Where is the value in software development now?

As a product developer, knowing that people are using your product gives you confidence. It’s a vote of confidence in your judgement. You’ve found or picked a problem that needed solving. It also says that what you do and how you work is valuable - especially at the start. However, more recently, AI has made me question where my value actually lies, and what it actually is.

Back in the early days of ScriptRunner as a commercial operation, I would have 10 ideas a day, but only time to implement one or two. Of course, when I was still working a day job, I’d have even less time. As ScriptRunner grew, I’d have to persuade others on the team that an idea was worth implementing, worth spending time and resources on. Today, if you’re using AI to write your code, if you have 10 ideas, you can pretty much implement them all by lunchtime.

Writing code is no longer the hard or time-consuming part. The question has flipped from "can we build it?" to "how do we choose the right thing?"

My initial approach to using AI was treating it as an assistant. The result was that I found myself as the human reviewer of AI code. Not only was it the totally wrong way to work, but it was demoralising. I was correcting AI’s homework, finessing its output to get it past sniffy human reviewers.

The reality was that I actually found it physically uncomfortable. AI codes so quickly that I became increasingly unfamiliar with code shipped under my name. Where, in the past, I would know it inside out.

The joy drained out of the process for me. I wasn’t getting a buzz from elegantly refactoring code. And I wasn’t getting dopamine hits from seeing the green ticks appear as a piece of code I’d crafted passed its tests.

The AI was driving me, not the other way around.

The more troubling thought, however, was that 30 years of my knowledge was now worthless. How to use Maven, Gradle, OSGI and hundreds of other things - now unnecessary. So how can you feel valuable when your value has been taken away?

One day, I was observing and discussing AI coding with a colleague. They were working with it (and agentic development) in a very different way, firing off six concurrent agents and just checking in to guide them. It occurred to me that I could try something different: stop looking at the code entirely, to stop checking up on AI line-by-line.

Now, the primary surface I’m interacting with is the app's behaviour. Not the code.

It’s a fundamentally new model for me. I have agents that ‘work’ as a team of trusted senior engineers. I give them intent and an outcome, but not a specification. They ask clarifying questions, build, deploy, test, and hand back a working thing to critique or steer.

It’s about letting go of hypothetical or assumed value and seeking the actual value. The value of coding is often expressed through ideas like ‘elegant code’. I spent years hand-crafting ScriptRunner and, of course, it wouldn’t have been built if I hadn’t done it. But the value of a product like ScriptRunner isn’t in the elegance of its code. It’s in how it solves problems.

I’ve realised that the code was never what drove me. Shipping a feature and having a user say “this solved my problem" in the forums (and, later, in the product reviews on the Atlassian Marketplace) was where the excitement was.

My value lies in understanding what customers need and being motivated to work on engineering solutions that matter enough to customers to pay for them. You might say that I translate lived frustration into features.

With AI, I’ve lost a recognised skill that differentiated me from people who couldn’t do it. However, I’ve gained this other opportunity. Now I can build more stuff, and the challenge is knowing what to build, whether it’s worth building, and what the customer thinks about it. I simultaneously retain the advantage of being able to understand line-by-line what's built.

"Where is my value now?" is a question that everyone needs to ask themselves. The answer is different for each person and every role or sector. It's genuinely hard, too. Nobody should be left to work it out alone.

There are many examples of things we did in companies that were just convenient or limitations. Before the invention of word processing, businesses had typing pools. Rooms full of typists. Organisations didn’t need typists; they needed accurate documents in a world where they could most readily be created with a typewriter.

Businesses succeed by meeting customer needs at a price that the customer will happily pay. That’s why we need to seek out the value – and not just because of AI, but all the time. I have to be honest: I have more confidence in AI-written code than in any human-written code, even my own. That’s especially true when I’ve got AI working through my code for errors or vulnerabilities. But it’s hard to let go of the thing that I thought was my value. I was clinging on to the old way. When I just cut myself loose from all that, I realised I’ve never felt more excited about what I do.