Claude Code Singapore · Real Projects · Non-Technical Builds
The best way to understand what Claude Code can do is to see what people with no coding background have actually shipped. Not hypothetical use cases, not screenshots from a demo — real things that are running today, built by professionals who couldn't explain the difference between JavaScript and Python if you asked them.
Here are five real builds from non-technical professionals in the Cowork SG community. For each one, we cover what they built, how long it took, what they struggled with, and what surprised them along the way. The details have been edited slightly to protect privacy, but the builds are real and so are the outcomes.
None of these professionals could explain what a for-loop is. All of them shipped working software.
Sarah runs a one-person HR consultancy focused on SMEs. Her work involves a lot of proposal writing — scoping statements, engagement structures, deliverables, fees. Every new client needed something customised, but the underlying structure was about 80% the same across all of them: an introduction, a diagnosis section, a proposed approach, a timeline, and a fee table.
The problem was that she was rebuilding that 80% from scratch every single time, or copying and pasting from old proposals and spending 45 minutes cleaning up references to the previous client. It was the kind of repetitive, skilled task that felt beneath the level of thinking she wanted to be doing — but it also couldn't be entirely delegated because the framing still required her judgment.
What she built with Claude Code: a simple web form that sits on her laptop. She inputs the client name, industry, the scope of work she's agreed to, the budget range, and the timeline. She clicks Generate. The tool passes those inputs to Claude, which writes the full proposal body in her voice — with her standard structure, her typical phrasing, her formatting preferences — and produces a PDF she can review and send.
The setup took a full Saturday afternoon, including the time to get Claude Code installed and learn how to navigate it. The actual tool was built in about two hours of that session. She spent the rest of the time testing it, giving Claude feedback on the tone and structure, and refining the prompt it uses to generate proposals until it matched her standards.
"It debugged itself. I didn't have to do anything except describe what I wanted and approve the changes. I expected it to feel like programming. It felt like briefing a very attentive assistant."
She now sends out proposals in under 15 minutes per client. The time she's saved over three months has more than covered the cost of her Anthropic API credits many times over.
Marcus runs a four-person digital marketing agency in Singapore. His clients are SMEs and mid-sized local brands. Every Friday, his team sends performance reports to each client — covering campaign results from Google Ads, Meta Ads, and Google Analytics. At its worst, the reporting process was taking 2 to 3 hours per client. With six active clients at any given time, that meant up to 18 hours of Friday afternoons disappearing into copy-paste, chart-making, and number-formatting.
He knew the process could be automated. He'd been quoted $12,000 by a developer to build a custom reporting dashboard. He decided to try Claude Code instead.
What he built: a script that runs every Friday at 4.30pm. It connects to the Google Analytics API and the Meta Ads API, pulls the week's performance data for each client, feeds it into Claude with a brief, writes a plain-English performance summary in the agency's reporting voice, formats everything into a clean HTML email template, and sends it to each client automatically at 5pm.
The build took two evenings spread across a week. The first evening was mostly getting the Google Analytics API key set up — navigating Google's developer console, granting the right permissions, and testing that the data was flowing correctly. The second evening was refining the summary language and getting the email formatting right.
"I thought I'd need a developer. The hardest part was getting the Google Analytics API key sorted — not the coding. Claude Code handled every line of code itself. I just described what I wanted and answered its questions."
His Friday afternoons now belong to him. His clients receive their reports reliably at 5pm every week, often with better written summaries than the agency used to produce manually.
Priya had been running her executive coaching practice for four years out of a Wix website she'd built herself when she first started out. It was functional, but dated. The design hadn't aged well, the load speed was poor, and she'd been quoted $6,000 to $9,000 by web design agencies to rebuild it properly. She'd been putting it off for 18 months, telling herself she'd sort it out "when things slowed down".
Things never slowed down. She decided to try Claude Code after seeing another Cowork SG member share their experience in the community group.
What she built: a clean four-page website — a homepage, a services page, a testimonials page, and a contact page. The homepage includes a lead capture section offering a free career planning worksheet (a PDF she'd already written) in exchange for an email address. The contact form is functional and sends enquiries directly to her inbox. The design is professional, fast-loading, and fully mobile-responsive.
She spent Saturday on the design and structure, giving Claude Code detailed descriptions of how she wanted each section to look and what it should say. She spent Sunday on the lead capture form, the PDF download flow, and testing everything on her phone. By Sunday evening it was live on her domain.
