Not long ago, building software had a clear rhythm. You'd figure out the problem, design a solution, write the code, test it, ship it, and monitor it. Then you'd do it all over again. There was satisfaction in that loop, and there was a kind of quiet pride in doing it well.
But now, things feel different. You're still solving problems and writing code, but the rhythm feels… different. Some parts are faster, some are messier.
That’s because we’re not doing it alone anymore.
AI is now part of the team.
And it's not just speeding up our typing. It's changing how we think about the whole process, from the first idea to the moment we go live.
AI changes the Shape of the Loop
Let’s recall the engineering workflow:
- Understanding the Problem: Talking to people, asking questions, and clearly defining the problem.
- Choosing the Right Solution: Considering current systems, trade-offs, and deciding the best solution that fits with the current context.
- Planning the Work: Writing clear, actionable implementation plans and system design documents.
- Writing Code: Writing code that’s clean, readable, and easy to maintain.
- Testing and Fixing Bugs: Making sure everything works as expected and quickly fixing issues.
- Deployment and Monitoring: Launching the changes safely and closely monitoring its performance.
Every step was human-driven.
AI doesn't just make individual steps faster, it blurs the boundaries between them.
- While you're writing code, your AI tool suggests structure, edge cases, and even documentation.
- When debugging, AI surfaces possible root causes before you’ve even hit the logs.
- Before you start coding, you can ask AI to scaffold the design doc or break down tasks in Jira.
- During QA, AI can generate test cases from API contracts or user stories, reducing manual work.
The linear path becomes more like a web. You're jumping back and forth more often, and the feedback loop tightens dramatically. Instead of waiting for code review to hear about a potential issue, AI may flag it right in the editor.
It’s less “step-by-step”, and more “everything, everywhere, all at once”.
How to adapt
We need a new mental model and here’s how engineers and teams can rethink their workflow:
Start with “How can AI help?”
Every task you pick up, ask: What part of this can AI do faster, or help you think through better?
- Writing a migration plan? Ask AI to generate a first draft.
- Building a feature? Let AI write the boring boilerplate.
- Debugging an issue? Summarize logs and ask AI for a starting point.
Move from “Write Code” to “Shape the System”
You should focus on the why and how well. Think:
- Is this code scalable?
- Are we choosing the right trade-offs?
- Will it be maintainable?
AI can generate code, but you’re still accountable for the quality.
Treat Docs and Context as Code
If you want AI to help meaningfully, you need to give it the right context:
- Centralize your decisions in the codebase.
- Keep README, design docs, and task descriptions up to date.
- Log the “why” behind decisions in places AI can read later.
Think of this as writing for your future self and your future AI collaborator.
Tighten the Loop
Old workflow: Design → Code → Review → Test → Deploy → Monitor.
New workflow: Think → Prompt → Code → Review → Prompt again → Ship faster.
You might ask: where are Design, Test, Deploy and Monitor in the new flow? They're still there, but more blended into the process:
- Design happens during Think and Prompt, with AI helping shape ideas early.
- Testing blends into Code and Prompt, as AI suggests and runs test cases continuously.
- Deploy and Monitor are rolled into "Ship", once a solution passes review, AI helps push it live and monitor it right away.
These stages haven’t disappeared, they’re just no longer siloed. They happen earlier, more often, and more automatically.
This means:
- Shorter feedback cycles.
- More experimentation.
- Smaller Merge Requests, reviewed faster with AI-assisted coding.
In product development, time-to-insight becomes more important than time-to-code.
Rethink Team Planning
Instead of estimating “how long will it take to code this”, ask:
- “How long to align on the solution?”
- “How do we co-build this with AI?”
- “Where do we still need deep human judgment?”
Time saved in coding should be reinvested in better testing, tighter design, and learning.
Welcome to the new engineering workflow. It’s still your craft, but now with a jetpack.