I am not a developer. I have an English literature degree and have worked in international marketing and corporate training as an instructor. The closest I ever came to coding was DOS, BASIC, and HTML in the early-to-mid 1990s. Then one day in mid-February 2026, I was writing a training proposal email as usual. I would look at the screenshot of a training request from a client’s manager, find a matching curriculum, add a suitable greeting, and write the email. I have done this for 14 years, and honestly, it follows the same pattern every time. “Couldn’t I just have AI do this repetitive work?” That thought was the beginning. The moment I asked Claude to “build a web app that automatically generates training proposal emails,” my 1.5-month journey began. In 45 days, I built 43 web apps, almost one a day. 40 of them have AI built in. And I did not type a single line of code myself. AI wrote all of the code, and all I said was “I need something like this,” “change this part,” and “this isn’t working.”
My rule: "If I do it three times, I automate it"
The driving force behind those 43 apps comes down to this: I don't put up with annoyances. The moment I think "ugh, I have to do this again," I stop what I'm doing and spend time automating or simplifying that task. Even if sending one email takes just 5 minutes, if I'll be repeating the same task every day, I think it's better to invest 3 hours now and build a web app. That's exactly what happened. Writing a training proposal email used to take 15 minutes, so I spent a whole day building an automation app for it. It felt crazy. But since then I send the same email in 2 minutes. At 60 emails a month, that's 13 hours saved in the first month alone.
This kept repeating. I built CallArchive because taking notes on calls was a hassle. I built LectureTrack because tallying my teaching results by hand in Excel was annoying. I built PayCheck because I was tired of switching between my bank account and a spreadsheet to confirm training payments. I built the New Training Schedule Registrar because I hated clicking through dates to add events to Google Calendar. It wasn’t about the time: a two-second task done three times a day is only six seconds. I just hated the feeling of “I have to do this again.”
Some nights I'd be working at 11 p.m., notice something annoying, and instead of saying "I'll do it tomorrow," I'd spend until 4 a.m. building an app to fix it. It may look inefficient. But from the next day on, that annoyance is gone for good. Isn't that a fair trade: one painful night for lasting relief? I've made that trade 43 times in 45 days, haha. The habit has had a side effect, too. As I automated my annoyances, my entire work process gradually came into view: where time leaks, which tasks truly matter, and which ones I only do "because I have to." Building the apps was itself a way of re-engineering my work.
First weapon: 16 personal tools that automated my work
The first things I built were tools to automate the work I repeated every day. Once I got a taste, I couldn't stop. 'OO RAG' was my most ambitious project. Over 14 years of corporate training, I have accumulated more than 120,000 PPT slides. I put all of them into Supabase, a vector database. Now when I type "Find slides on digital marketing," the AI quickly picks the most relevant slides out of 120,000. I didn't stop there. Based on the retrieved slides, it automatically generates a training proposal and even drafts the email. Writing one proposal used to take an hour; now it takes 10 minutes. I also added voice input, so when I'm driving I can say, "Make me a proposal for next week's Samsung Electronics AI training," and the draft is ready by the time I arrive.
The AI Training Proposal Email Generator is the app that started this whole journey. When a client sends a training request, I upload a screenshot of it. The AI extracts the text with OCR, analyzes the request, and automatically matches the best-fit program from our 230 training courses. It then drafts a customized proposal email that I can send straight from Gmail. The proposal emails I used to write, at least 5 or 6 a day, now take a few clicks.
The OOTrack lecture dashboard shows my teaching results at a glance. It pulls training records from Google Sheets and charts monthly revenue trends, revenue share by client, and breakdowns by course type. Claude AI is layered on top, so it even generates insights automatically, such as how first-half revenue this year compares with last year. When I need to report results, this one dashboard is all I need.
The New Training Schedule Registrar is small, but I use it every day. When a training schedule is confirmed over KakaoTalk or email, I copy the message as is or upload a screenshot. AI automatically extracts the date, time, location, and training topic and adds it to Google Calendar. I used to open the calendar, click the date, type the title, add the location… several times a day. Now it takes two seconds: “paste → register.” It is the classic app that makes people ask, “Why automate something that takes five minutes?” But in those five minutes, what I do is read the information, remember the date, find that date in the calendar, and type. That is not my job. It is something AI can do.
The CallArchive call dashboard is another one I use every day. When I talk with a client’s manager, the recording is saved automatically, and this app extracts the text with the Whisper API and Claude summarizes it. If I ask the AI chat, “What did Manager Kim at Samsung Electronics say last time?”, it searches the call records and tells me. I used to take notes during calls and couldn’t focus on the conversation. Now I don’t need notes at all, so I can give the call my full attention.
