AI tools for Product Managers
I experimented with some AI tools to build this website and app and here’s what I learnt
Please note: I am not particularly technical, although I understand basic web/app infrastructure and jargon. And I get bored easily with detailed tasks which is why I enjoy working with more talented and detailed engineer/design friends ;)
I used two tools and will review them separately with my learnings:
As the package says - for frontend website design
Pros:
- Incredible customer support including 1-1 chat to troubleshoot and a plethora of youtube videos
- Good flexibility in design if you know the basics of containers + HTM/CSS as you can paste your own code
- Fairly intuitive options for elements and good breadth of templates to use
Cons:
- Not that much actual AI despite the branding
- Not hugely intuitive: quite difficult to get your head around the individual functionality (e.g. layout design and formatting) > better to read instructions and watch support videos first instead of diving head first (the latter being my typical style)
- Limited ability to connect backend components even if you wanted to
For non-engineers to build simple apps.
Pros:
- The AI agent is pretty incredible - you can use natural language to ‘instruct’ the tool as you would a team of developers as a PM
- Can achieve in hours what you would achieve in weeks
- Great for quickly prototypin
Cons:
- Maxes out use after a period of time, which you’re blocked for 4hrs - so quite stop/start
- Awful error troubleshooting by the agent itself, requires you to follow the code and support heavily to troubleshoot and error handle
- I worry about the environmental impact of using this type of tech…
Overall learnings
1. Start with a tool like Perplexity to scan the market for the best tool; you can compare prices and reviews too
2. Choose a tool with online learning, FAQs and/or Reddit forums to help troubleshoot (Code DesignAI was great for this)
3. Most tools are either front-end focussed (e.g. Code Design) or backend focussed (e.g. Replit). It is surprisingly hard to combine the two (e.g. building a backend in Code Design or adding a nice frontend/Javascript to Replit)
4. Error handling and troubleshooting in Replit is extremely difficult. Understand the code changes and being able to trouble-shoot using the dev tools is the best way (so requires technical knowledge)
5. For any AI tools, think very carefully about how you instruct the agent. Making things logical and step-by-step as if you were building it helps avoid errors and complex rollbacks etc (it is better to think through the application e2e initially, how you would build it step by step, adding complexity and then instruct the agent in this way)
6. AI agents tend to get creative and cause their own problems - be cautious about accepted suggested next steps.
Key takeaway
PMs should use tools like Code DesignAI/Replit for simple prototyping and testing e.g. landing pages, simple features. Continue to use a good product process rather than over-relying on the AI tool. Beyond MVP or when design is important, continue to rely on the specialists in your product team 🙂