest in those relationships. If you're good at spotting problems, AI helps you solve them faster.

If you don't have any of that? AI just helps you produce mediocre work at scale. Which is worse than producing it slowly, because now there's more of it.

The year-end review that changed how I run my agency wasn't about tools. It was about knowing which clients to keep, which services to offer, and which direction to grow. AI didn't tell me any of that. Experience did. AI just made the execution faster once I knew what to execute.

The real competitive advantage for a one-person agency isn't which AI tools you use. It's speed of decision-making. No committees. No approval chains. No stakeholder alignment meetings. A client asks for something on Monday, and it's done by Tuesday. Try doing that with a 10-person team and a project management tool.

My Actual Daily Stack (The Honest Version)

Here's everything I use. The full list.

• Claude Max plan with custom skill folders (development, bookkeeping via Lexware, content/SEO pipeline, and more)
• Directus as my headless CMS
• Next.js + Coolify on Hetzner servers for frontend and hosting
• Pixabay for stock images when I need them

That's it. No Zapier. No Make. No ChatGPT. No Copilot. No Notion AI. No Jasper. No Surfer SEO.

Total monthly cost for the AI side: under $100. Compare that to the $3,000-$12,000/year that most AI tool stacks cost, or the $24,000-$60,000/year for a virtual assistant.

The stack is small. That's the point.

Fewer tools used deeply will always beat more tools used badly. I'd rather know one tool inside and out, build custom workflows that fit exactly how I work, and never wonder which of my seven subscriptions just broke the integration with the other six.

If you're running a solo operation or a small team and you're drowning in AI tool subscriptions that all promise to "10x your productivity," maybe the answer isn't adding another one. Maybe it's picking the best one and going all in.

That's what I did. It's working. I'm not saying it'll work for everyone. But if you're curious about what a real solo agency stack looks like in 2026, now you know mine.

If you want to talk about building a setup like this, or you're a business owner wondering what AI can actually do for your specific situation, let's talk. No pitch deck. Just an honest conversation about what's possible.