Here are the AI agents I've tried for this series
Those tools can help anyone create digital products, no coding or extensive tech skills required. However, even though it looks like the era of product creators is coming, at this moment in time, it's not yet here. AI agents will not take our jobs just now, but we shouldn't feel too safe either.
Good to start with, bad for scale
All my engineering friends say the same thing: AI agents generate code that is not ready for production use or scaling. Engineering leaders are really worried that more and more code will be autogenerated, introducing a magnitude more bugs and maintenance issues. They are excited that now their developers could be more productive, but at the same time they are very scared to lose control of their codebase. This leads to a cautious approach in most organisations with introducing AI code generators. CTOs want to make sure the code that is being added is of high quality, security, scalability and maintainability.Where does that leave AI coding agents?
They are being used anyway, just not necessarily as advertised. Instead of creating production-level apps, AI agents are used for quick prototyping, testing and even MVP creation. And that's a huge value! Back in the day, if you had an idea for a digital product, you needed either to code it yourself or find someone who could code. That meant spending lots of time and money until you have something tangible you can validate. Now, you can create simingly working product in minutes with the help of AI agents. You can show an interactive prototype or even an MVP to your potential customers and see if it's working for them. If it's not, you can keep iterating. If it is, you have a much stronger case to turn it into a real business, get funding and co-founders.
What is AI coding good for now?
Quickly visualising an idea, making something clickable for a presentation, proving a concept and gauging interest - all those could be helped by agentic AI. Even going from zero to one might be possible with autogenerated code and design. But later, and for now, you need to accept it as a must; you need to throw all the auto-generated code away and start over with proper engineering.Good to learn, bad for being challenged
PMs can get a lot of benefits from using AI agents even if the outcomes, like a product it generates, are disregarded. AI agents mimic PMs' interactions with engineers. For an agent to generate something tangible, you need to explain what you need. Similarly to working with human engineers, you need to be clear, you need to think of most of the use cases and limitations, and you need to handle exceptions. All this could be practised by using AI agents, and it could make you a more confident PM.However, this comes with a crucial caveat. Unlike working with real engineers, AI agents won't ever challenge you. Those systems are trained to be the most accommodating, so they will try to understand you even if you write badly. Human engineers would be more demanding, and for a good reason. So don't rely too much on AI agents being able to understand you; they have infinite patience. Be sure you harness your communication skills and make friends with engineers.
AI coding agents = myth busters
Early in my career I was arrogant (was?). I thought I knew better than anyone, and if only I could make all the calls, all the decisions, I would create the best products out there. Reality has predictably crushed that fantasy. And now, AI coding agents put the last nail in the coffin of that dream.A diverse, multitalented team will still create better products than one brilliant individual ever could. That's just a given. You need different perspectives, creative tensions, even constructive conflict. You need group dynamics. To be fair, some AI agents are trying to mimic that too by offering ideas on how your product could be improved. Yet even with AI help, a single person would find it much harder to create successful, innovative products than a diverse team of specialists would.
AI agents are here to stay. The value they provide now might be reasonably low, but they are learning and improving with unimaginable speed. Depending on the nature of your product, you might want to jump on the bandwagon right away or wait it out a bit until the early adopters help to make AI agents better and more capable. Still, I would recommend anyone in the product-building community to at least try some of the AI agents and see how they work. It's a fascinating piece of tech that could end up revolutionising how we build products.
Good luck vibing.