"The site Claude built was honestly better than what I had on Wix — cleaner, faster, and it actually represents how I think about my work. And I own every file. No monthly subscription, no platform holding my content hostage."
Within the first month of launching, her enquiry rate increased noticeably. She attributes part of that to the improved impression the site makes and part to the lead magnet, which she's now testing with different versions — another Claude Code weekend project.
James owns three retail outlets across Singapore. His point-of-sale system — like most POS systems — exports sales data as CSV files at the end of each month. For years, he'd been opening those CSVs in Excel, manually building charts, cross-referencing numbers between outlets, and writing a summary paragraph to share with his operations manager. It was tedious, it was time-consuming, and it was the kind of task that kept slipping to the bottom of his priority list until mid-month had passed and the data felt stale.
What he built with Claude Code: a script that does all of it automatically. He drops the CSV files from all three outlets into a designated folder on his laptop. He runs the script. Claude Code reads the CSVs, calculates outlet-by-outlet performance, generates comparison charts, writes a plain-English summary of the key findings and any anomalies that need attention, and packages everything into a formatted PDF that goes directly to his email and his operations manager's email.
The first evening was setup — getting Claude Code installed and having Claude build the first version of the script. The actual code was written in under 20 minutes. He spent the rest of that evening testing it with real CSV files and giving feedback on what the summary should highlight and how the charts should be labelled.
"I didn't know what a CSV parser was. I still don't. Claude handled it. I just described what I wanted to see in the report and it built the whole thing. First time I ran it on real data, it worked."
His monthly reporting now takes about five minutes — the time to drop files into the folder and send the PDF link to his team. He's started using the same approach for his supplier invoice tracking and is considering having Claude Code build a simple stock management dashboard next.
Rachel runs a corporate training practice focused on communication and leadership programmes. Her clients are mid-sized companies who bring her in to run workshops, often with 20 to 60 participants. Every new engagement requires a customised set of onboarding documents: a programme overview, a pre-work email for participants, a facilitator brief, a schedule with session timings, and a set of pre-programme survey questions tailored to the company's context.
Creating this pack used to take her half a day per new client. It wasn't intellectually demanding — it was mostly adaptation and formatting — but it was time she would rather spend on programme design and client relationship work.
What she built: a web-based form (running locally on her laptop) that asks for the company name, the number of participants, the focus areas they've agreed to cover, the programme dates, and any specific context she wants to incorporate. She submits the form, and the tool generates a complete onboarding pack — all five documents, properly formatted, using her templates and her voice — as a ZIP file she can download and send directly.
The build took three hours on a weekday evening. She started by describing the structure of each document to Claude Code, showing it examples of her existing templates, and asking it to build a form-to-document generation system. Claude Code built the form interface, the generation logic, and the ZIP file packaging. She spent the last hour testing it with three real upcoming clients and refining the output until it matched her quality standard.
"It felt like having a developer on speed dial. Except the developer knew my templates, my language, and my client context. I've used it for six new clients since and I haven't had to touch the code once."
She now onboards new training clients in under 30 minutes from signed contract to documents in their inbox. She's shared the tool concept with two colleagues in adjacent fields who are now building their own versions with Claude Code.
None of these five professionals wrote code. They described what they wanted in plain English. They answered Claude Code's clarifying questions. They reviewed what was built, gave feedback, and iterated. They tested. They shipped.
The common thread across all five is not technical skill — it's specificity. Every one of these builds started with a very clear problem statement: here is what I'm doing now, here is what it costs me, here is exactly what I want the tool to do instead. The clearer the description, the better Claude Code performs. Vague requests produce vague results. Specific problems with defined inputs and expected outputs produce working software.
This is the same principle that underlies good AI prompting in any context. If you want to go deeper on how to write instructions that produce consistently high-quality results, our guide to outcome-first prompting covers the approach in full — and it applies directly to Claude Code instructions as well as regular Claude conversations.
If you're ready to try building something yourself, the place to start is our step-by-step Claude Code setup guide for non-technical professionals. It covers the installation process in plain English, flags the moments where most non-technical users get stuck, and walks you through your first session from start to finish.
The builds in this post represent a few hours each. The professionals who made them are not exceptional — they're just willing to sit with the discomfort of learning something new long enough to get through to the other side. That willingness is the only prerequisite that actually matters.
Cowork SG is where Singapore professionals share what's actually working with Claude — hands-on, no jargon, always free.
Join Cowork SG Free →