The Actual Automation Stack a Solo Founder Uses (Not Your Fantasy)

The fantasy automation stack: OpenClaw agents running overnight while you sleep, building products in the background, scaling infinitely.

The actual automation stack: Notion AI, Claude, Figma Make, Copilot 365, Make.com, and a human at every decision point.

Ali Parandeh runs Build Your AI solo. He’s an engineer who thinks about production systems, cost efficiency, and safety. His stack reflects that: practical, integrated, and deliberately incomplete. He’s not using the hottest new agent framework. He’s using what works.

The Foundation: Claude + Notion AI

Claude is the workhorse. “I use Claude for most of my workflows,” Parandeh explains. “Whether I’m planning, writing, designing, or building, Claude helps me think through problems.”

But Claude alone is a context-switching tax. You have to context-switch to the Claude interface. Notion AI removes that friction.

“Notion allows me to install and configure custom AI agents. It comes with pre-configured integrations to Google Docs, Google Workspace, Microsoft 365, and you can add your own MCP and customize the system prompt. It’s got all the security infrastructure in place — ISO 27001, SOC 2 compliance.”

This is the pattern in his stack: convenience paired with safety guardrails. Notion agents are less flexible than OpenClaw, but they’re faster to deploy and you don’t end up with an $800 overnight bill. As a solo founder, that tradeoff is obvious.

For serious thinking work — planning products, designing workflows, iterating on copy — Claude still gets the job. Notion AI handles the quick wins in his existing tools.

Design & Presentation: Figma Make

Most founders outsource design or use templates. Parandeh designs his own, which sounds inefficient. It’s the opposite.

“I do design user interfaces, presentations, all my training slides in Figma. Figma allows me to adjust the tone of content on the fly, generate presentation notes, generate images for slides. It comes with really good templates.”

The core tool is Figma Make, Figma’s AI-powered design assistant.

“Figma Make allows me to produce really good-looking user interfaces for prototypes when I’m showing a vision to clients. Or when I’m iterating on product ideas with users, I can send them mockups immediately. It’s fast feedback.”

He built his entire website with this pipeline: Notion AI for copy iteration, a tool called Relume for layout, Figma for design, JetBrains for animation code. The AI generates, he refines, and the output has his signature on it.

Workflow Automation: Make.com + Copilot 365

When a task is repetitive and the input/output is clear, you don’t need OpenClaw. You need Make.com.

“Workflow automation tools like Make.com or Zapier let you automate with a graphical interface. You don’t need technical expertise. Or you can ask Claude to produce a script if you need something custom.”

Parandeh’s philosophy: use the simplest tool that solves the problem. Make.com for multi-step workflows. Custom agents in Copilot 365 for specific knowledge tasks that live in Microsoft 365.

“Microsoft Copilot 365 has custom Copilots. I’ve created my own for certain scenarios. The moment it’s not producing the same level of quality as third-party tools, that’s when I evaluate alternatives. But Microsoft’s got legacy infrastructure, Microsoft Graph API, and tools like Copilot Agent Studio that lets you build custom agents.”

This is the reality: most solo founder work lives in email, documents, spreadsheets, and notes. Those tools now have integrated AI that’s good enough. You don’t need a standalone agent framework.

Video & Creative: CapCut AI + Vidonary

Parandeh learned video editing in CapCut to make social media content. He uses CapCut AI for the heavy lifting: script optimization, subtitle generation, transitions.

“Even my website was built with AI. I spent two or three days with Motion AI iterating on copy to make sure it’s tight and knit. It follows a 21-letter sales letter framework.”

This reveals the pattern: AI handles the volume work (variations, iterations, options), he handles quality control and direction.

The Product Example: DIY Quiz

The best example of his stack in action is the quiz on his website. He needed a lead magnet scorecard. Instead of buying Scorecard SaaS and paying a subscription, he built it himself.

“I created a quiz with HTML, CSS, and Claude did it all with my design system. I copy-pasted the entire CSS from my website into Claude and said ‘build me a quiz with reporting.’ The questions use a framework Daniel Priestley teaches: ten questions on best practices, ten on current reality, ten on ideal reality. It gives you a report.”

He learned the framework. Claude generated the questions. He edited the questions. He deployed the quiz.

“The most valuable thing is the frameworks and the way you think. It’s not the coding part. That’s where AI helps — it’s not replacing the thinking. It’s letting me move faster on the execution without losing the thinking.”

Why This Stack, Not a Unified Agent Platform?

Parandeh’s stack is fragmented by design. Different tools for different jobs. OpenClaw would centralize it, but at what cost?

