Builders Foundry #5: Agent or agentwashing? Spot the difference

For people who build software. We focus on AI, design, engineering, cloud, and startups without the hype.
No vendor pitches or buzzwords. Just real architectures, production lessons, and honest trade-offs. For engineers, architects, platform teams, and technical founders.

🤖 1. Should This Even Be an Agent?
Jacques van der merwe
AI agents are rapidly becoming the answer to everything.

But what was the question?

Before adding an LLM, RAG pipeline, tool calling, memory and orchestration to an application, there is a more important architectural decision to make: does this problem actually need an agent?

In this session, we'll work through a practical framework for answering that question.

Using one business problem throughout, we'll progressively move from deterministic application logic to an LLM, retrieval, tool calling, an autonomous agent and finally a multi-agent architecture. At each step, we'll examine what new capability we've gained, what complexity we've introduced, and whether moving further up the autonomy ladder is justified.

We'll look at questions such as:

When is traditional code still the better solution?
When is a single LLM call enough?
When does RAG solve the problem without an agent?
When should an AI system be allowed to take action?
Where should humans remain in the loop?
When does multi-agent architecture represent a real boundary rather than just multiple personas?

This is an architecture-focused presentation rather than a live-coding session. Microsoft technologies such as Microsoft Foundry, Microsoft Agent Framework, Azure AI Search and Entra will be used to ground the concepts.

You'll leave with a simple decision framework you can apply when evaluating your next AI use case:

Use the least autonomous architecture that solves the problem.

🤖 2. Q&A session
An open chat with the host for everyone joining online.
We will try to record the session, but joining live is better since that is where the Q&A happens.