🧠 AI-First Engineering in Practice

🧠 𝗔𝗜-𝗙𝗶𝗿𝘀𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗶𝗻 𝗣𝗿𝗮𝗰𝘁𝗶𝗰𝗲 I’d just finished building a backend feature — pairing with my AI agent throughout the process. It handled the actual coding while I designed and described what I wanted, reviewing, and refining the output.

In parallel my other agent picked up the frontend work from the same ticket description — including translations, the last step before release. Painfully this hadn't been automated and a bunch of json files.

The functional work was easy to review, but it had also updated dozens of translation files for a new field: “Total RRP.” Everything looked tidy — until I spotted the Spanish one: RRP Total. It looked English, if a PR landed in your review queue with English-looking text in a Spanish file, wouldn’t you question it? It could be a simple copy and paste catch.

So I did.

The agent replied:

“𝗧𝗼𝘁𝗮𝗹 is also a Spanish word.” “In Spanish, adjectives follow nouns — so 𝗥𝗥𝗣 𝗧𝗼𝘁𝗮𝗹 is grammatically correct.” “𝗥𝗥𝗣 stays unchanged for consistency across locales.”

Fair point — next I checked the German translation: Gesamt-UVP. Why UVP instead of RRP?

The agent explained that 𝗨𝗩𝗣 (𝗨𝗻𝘃𝗲𝗿𝗯𝗶𝗻𝗱𝗹𝗶𝗰𝗵𝗲 𝗣𝗿𝗲𝗶𝘀𝗲𝗺𝗽𝗳𝗲𝗵𝗹𝘂𝗻𝗴) is the standard German equivalent of RRP and already consistent elsewhere in the codebase. That’s context-aware reasoning — the kind you’d expect from a engineer using their brain.

Questioning an agent is exactly like questioning an engineer. Engineers and AI make mistakes that's the whole point of reviews to catch something the other missed. Sometimes AI catches yours, sometimes you see theirs, this is just a simple example.

There’s a misconception that “AI-first” means letting machines run wild. It doesn’t. It means designing workflows where humans and AI collaborate efficiently. In this case, the AI handled the repetitive updates, grammar, and consistency — and I focused on review, validation, and judgement.

That’s collaboration, I pair programmed with one agent on the backend and peer reviewed the other on the frontend. I decided the roles based on tasks.

AI-first engineering isn’t magic. It’s good engineering practice :

✅ You think and guide while it produces the code you want.

✅ Automate the mechanical work.

✅ Question and review as you would any engineers output.

✅ Keep strong guardrails.

The speed AI brings doesn’t remove responsibility — it raises it. Senior engineers now have more accountability, not less. When tools can update fifty files in seconds, your value isn’t how fast you code — it’s how well you question.

So yes, “RRP Total” turned out to be correct. But the real win was the process: reviewing, questioning, reasoning — human and AI, working side by side.

𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗔𝗜 𝗶𝘀 𝘀𝘁𝗶𝗹𝗹 𝗷𝘂𝘀𝘁 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 — only faster, smarter, and a little more collaborative. 😉

#AIEngineering #AIFirst #SoftwareEngineering #TechLeadership #DeveloperExperience #AIInPractice