The AI Detail Page Maker is a bit of an oddball. It automatically builds training course detail pages for the Nanum Consulting website. Claude writes the copy based on a 13-step persuasion structure, Gemini generates the illustrations, and Canvas automatically layers on the text overlays. You can choose from 16 art styles and 30 Korean fonts. A detail page that used to take a week through a designer now comes out in 30 minutes. I didn't build this because something was "inconvenient." I built it to make something possible that was previously impossible. A one-person company can't hire a designer. It isn't perfect yet, and there is a lot of room to improve it as I use it.
The AI PPT Generator is linked to Nanum RAG. The AI picks slides that match your topic from a library of 120,000 and assembles them into a new PPT file. Tell it "Make a PPT for a 30-minute lecture on AI-powered marketing strategy," and it finds the relevant slides, arranges them in a logical flow, and generates new ones to fill any gaps. This app isn't perfect yet, and there is plenty of room to improve it as I use it. When I grumbled, "Why doesn't it work like Genspark AI Slides?", Claude replied, "Those were built by dozens of experts over several months."
The AI SNS→Blog Converter automatically turns the short posts I put on Facebook or LinkedIn into the longer format suited to Naver Blog or Tistory. Write 5 lines on social media and a blog post is created automatically, which sharply cut the effort of running several channels with the same content. This app, too, started from a classic pain point: "I want to post my best social media writing on my blog too, but rewriting it every time is a hassle." I took that hassle away.
The AI Curriculum Generator builds an hour-by-hour module plan from just a training topic and audience. Enter “6-hour AI training for middle managers in manufacturing,” and you get a finished curriculum with learning objectives, hands-on exercises, and time allocation for each module. When I design a new course, I no longer have to start from a blank page, which is a huge relief. Later I plan to connect it to RAG so it can draw on the courses I have already built.
The Curriculum Editor converts curriculum text into HTML that can go straight onto the website. The AI also generates the SEO copy and FAQ automatically, and uses RAG to look up and insert details on past on-site clients.
The Training CRM Sales Opportunity Manager is a CRM that manages sales opportunities as a pipeline. It tracks each deal stage by stage from lead to close, and handles sending proposal emails and issuing quotes automatically, all in one place. It is connected to Gmail, so I can send emails right from it. Before, I would hunt through rows in an Excel sheet, wondering, “What stage is this deal at now?” I built it to put an end to that.
PayCheck, my payment manager, tracks payments for training fees. Paid, unpaid, and long-overdue invoices are visible at a glance, so I no longer dig through Excel at month-end reconciliation.
The Nanum Consulting Marketing Dashboard and the Nanum Campus Marketing Dashboard visualize real-time GA4 data for the Nanum Consulting website and the Nanum Campus online training platform, respectively. They show which course pages get the most views and which keywords bring in search traffic, and Claude AI even generates weekly insights automatically. In the long run, I plan to connect them to Naver and Google ad accounts.
The Open Course Participant Manager is an app for managing participants in our regularly run open courses. It handles everything in one place: accepting registrations, confirming payments, and sending AI-written notice emails. Its companion, the Training Fee Receipt Issuer, automatically issues payment receipts to participants as PDFs.
Thanks to these 16 tools, my daily work hours have dropped a lot. (In exchange, I now spend more time building things with AI. 😅) And with the time left over, I built 27 apps. Solving one pain point frees up time, and that time goes to solving the next one. I keep wondering, “What should I build next?”
Second weapon: 15 AI demo apps for the training room
I teach companies how to use AI in their work. But I always had a worry. When I show AI use cases in a slide deck, everyone nods. But when I say, “Now try it yourself,” most people freeze. They open a ChatGPT window and have no idea how to write a prompt. That was a pain point too: the pain point in my own training. When learners get stuck in the exercises, the training’s impact is cut in half.
So I thought: what if I gave learners AI web apps built for specific purposes? Tools where you just fill in the blanks and get a result, with no need to agonize over prompts. That is how I built 15 AI practice tools organized by job function.
Meeting Minutes Wizard is the app that gets the best response in training. Copy and paste your meeting notes as they are, and AI automatically extracts the key summary, decisions, action items, and owners. The most rewarding moment is when learners try it with their own team’s real meeting notes and say, “Wow, this actually gets organized!”
One-Line Report → Full Version is an app for busy professionals who need to brief their boss but are short on time. Type a single key line, such as “Reasons for the 15% year-over-year increase in Q3 sales and Q4 strategy,” and it automatically writes a one-page A4 report with an introduction, body, and conclusion.
The Excel Insight Extractor lets you paste in CSV data, and AI pulls out trends, outliers, and wording ready for executive reports. Even people who don’t do data analysis can get insights just by copying their Excel data.
Enter two company names into the Competitor Comparison Analyzer and you get a strengths/weaknesses comparison table and strategic implications. Enter a business idea and target market into the Market Entry Strategy Generator and it automatically produces a SWOT analysis and three entry scenarios. Enter just the client and their needs into the Proposal Cover Page Generator and an Executive Summary is written automatically.
The Multi-Platform Social Copy Generator is a fun one. Enter one marketing message and it converts it into versions optimized for Instagram, LinkedIn, and Facebook. Same content, but Instagram gets emojis and hashtags, while LinkedIn gets a professional tone.
