If I were a product manager today, I’d start consistently building my own AI ecosystem >>today<<.
Why?
It takes a fair amount of time to:
a. try new things,
b. overcome learning curves,
c. understand limitations,
d. develop your own tools,
e. develop an intuition for solving issues with AI.
As much as AI makes you faster, speed isn’t everything. Quality still matters and it’s not a “given” unless you master prompts and automations properly.
By “ecosystem”, I mean:
Collecting prompts
Prompts are crucial and it’s easy to underestimate how much of a difference a well-crafted prompt makes. Keep a list/database of prompts categorised by use-cases and type. The critical part is testing and adapting them to your case (or data format you typically work with). Where to find them? The internet and LLMs.
Meta prompts
Once you develop a suite of meta prompts, creating other prompts becomes so much easier and… fun. Meta prompts are prompts creating other prompts, helping you create highly sophisticated prompts, including synthetic examples, error and edge-case scenarios. Best ones simply guide you through the process.
Prompts at your fingertips
Familiarise yourself with productivity apps to have these prompts easily searchable and “paste-able” via a shortcut. This feature is often called “snippets” or in case of AI interfaces, “presets”.
Various models at your fingertips
All models have shortcomings and it’s necessary to develop a sense of working with them. You might prefer one over the other. You got to be using more than one to develop your own sense and taste. Don’t buy into one provider (such as OpenAI or Claude), instead subscribe to a service like Raycast Pro to switch between models quickly.
Create an API account at a leading LLM provider
You’ll need an API key for some tools. Sign up for an API account, top it up and » set alerts and limits «. With automations, it’s easy to create never-ending loops draining your balance.
Automations
This is a time-sink initially but once you set it up, it feels like magic. Your ecosystem should include at least one automation platform or AI Agent such as make or Lindy AI. You should build 2-3 automations to get an understanding how they work and their limitations. Connect it to your services and/or to Slack. Understand the key difference between using “watching modules” and webhooks. Connect at least one AI provider. The goal here is to be aware of what’s possible so when you come across a mundane task that, you’ll immediately have an idea how to automate it.
↯
It takes fair amount of time to create such ecosystem. As a product manager, your time is scarce. Consistency is key. Spend 1-2 hours per week working on that ecosystem and it will save you a lot of time in a near future.
↝ Join the discussion at LinkedIn.