“I didn’t find the need to purchase a Mac mini and set up all those infrastructure requirements. I’m happy with tools I already have access to. I’m quite happy with those.”

The hidden cost of OpenClaw — safety guardrails, monitoring, infrastructure — is time he’d rather spend on the product and the business.

“For most business owners and engineers, you just need a general assistant that can do a lot of your workflows. More advanced use cases, you go down to workflow automation tools. You don’t necessarily need OpenClaw unless you’ve thought through the security layer and cost implications.”

The Quality Control Layer

The thread connecting every tool is human judgment. “If you just get AI to do the entire pipeline, you’re not going to get a hundred percent. You’re going to get seventy, eighty percent. Most of the time you won’t get something you like. It won’t have your signature on it.”

He quality-controls at every stage. Notion AI suggests, he refines. Figma Make generates layouts, he adjusts tone. Make.com runs workflows, he monitors them. Claude generates code, he reviews it.

“You design the pipeline and get the tools to add the work, but you quality-control the outputs between each stage. That’s a manufacturing pipeline. You’re in the middle.”

This is the real lesson. The automation stack isn’t a set of tools. It’s a workflow where AI handles the volume and you handle the decisions. Tools matter, but the discipline of staying in the loop matters more.

For a solo founder, that loop is the whole business model. It’s what prevents the $800 bill. It’s what keeps the work from looking like everyone else’s. It’s what scales you from good to great.

FAQ

What’s the fastest way to set up an automation stack as a solo founder?

Start with what you already use: email, documents, notes. Add Notion AI or Copilot if you’re in Microsoft 365. Both have custom agents built in and don’t require extra infrastructure. For complex workflows, use Make.com. For one-off tasks, use Claude. You don’t need OpenClaw unless you’ve hit the limits of these simpler tools and have time to manage the safety layer.

Should I use Make.com or Zapier?

Make.com is slightly easier to learn and has more integrations. Zapier is more established. For solo founders, either works. The difference is small. Pick one and learn it well instead of bouncing between both.

Can I replace my entire stack with OpenClaw?

You could, but you’d spend more time building and monitoring safety guardrails than you save on setup. For solo work, the simpler approach — Notion agents, Make.com, Claude for thinking — is faster and cheaper. OpenClaw makes sense if you have a specific use case (agents that need local control, heavy integration with custom systems) and the engineering time to maintain it.

What does Parandeh use for writing and copy?

Claude for first drafts and iteration. Notion AI for quick refinement inside documents. For website copy, Motion AI (which is really Claude optimized for marketing). He spends a few days iterating with the tool to nail the framework and tone, then stops. The copy has his thinking in it.

How do I know when to use Make.com vs. Copilot vs. OpenClaw?

Make.com: when you have a repetitive multi-step workflow with clear inputs and outputs. Copilot: when the task lives in Microsoft 365 and doesn’t need custom integrations. OpenClaw: when you need local control, don’t have a paid subscription budget, and have engineering bandwidth to handle safety. For most solo founders, Make.com and Copilot cover everything.

Does the Figma Make pipeline actually save time?

Yes and no. It saves time on iterations — Figma Make generates multiple design directions quickly. But Parandeh spends time reviewing and refining each one. The real win is that he doesn’t need to hire a designer. That’s the cost he’s avoiding.

Why didn’t Parandeh build his website with a no-code tool?

He wanted control over the copy and the design system. He used Relume (template + AI layout), Figma Make (design), and JetBrains (code customization) because that gave him leverage and quality. A no-code tool would have been faster but less flexible.

Can I use this stack to build a SaaS product?

These tools handle supporting work: copy, design, automation, admin. For the product itself, you still need real engineering. If you’re a developer, you use Claude to move faster. If you’re not, you use Parandeh’s approach: outsource or use no-code with as much AI assistance as possible, and quality-control at every stage.

What’s the monthly cost of this stack?

Claude Pro ($20), Microsoft 365 or Notion Plus ($10-15), Make.com ($10-20 depending on usage), Figma ($12), CapCut Pro ($4.99). Maybe $60-70 a month if you use everything. Much less than hiring an assistant or buying specialized tools.

Why does the quiz example matter?

It shows the workflow: learn a framework (10 + 10 + 10 questions), generate variations with AI, edit and refine, deploy. The result looks custom and professional because Parandeh did the thinking. AI handled the generation. That’s the pattern that scales — AI for volume, you for judgment.

Is Parandeh’s automation “real” or is he just using marketing tools?

It’s real — it’s genuinely saving him time on repetitive work. But it’s not the autonomous agent fantasy. It’s human-in-the-loop at every step. That’s the practical version of automation that works for solo founders. Full autonomy would cost him safety and quality.