The Customer Persona Generator takes your interview notes and produces a persona card with name, age, occupation, pain points, buying motivations, and more. The Press Release Generator takes an event or achievement and automatically writes an inverted-pyramid press release with 3 headline options. The JD Generator turns a job title and role into a complete job posting.
Receipt→Expense Report takes a receipt photo or text, automatically extracts the line items, and turns them into an expense report form. The Financial Statement Translator turns statements that are nothing but numbers into explanations that non-specialists can understand.
The Quote Generator produces a clean quote layout with VAT calculated once you enter items, quantities, and unit prices. The Price Negotiation Simulator takes your current quote and target price and suggests three negotiation scenarios.
The Contract Risk Clause Detector takes contract text, highlights clauses that are unfavorable to you, and suggests revisions.
Using these 15 apps in training has clearly made a difference. Learners put in their own real work data and see results right away. Then they get strongly motivated, thinking, “I have a pain point like this too. Maybe I could build something for it?”
Third weapon: 11 tools built for B2B sales training
These are the tools I use when teaching B2B marketing and sales, my area of expertise. B2B sales has its own frameworks, so I built tools in which AI fills them in automatically.
The AI Target Customer Buying Center Analyzer automatically analyzes the Buying Center, a core concept in B2B sales. Enter a company name and a product, and the AI identifies who fills each role (decision maker, influencer, user, buyer, gatekeeper) and what value to pitch to each. In training, when participants enter their own clients, they get it right away: "So this is the message I need to deliver to this person."
The AI IPP Persona Generator automatically builds an Ideal Prospect Persona (IPP): a profile of your ideal prospect. It describes in detail who needs your product most, including their job role, pain points, and buying motivations.
The AI Empathy Map Generator fills in all six areas of what your target customer sees, hears, thinks, feels, says, and does. The AI Customer Journey Map Generator draws the full path from awareness to interest, search, purchase, and sharing, based on the AISAS model. A team assignment in a marketing class that used to take an hour now gets a first draft in 3 minutes, leaving the remaining time for discussion and revision.
The AI VP PPSDN Generator builds the value proposition, the most important element in B2B. It automatically derives five components (Problem, Pain, Solution, Differentiation, and Need-payoff) and completes an integrated VP statement.
The AI Inbound Content Marketing Idea Generator takes your company and product information and produces content strategies and ideas for 6 channels at once: blog, webinar, SEO, AEO, eBook, and YouTube. The AI Outbound Sales Strategy Generator automatically creates 7 outbound approaches, including cold calls, email, and social selling.
The AI Elevator Pitch Generator automatically writes a sales pitch in 7 steps, built to win someone's attention within 30 seconds. In training, participants create a pitch for their own company's product and use it to practice delivering it.
The three B2B VWE calculators are a different kind of tool. They compare the TCO (total cost of ownership) of a competitor's product against ours, and chart how much our solution saves over a few years. I built three versions: one for Nanum Consulting, one for Hyosung Information Systems' AI Ready Storage, and one for Kyungdong Navien's integrated solution. They are meant for training, but they also work as practical tools in real client sales meetings.
I never reveal these tools at the start of a session. In an 8-hour training, I spend 7 hours building the foundation with theory and case studies, then reveal them in the final hour and stress that AI now has to be applied to marketing and sales as well.
The biggest insight: “I didn’t learn to code. The language of problem-solving changed.”
A month and a half ago, I would think, "Which Excel function should I use to do this faster?" Now I think, "What web app could I build to do this automatically?" I didn't learn to code. I learned a mindset: "This can be automated." I can't stand repetitive work. Once I do the same task three times, the thought "I should turn this into an app..." pops up on its own.
It's almost embarrassing to call this a tech stack. The frontend is HTML, CSS, and JavaScript, all in a single file. The backend is Google Apps Script, the database is Google Sheets, the AI is the Claude API, and deployment is GitHub Pages. At first I didn't even know what GitHub was. I understood that git push meant "pushing" something, but had no idea where it was being pushed. Now I manage 43 repositories.
I also realized one more thing. In the AI era, the line between developers and non-developers is much thinner than you might think. What matters is “knowing what needs to be built.”
Domain knowledge built over 14 years in the training field — which repetitive tasks are painful, where participants get stuck, what clients really need — that instinct was the real engine behind all 43 apps. AI was simply the translator that turned it into code.
In the end, all 43 apps have one thing in common: every one started with "This is annoying." Annoyances in my work, in my training, in my clients' work. I didn't put up with them. I fixed them on the spot, however many hours it took. That "impatience" produced 43 working apps in 45 days. What will the next app be? I don't know yet. But I do know that tomorrow I'll think again, "Couldn't I just build an app for this?" And within 30 minutes of that thought, a working app will exist. Though of course, I'll spend hours agonizing and fixing it over and over.
To ne continued