I’m Mitchell from AppSumo. I sat down with David Kelly, the general manager of AppSumo Originals, to unpack how a tiny team of seven runs a business that does the equivalent of $6–7M per year in recurring revenue and serves roughly one million registered users across products like TidyCal, BreezeDoc, SendFox, and King Sumo.
🚀 What AppSumo Originals is and how we measure success
AppSumo Originals is our collection of first-party products built and launched directly by AppSumo. David manages the Originals portfolio and the results speak for themselves: three separate products have already crossed the $1M revenue mark, and a fourth is coming soon.
Why that matters: Most SaaS companies see ~4% of products reach the million-dollar mark. Our process pushes that to ~30–40% by prioritizing speed, repeatable decision frameworks, and ruthless iteration.
🧠 The 4 AI tools that changed everything
We rely on a short, practical AI stack that accelerates work without replacing the team. The four tools we mention most often are:
- Chatbase — powers contextual chatbots that reduce support tickets by handling common product questions using our knowledge bases.
- ChatGPT (03 model) — our go-to for reasoning, planning projects, drafting copy, and synthesizing data.
- Gemini 2.5 Pro — excellent at analyzing screenshots and Google Drive integrations; we use it to generate in-app copy and product documentation.
- Schumer Prompts / prompt libraries — pre-vetted prompts we adapt to collect feedback, analyze reviews, and centralize insights for product decisions.
💬 How AI reduced our 5–10 hour processes to ~1 hour
Here’s a practical example: redesigning settings copy or switching around in-app controls used to spark multi-hour debates between product, design, and engineering. Now we send a screenshot to Gemini, ask for clarity and UX improvements, and get structured recommendations in minutes.
That single interaction moves a task that previously took 5–10 hours to about an hour end-to-end. The ROI compounds: faster launches, quicker experiments, and less time spent on low-leverage debates.
🧾 ChatGPT + prompts: how we gather and act on feedback
We use ChatGPT extensively for internal planning and research: conversion rate analysis, monthly project plans, template creation for BreezeDoc, and subscription level modeling for TidyCal. David shared exact prompts he uses — many adapted from Schumer Prompts — to collect, synthesize, and prioritize feedback.
These prompts let us take disparate data (reviews, support tickets, feature requests) and centralize it into clear recommendations for the roadmap. AI doesn’t replace judgment — it accelerates decision-quality and surfaces what’s actually worth building.
🧰 The product modeling spreadsheet (how we pick winners)
We don’t rely on intuition alone. David walks us through a product modeling spreadsheet that scores ideas across an adapted RICE framework: Goal (how well it supports our yearly objectives), Impact, Ease (how simple it is for a small dev team), Viral (built-in growth/leverage), and Excitement (how motivated the team is to build it).
Each idea gets a numerical score. We prioritize quick builds where the development timeline is short, because the shorter the launch, the faster the ROI and the cheaper the experiment. That’s how TidyCal went from idea to first version in about five weeks.
📈 TidyCal case study: launch fast, recoup faster
TidyCal is our poster child for this approach. From concept to version one took roughly five weeks. Development and time costs landed around $30,000 — and we recouped that in roughly 30 days (maybe even less).
Key takeaways from the TidyCal playbook:
- Keep scope tight for version one.
- Use AppSumo and organic product-market fit to amplify traction.
- Measure cost-to-launch and set a breakeven target — speed shortens that window dramatically.
📚 How we learn, test, and adopt AI
David’s advice for getting good at AI is pragmatic:
- Use AI in personal projects first — planning trips, birthdays, or small tasks to build the habit.
- Surround yourself with curious people and an entrepreneurial community who share workflows and use cases.
- Follow reliable sources — LinkedIn posts from practitioners, prompt subreddits, TechCrunch, TechMeme — and try products from marketplaces (we list many on AppSumo).
- Fail fast and iterate — even if 50% of experiments fail, the frequency of experiments finds the 50% that work.
🧩 Practical next steps for founders building with a tiny team
If you’re bootstrapping SaaS or AI with a small crew, here’s a compact checklist we actually use:
- Create a simple product-modeling spreadsheet that scores ideas against company goals, impact, development ease, virality, and team excitement.
- Prioritize short development timelines — a 4–6 week MVP uncovers product-market fit quickly.
- Automate low-level support with a knowledge-driven chatbot (Chatbase worked for us) to free senior staff for higher-ROI tasks.
- Use ChatGPT (or similar) for planning, copy drafts, and quick synthesis. Use Gemini to analyze screenshots and in-app copy.
- Leverage pre-vetted prompts (Schumer Prompts, prompt communities) to standardize data collection and feedback processing.
❓ Frequently Asked Questions
How many people run the Originals portfolio?
Seven team members at the time of the conversation. That works out to roughly a million dollars of ARR per teammate — a sign of efficiency, not magic.
Which AI tools gave the biggest ROI?
Chatbase reduced support ticket volume immediately. Gemini improved in-app copy and UX decisions. ChatGPT accelerated strategy, planning, and copy work. Combined, they turned long processes into fast, repeatable ones.
Can I copy your product spreadsheet?
Yes — the methodology is the important part: score ideas by goal alignment, impact, ease, viral potential, and team excitement. Keep it simple and use it to prioritize fast experiments.
How fast should I aim to launch an MVP?
Aim for 4–6 weeks for a version one that proves the core value. Shorter timelines mean lower cost and faster breakeven.
Where should I start learning prompts and AI workflows?
Start with prompt libraries like Schumer Prompts, explore prompt subreddits, and practice in low-stakes personal scenarios (trip planning, event prep). Then apply successful patterns to product work.
🏁 Final thoughts
Running a $6–7M/year recurring business with a seven-person team isn’t a fluke — it’s a discipline. It’s about combining a tight product-selection framework, fast execution, and pragmatic AI adoption that amplifies the team’s leverage. If you’re building SaaS or bootstrapping AI, prioritize speed, centralize feedback, and use AI to accelerate the high-leverage parts of your workflow.
Want the links and exact prompts we mentioned? Check the AppSumo post and David Kelly’s profiles for the full resources and templates.
