“Building apps before 2022 was always challenging. Even with no code, you had to learn a tool or platform’s specific way of building stuff,” says Dan Parry, cofounder of voice survey platform Cluso and an educator focused on a blend of AI, no-code tools and entrepreneurship.
Dan has been at the forefront of the no-code scene since 2016, building products with tools like Bubble and Softr. Now, he’s moved into building primarily with AI. “Over the last few months, there’s been an increase in tools that allow you to ‘prompt’ an app into existence,” he says. “You might not know developer terms but can now type plain English into a text box to get it to create your vision.”
In our Startup Life newsletter, Dan gave us his top tips for building your first product with AI.
Understand the lay of the land
There are lots of tools out there. Some of the most popular ones right now are Replit, Lovable and Bolt.new. Each has its own distinct approach. Replit acts more like a thinking partner and is more developer-friendly — it analyses code and works through problems systematically. It shows you the code being developed that you can plug into a bigger project. It feels closer to development environment support tools like Cursor and Windsurf. By contrast, Lovable and Bolt take a more direct prompt-to-build approach, turning your writing straight into applications. Also, some tools only work in a specific coding language, whereas others like Bolt let you build in multiple languages. Figure out your level of technical skill and then choose the appropriate tools that work for you. Then, it’s trial and error.
Be aware of how AI tools charge you
Free tiers are great for learning, but make sure you know their limits. Some count tokens (think of these like words or characters), while others — like Claude or Vercel’s V0 — count the number of messages you send. Understanding this helped me to avoid any surprise bills. Monitor your usage. Keep track of how many messages or tokens you’re using — it’s easy to get carried away with back-and-forth conversations with AI, which can quickly eat into your budget. I regularly check my usage dashboard and set up alerts when tools have that feature.
Create a product requirements document first
Before you start any AI project, write down exactly what you want to achieve. This document should be your roadmap that keeps you from going off track. I’ve learned that AI can sometimes try to do too much — feature creep (adding too many features or overcomplicating a product) is a real thing when it’s so easy to build — or veer in unexpected directions, so having these ’guardrails’ is essential. You don’t have to do this yourself, you can ask Claude or ChatGPT for help before heading to one of the app building tools:
- Tell it your idea, your desired outcomes and your unknowns
- Get it to ask you questions to clarify what you need. Ask it to tell you what it thinks you want it to build and the proposed steps to get there. You want to see if it understands your scope, your technical or integration needs, any data or security restraints and your overall vision.
- Ask it to generate a product requirement document and a style guide.
Get good at writing prompts
Learning how to talk to AI is like learning a new language. You have to craft thoughtful, clear and specific prompts to get what you want — do not be polite and British with it. Think of it like receiving directions in a foreign city — the clearer and simpler someone is, the better the outcome for you. It will help you save money too! Most tools have great educational pieces on their blogs like Lovable’s guide or Replit’s top tips. Use AI to help you. Ask it to create succinct prompts for creating your product. I use voice-to-text in ChatGPT to reduce the level of thinking required to translate my thoughts. I’ll say, ‘Hey, this is what I want to do, what’s the perfect prompt for you to be able to do this effectively?’ I’ll get the results and then plug them into the app builder.
Start small and iterate
When I first began working with AI, I made the mistake of trying to do everything at once. Instead, you want to break projects into smaller pieces to build and test. This will help you understand what works and what doesn’t — it’s harder to problem solve when it’s all in one big project.
On the subject of… Building with AI
1. Here are free AI courses to get you started.
2. AI makes it faster to test things, but “taking some time still to ponder, research and validate an idea first, before jumping right into building, is still very valid.”
3. Your prompts need context and constraints.
Read the orginal article: https://sifted.eu/articles/how-to-build-with-ai/
