Read Part 2:
1 Mastering prompt engineering: clear thinking and being goal-oriented.
Which are also the fundamental traits for any knowledge worker.
Clear thinking allows you to precisely define problems.
Being goal-oriented ensures that your prompts are tailored to achieve specific outcomes.
Applying these qualities in implementing GenAI features:
Clear thinking means process-oriented thinking. Whether you’re replacing parts of a workflow with GenAI or rebuilding it with GenAI first, you need to break down the process into manageable granularity.
Being goal-oriented means focusing the final outcomes. Think about sales, modern software tools drive efficiency in the selling process. GenAI products would aim to do the sells.
2 Domain expertise matters more
Product software engineers must have an understanding of the domain they are building for.
For example, if you are working in advertising technology, a deep understanding of the advertising business can be more valuable than pure engineering expertise. This contextual knowledge is what enables you to create truly impactful solutions.
3 It is engineering and machine learning
Unit tests are still crucial. Building test evaluations (e.g. LangSmith) ensure that AI performance meets expectations.
Make agentic tools more adaptable. Consider adding some tolerance to user inputs and improving error messaging to guide LLM more effectively when things go wrong.
Not every problem requires an LLM-based solution.
4 Chat interfaces are not the only form of UX
Forms, buttons, and other traditional user interfaces can be highly effective, especially on screens. It's crucial to remember that different UX elements serve different purposes, and the best interface is the one that suits the user's needs and context.
5 Agents = Microservices
Agents can be thought of as microservices. They pass messages to each other like actors, enabling modular and distributed solutions.