- Seed to Sequoia
- Posts
- 3 things you could build with AI tools in a day or less
3 things you could build with AI tools in a day or less
It's amazing what you can get done in GTM or product with new AI tools
We’re excited to be launching Skyp on stage at SaaStr Annual next week! Reply or message me if you will be there and want to meet up. I’ll be on the demo stage 10-11a Thursday exactly one week from right now (as I send this)! Want to try Skyp before the world hears about it? Join the waitlist. We are very close to hitting our limit of founding customers–if you want more leads, now’s the time.
Building in public is so hot right now, I thought I’d dip a toe in. Here’s what I’ve been working on this week, as inspiration for what you could build with AI tools. Or as a way for me to justify how I spent my time.
Will all of this work?
No. But some of it will. And that’s the game we’re all playing.
Conference app hacking
We wanted to schedule meetings at a conference we’ll be at soon (not saying which one 😎 ) but the app only allows 2 meeting requests. And who ever accepted a meeting request in a conference app, anyway?
So I opened up Cursor and built a scraper to go through the app programmatically, grab screenshots, OCR them using ChatGPT, and then save them to a CSV. That we could upload to Skyp and do outreach with.
This actually turned out to be hard. It is very easy with some websites or apps (e.g., TechCrunch’s disrupt website–I got something working well in under 10 minutes). But with this particular app and my particular setup, it took a while to get it to work. Still, it beat how long it would have taken to do the work by hand (we tried–it’s not really possible).
There are a lot of AI startups that claim to do scraping or other technical approaches to get you leads. I’d flip this and suggest you come up with the idea, open up Cursor or your AI tool of choice (windsurf?) and try it yourself. Instead of hundreds to possibly thousands of dollars, and long commitments, you can probably get something done in an hour or less with $20 invested in a monthly subscription. And if you can’t, then you can pay them their money for the easy button.
ROI calculations
Skyp has a pretty obvious ROI. Pay money, spend fifteen minutes setting up a campaign, and get leads. Some of which turn into revenue. The beauty of this kind of business is ROI calculators are unnecessary. And in case I forgot to mention, we’ve settled on “pay what you want” for founding customers so the ROI is pretty easy to calculate (Waitlist here).
Obvious ROI is not the case for all companies. Several clients have struggled with this–the ROI calculation is complicated, but nobody has the patience to walk through excel. I’m doing a guest post for Kyle Poyar’s Growth Unhinged newsletter breaking down how to do cool stuff with AI coding tools, and I thought, what better use case than an ROI calculator?
So I built an ROI calculator for client PicoMES. Check it out here. The prompt was pretty simple–here’s an excel, build an ROI calculator out of it that’s beautiful and a react web app. Ask these questions. Use colors from their website. That sort of thing. It takes skill–but maybe 1.5% as much skill as building a React web app by hand.
Because you’re in total control, unlike the scraping project above, I can guarantee that prompt will probably get you 95% of the way to “done”. You can iterate in 30 minutes with simple prompts and it will just work. Things might get tricky if you want to integrate it into Hubspot or your CRM of choice. But nobody said you have to! Don’t let perfect be the enemy of good. A lot of value in a beautiful ROI calculator is walking through it on a sales call with a prospect–no CRM integration required.
Product iterations
We’re heavily leaning on AI coding to build Skyp. The goal is to keep this team small, indefinitely. Julian building stuff. Me selling and marketing stuff, and chiming in on product from time to time.
We needed to iterate on prompts and the structure of agentic AI within the app. To do that on the production app would be complicated. So I used bolt.new to build an app purely for testing prompt, prompt structure, model usage, and the rest of it. Took about 30 minutes to build it. I’ve poured hours into iterating. What before would have taken maybe hours now takes thirty seconds. For example, “add an export to CSV button that exports all of the current prompts, settings, emails” was enough to get exactly that, so I could save what worked.
AI for the win
I think this AI-engineering for GTM is going to be a massive trend. It may already be. The same for product development. We probably saved an entire sprint–maybe two–with an app that took an afternoon to build.
If you have more GTM engineer questions or things to solve, ask me. Just reply. Happy to dig in with you on them in a proprietary fashion–or build in public. I do wonder if I was brought in to a larger company (say, $10mm + ARR) what kind of pains they have in this area, and what impact GTM engineering (as it’s starting to be called) could have.
Second, if you’ve done anything like this, I’d love to learn from you. What’s worked? What didn’t work? Just reply.
Don’t want to vibe code but still want to leverage AI for GTM? Have I got an app for you! Waitlist here. Want to hear our new tagline? “Skip hiring AIs with first names. Use AI to make your own name.”