Why I'm Adding AI to My Toolkit (Honestly)

I'm not an AI researcher. I don't have 20+ agent deployments under my belt. I'm a software engineer who's been building web apps, SaaS platforms, and ecommerce stores for nine years — and I'm now figuring out where AI genuinely helps in that work.

I'm writing this because most AI content online falls into two camps: breathless hype from people selling AI services, or deep technical content from ML engineers. There's not much written for builders like me — people who ship products and want to know where AI fits in honestly.

What I've actually used AI for

So far, the AI features I've built or integrated into client projects are modest but real:

Search and retrieval. Adding semantic search to an existing product so users can find things by meaning, not just keywords. This works well and is relatively straightforward to implement with embeddings and a vector store.

Content generation helpers. Giving users a starting point for text fields — product descriptions, report summaries, form pre-fills. Not replacing human judgment, but reducing blank-page friction.

Structured data extraction. Pulling structured information out of unstructured text — invoices, inspection reports, customer messages. This is where LLMs shine for me: messy real-world input, clean structured output.

What I'm skeptical about

Autonomous agents in production. The demos are impressive, but giving an AI agent real authority over production systems still feels premature for most of the teams I work with. The failure modes are hard to predict and harder to debug.

AI as a core value proposition. For most products, AI is a feature, not the product. If the underlying software doesn't work well, adding AI won't save it.

Replacing developers. AI makes me faster at certain tasks — boilerplate, research, debugging. But the hard parts of software engineering — understanding requirements, making architectural tradeoffs, communicating with stakeholders — are still entirely human problems.

How I'm approaching it

I'm treating AI the way I'd treat any new technology: learn it by building with it, be honest about what I don't know, and only add it to client projects when it solves a real problem. I'm not going to reposition my entire practice around AI just because it's the current hype cycle.

The builders who will do well with AI are the ones who already know how to ship software. AI is a powerful new tool, but it doesn't replace the fundamentals.