Ai and the return of true engineering

One of the stranger side effects of AI-assisted software development is that it may finally give engineering teams enough breathing room to engineer properly again. That sounds backwards at first, because most conversations around AI focus on acceleration. Faster delivery. Faster implementation. Smaller teams producing more software. More tickets completed in less time.

And yes, AI absolutely enables that; but I increasingly suspect the more interesting long-term impact is not speed. It is the possibility that engineering teams might finally have enough leverage to spend time on the parts of software delivery that have slowly been squeezed out over the last decade.

A black computer keyboard on brown wooden desk with the words Do More written on the computer monitor.

For years, many organisations have optimised themselves entirely around throughput. Short-term delivery pressures became constant behaviours, planning became compressed, and documentation was treated as overhead. Discovery sessions have shortened, and architecture discussions have been skipped because “we need to start building”.

Somewhere along the way, many teams accidentally convinced themselves that intentional engineering was the enemy of agility.

Agile Was Never Supposed to Mean “No Thinking”

One of the most frustrating things I have watched over the years is agile gradually becoming shorthand for skipping engineering discipline.

Documentation became “anti-agile”
Architecture became “waterfall”
Planning became “slowing delivery”

The irony is that genuinely good agile teams never behaved like that at all; strong agile teams collaborate heavily, continuously refine understanding, challenge assumptions, and iterate intentionally, breaking work into manageable chunks while maintaining a coherent understanding of the wider system.

The problem was never agile itself. The problem was organisations trying to optimise every process purely around delivery speed. AI changes those economics dramatically.

An iPad on a wooden desk with the words ‘Slow does not always mean slow’ written on the screen over and over. With a glass of iced tea alongside the iPad on the desk.

AI Makes Engineering Support Work Cheap Again

One of the most interesting things about modern AI tooling is not actually code generation. It is everything surrounding the code.

Imagine a discovery workshop where:

the meeting transcript is captured automatically
the whiteboard sketches are analysed
contradictions in requirements are highlighted
RAID logs are scaffolded
architecture summaries are generated
missing questions are identified overnight

Suddenly, many of the supporting engineering artefacts that teams previously neglected become almost free to produce (the less we talk about new token billing the better, until Monday).

That matters enormously.

For years, teams avoided documentation because writing and maintaining it required significant human effort. Now the economics are changing. AI can produce first drafts at extraordinary speed, which means engineers can spend more time refining ideas instead of starting from blank pages. And that is a genuinely exciting shift.

The Return of Collaborative Engineering

One of the things I quietly worry about with AI-assisted development is organisations using it purely to increase throughput. More tickets, more code, and more output. But perhaps the opportunity is in an entirely different kind of increase.

AI can and should create space for more:

discussion
experimentation
design reviews
human-centric approaches
architecture sessions
mentorship
pair engineering
collaborative problem solving
and finding fun in what we do

A photo of a group of people having fun together at work, with the person sat at the computer wearing a cardboard robot head.

Some of the best engineering work I have ever seen did not emerge from isolated coding sessions. It came from people standing around whiteboards, challenging assumptions, debating trade-offs, and gradually refining understanding together.

That kind of collaborative engineering became harder when every organisation was under relentless pressure to increase delivery speed. AI may finally give us enough leverage to reclaim some of that time.

Engineering Craftsmanship Was Never About Typing Speed

There is a dangerous misconception emerging that AI reduces the importance of engineering skill because implementation itself becomes easier, and open to more people with less experience. I increasingly think the opposite is happening. When implementation becomes cheap, the quality of thinking surrounding implementation becomes more important; people who have ideas can now access a way to rapidly prove them. Just so long as they have an experienced engineer help productionise it before risking their future client’s data!

The future value of engineers can now increasingly come from:

judgment
design
reasoning
intentionality
validation
understanding trade-offs
preserving maintainability
communicating clearly

In other words:

craftsmanship.

Perhaps the strangest irony of the continuing acceleration of AI is this: the more powerful and less dependent on human effort our tools become, the more important the judgement of a human engineer becomes.

A photo of some drinks cans painted green against a grey wall. The cans have letters spray painted onto them and spell our ‘Power to the people’.