Built by a Non-Coder with Cursor and Claude Code — How Profit AI Turned a Client Spreadsheet into a $30K-MRR Shopify App in 5 Months

Profit AI is a profit-analytics (FP&A) suite for Shopify stores. Instead of top-line revenue, it surfaces *net and contribution profit* in real time — after COGS, fees and ad spend — and bundles goal-setting, ad-platform connections and team task management into one place. Its audience is e-commerce operators and the marketing/consulting agencies that serve them.

It was built by Jack. For years he consulted for e-commerce brands and agencies, and each time he built the same kind of meticulous spreadsheet to compute profit, ad ROAS and scenario planning for a client. One day he realized that the *thing he kept rebuilding* was itself the product. That’s the crux: the idea didn’t arrive from outside — it was hiding inside the manual work he was already repeating.

Jack is not an engineer. Even so, using AI coding tools (Cursor and Claude Code) he translated the logic of his client spreadsheet straight into a Shopify app and shipped it. It launched in December 2024 (approved on Christmas). Hiring no engineers and never learning to code, he finished it by prompting the AI over and over and correcting its mistakes — which is exactly the most reproducible part of the story today.

Profit AI also layers a thin ‘done-for-you’ service on top of the SaaS. For a large Miami/New York agency, a Slack bot powered by Claude takes data requests and delivers custom data into spreadsheets. On top of the monthly subscription, human help lifts both ARPU and trust.

The result: about $147,000 total in roughly 5 months, with $30,000 MRR from the app alone (~$40,000 including services). And through all of it he ran no ads and staged no big launch — the growth came entirely from word of mouth.

The repeatable playbook

1. 1Don’t hunt for a new idea — inventory the manual work you’re paid to repeat
2. 2Pick the one you solve in a spreadsheet (the sheet’s logic is your spec for the AI)
3. 3As a non-coder, translate that logic into an app with Cursor + Claude Code (imperfect is fine — keep fixing the AI’s mistakes)
4. 4Ship first to people who already know the pain (your client network) — word of mouth for first-mover speed
5. 5Don’t set the price, discover it: launch high ($5,000), cut fast if nobody bites ($800)
6. 6Fill the product’s last mile with a human service (e.g. a Slack bot) to lift price and trust — automate it later

Deep dive

【Deep dive】This case is more than ‘a non-coder built it with AI.’ What actually worked comes down to four things: (1) how he chose the seed, (2) a realistic way to *finish*, (3) price *discovery*, and (4) layering service onto SaaS. Let’s break them down.

■ The seed was the manual work he already did every time. Jack didn’t go hunting for a new idea. He productized the very thing he had rebuilt for client after client as a consultant — a profit-analysis spreadsheet covering COGS, fees, ad ROAS and scenario planning. That’s the decisive difference from most indie builders: before writing a line he already had (1) proof the pain was real (he was *paid* to solve it), (2) a settled shape of the solution (the spreadsheet logic *was* the spec), and (3) knowledge of who the first customers were (past and present clients). ‘The chore you keep repeating is the draft of your product’ is a directly reproducible way to find ideas.

■ Finishing as a non-coder: the stamina to prompt and fix. Jack can’t code. Using Cursor and Claude Code (some summaries also mention Codex) he translated the spreadsheet logic into a Shopify app and shipped it. The lesson is *not* ‘AI does it all.’ The opposite: the keys are having a *well-defined spec* (the existing spreadsheet) and the stamina to keep pointing out and correcting the AI’s mistakes. With a vague spec, AI coding wanders. Jack shipped fast because what to build already existed fully in his head and in the sheet. He reached launch in December 2024 without hiring an engineer.

■ Discovering the price: $5,000 → $800. This is the most operational lesson here. Jack launched the paid plan at an aggressive $5,000/month — and got only a couple of takers. He quickly cut it to $800/month, and deals started moving. Two takeaways. (1) Price is not something you *set* but something you *discover*: start high, watch the response, drop fast if nobody bites. (2) Even so, $800/month is an order of magnitude above a $10 consumer app. In B2B (operators/agencies), even 38–39 customers clear $30K MRR on price alone. If an indie’s goal is ‘cover living costs with few customers,’ deep-pain B2B at high price × few accounts can be more realistic than thin-margin consumer volume — and this is a live example.

■ Layering a thin ‘done-for-you’ service on the SaaS. Profit AI doesn’t stop at software. For a large Miami/New York agency, a Claude-powered Slack bot takes data requests and delivers custom data into spreadsheets — a *service*. That lifts MRR from $30K (app alone) to ~$40K including services. The implication is clear: in early B2B SaaS, filling ‘the last step the product can’t reach’ by hand raises price, retention and trust at once. You can productize (automate) that human step later — *not* trying to solve everything with product on day one is what makes launch fast.

■ $147K on zero marketing, pure word of mouth. Jack has run no social virality, no ads, no big launch. Growth is 100% word of mouth. It worked because, at the seed stage, he already held the audience that *knew the pain* — past/present clients and the surrounding agency network. In other words the acquisition channel was baked into his pre-existing human network. Meanwhile, 38–39 paying out of 122 total users means conversion still has room. Flip it around: he reached $147K with the entire marketing upside still untapped.

■ How to reproduce it (the non-coder pattern). The template: (1) inventory the manual work you are *paid to do* or keep repeating; (2) pick the one you solve *in a spreadsheet* (the spec is already settled — ideal for AI coding); (3) with Cursor/Claude Code, translate that sheet’s logic into an app (it needn’t be perfect — keep fixing); (4) ship first to people who already know the pain (word of mouth is your first-mover speed); (5) price high, and cut fast if nobody bites; (6) fill the product’s last mile by hand, automate later. The numbers (38–39 accounts for $30K MRR) show that ‘few high-price B2B customers’ can be a realistic living-cost path for a solo non-engineer.