Gbenga Akindolani |AI Engineer | Automation Expert

If you're in the automation space, whether you're building in n8n, Make.com, or Zapier, your ability to connect APIs and move data around is just the baseline. That's not what makes you a professional. And if you're just starting out, learn these principles early. You'll be ahead of 80% of people offering automation services. Here's the framework I follow for every project: 1. Understand the client's objectives before you touch anything Before I build a single node, I need to understand: What business problem are we solving? What does success look like? What are the expected inputs and outputs? What's the data volume? Who will interact with this workflow? What happens when something breaks? Getting this wrong means rework later, or worse, delivering something that technically works but doesn't solve their actual problem. If you're new, create a discovery call template with every question you need answered before starting. This saves you from scope creep and awkward conversations. 2. Build clean, structured workflows Your workflow should tell a story. Anyone looking at it should follow the logic from trigger to output without needing you to explain it. This means logical node arrangement flowing left to right, related operations grouped together, clear spacing between sections, and consistent naming conventions (not "HTTP Request 1," "HTTP Request 2). A well-structured workflow takes under a minute to understand. A messy one still confuses you after 10 minutes. If you're starting out, slow down. Don't just make it work. Make it readable. 3. Optimize before you call it done I constantly see workflows where people chain 4 or 5 nodes to do something simple. A Merge node, then Aggregate, then Split, then Filter, all to transform some data. A single Code node with 10 lines of JavaScript does the same thing faster and cleaner. Optimization matters because it reduces execution time, lowers operational costs (most platforms charge per operation), makes debugging easier, and reduces failure points. Another thing: stop throwing AI agents into workflows that don't need them. Workflow automation is deterministic. If X happens, do Y. Predictable and reliable. AI automation involves language models making decisions or generating content. Both have their place. But if you're using an AI agent to route leads based on a dropdown value, you're overcomplicating things. A simple Router node does that better, faster, and cheaper. 4. Add error handling. Always. A workflow without error handling is a ticking time bomb. APIs go down. Rate limits get hit. Data comes in with unexpected formats. Tokens expire. If your workflow fails silently, your client won't know until the damage is done. Maybe leads stopped syncing to their CRM for three days. Maybe invoices weren't sent. Every workflow I build catches failures at critical points, logs errors with context, notifies the client immediately via their preferred channel, and implements retry logic where possible. If you're starting out, make error handling part of your standard process. Don't treat it as an afterthought. 5. Use sticky notes and node grouping Every major section should have a sticky note explaining what it does. Related nodes should be grouped together. A group labeled "Lead Enrichment" with your Apollo or Clearbit calls. A group labeled "CRM Sync" with your HubSpot operations. A sticky note explaining "This section handles retry logic." This helps clients understand what they're looking at (most aren't technical), and helps anyone maintaining the workflow later, including you six months from now when you've forgotten the details. Make your workflows self-documenting. 6. Provide external documentation This is what separates professionals from amateurs. And what most freelancers skip entirely. I deliver two documents with every project: Client Documentation is written for business users. It covers what the workflow accomplishes, how to trigger it, expected inputs and outputs, basic troubleshooting, and who to contact if something breaks. No technical jargon. Plain English anyone in the organization can understand. Developer Documentation is written for technical teams. It covers the architecture, all API connections, authentication methods, data schemas, error handling logic, edge cases, and step-by-step instructions for modifications. Why does this matter? Clients don't want to be dependent on you forever. They want to understand what they paid for. By providing documentation, you give them independence. And paradoxically, this builds trust and leads to more work. Here's the business case: this is a value-add you can charge extra for. When I pitch projects, documentation is a line item. I explain the difference between client and developer docs, who uses each, and why it matters. Most clients immediately see the value. They've been burned before by automation nobody understands. This alone can add 15-25% to your project value. 7. Hand over properly When I complete a project, I provide a live walkthrough, documentation files, and a Loom video for future reference. I set these expectations from the start so clients know what to expect, leading to clean project closures, testimonials, and referrals. If you're new to automation, remember to focus on quality, not just connecting nodes. Structure, optimize, document, and hand over your workflows properly. This way, you won’t just compete on price, but on the quality of your solutions. Happy New Year! 🎉 This year, I'll be sharing more tips on leveling up your automation skills and how to actually land clients as a freelancer. Follow along if you don't want to miss out.

Huge risk but worth it. Rooting for you mate

Same way Upwork laid off 25% of their workforce, claiming the need for smaller AI-driven teams. Now most freelancers are quitting and giving up due to high extortion of connects without landing a gig. Ngl, Tech isn’t the way it used to be like before in terms of Job hunting. Before, with 100 connects you could apply for up to 10 jobs and could convert up to 2 jobs. It is what it is. This is not me discouraging anyone because while the majority are complaining, some are still landing gigs back to back. Ciao

AI is making me less smarter but it helps me solve incredibly difficult problems. 💔

In a Greenhouse survey of 2,950 job seekers, 63% had been interviewed by an AI, yet 38% have walked away from a hiring process because it included one. Candidates reported nearly identical bias from AI and human interviewers: 36% felt age bias and 27% race or ethnicity bias from both, and 70% weren't told AI was involved. No matter how good the knowledge base is, the AI will still hallucinate at some point. Now picture it misunderstanding your answer, correcting itself halfway through the interview, and you're sitting there wondering if the wrong version is what got scored.

Yesterday my git push to GitHub kept failing, and Claude Code in VS Code kept dropping. Turns out it was a macOS Tahoe bug, once your Mac runs 49+ days without a restart, network connections start breaking. Mine had been up 112 days. One restart fixed everything. Crazy!

So many crazy things have been done with automation. I remember when @The_Christophxv got a client that wanted to spy on his girlfriend’s WhatsApp and wanted it automated so that any message sent to her, he would receive it, and any response sent out, he would receive it.