What Engineering Leaders Need to Prepare For

AI is changing software engineering by shifting engineers away from repetitive implementation and towards architecture, validation and business decision-making. Organisations that treat AI as a coding tool alone risk missing the bigger transformation.

That was the theme of Montash's panel, DevOps and Software Engineering in the Age of AI, hosted by Harry Glover and Abutalib Shah, with two guests operating at opposite ends of the market:

• Rocky Woestenborghs, Head of IT for Domestic Products at ING Belgium, responsible for the banking products used by retail and business customers, across a distributed team of 400+ engineers in a heavily regulated environment.
• Piotr Krzepczak, CTO at Bnewable, a Belgian energy company building and operating large-scale battery storage on its in-house Voltana platform, working at scale-up speed with less red tape.

The discussion covered how the engineer's role is changing, the junior talent pipeline, platform engineering, data foundations, the cost of running AI, and the guardrails needed to adopt it securely.

Watch the full webinar

What is "DevOps and software engineering in the age of AI"?

It is a change in where engineers spend their time and what they are accountable for, not a reduction in the need for engineers.

Instead of writing every line by hand, engineers increasingly use AI to generate, review and refine code. Their value shifts to defining the problem, setting the right context, validating the output, and confirming the solution works end to end. 

Rocky described the old delivery model, where a junior started at the bottom writing and testing code and worked upward with experience. AI now absorbs much of that lower layer, moving human focus to the top: requirements, system design and business judgement.

For engineering leaders, this changes what "good" looks like. Strong coding skills still matter, but so do business understanding, system thinking, and the judgement to tell whether AI has produced the right answer rather than just an answer.

How is AI reshaping the engineer's role?

Piotr framed the change as two shifts: up and left.

Up: engineers operate at a higher level of context. The question is no longer only how to build something, but what to build and why. As he put it, this is the kind of thinking organisations used to expect only from seniors.

Left: problem definition moves earlier. Engineers now shape specifications and system design before any code is written, and they set the guardrails that tell AI agents what to build, how, and what not to do.

Rocky made the same point from a banking perspective. Hallucination is still real, so nothing reaches production without validation. Prompting and giving the right context have become skills in their own right, which pulls engineering closer to the business it serves.

What happens to junior developers in the AI era?

Both guests named this as an unsolved problem.

For decades, junior developers learned through repetition: fixing bugs, shipping small features, and gradually taking on harder work. If AI absorbs those entry-level tasks, the traditional route into senior, business-literate engineering starts to close. 

Rocky put the commercial reality plainly: if an agent works around the clock, does not complain and does not take holiday, the incentive to hire a junior weakens. He said he has already seen recent graduates in large IT labour markets struggling more than before to find a first role.

Neither guest argued for avoiding AI. Their advice was to avoid over-relying on it:

• Keep building by hand. Piotr's guidance was to write some of the software yourself and use AI as a pair programmer rather than delegating fully, as a way to keep skills sharp and keep growing.

• Build domain depth. Knowledge of a sector such as fintech or energy lets an engineer solve business problems, not just generate code, which makes them far more valuable.
• Protect business continuity. Piotr warned against efficiency that leaves a single point of failure. A team of one person plus nine agents is fine until that person leaves. Organisations still need to hire and grow juniors to protect continuity.

Rocky also pointed to a wider responsibility:

"We have a social responsibility to make sure that when students are graduating, there is a future for them. We need to work together with universities to make sure the curriculum is sufficiently adapted to future needs."

Why is platform engineering the backbone of AI-accelerated delivery?

Piotr was direct:

"Platform engineering is no longer nice to have. It's the core of the delivery setup."

Access to AI tools is not the same as a consistent way to use them. Skip platform engineering, Piotr argued, and you end up with fragmented security, duplicated infrastructure and spiralling costs, including uncontrolled token usage.

 Platform teams now own more than CI/CD and infrastructure. They decide which models can be used, what data those models can be exposed to, and how agents are governed across teams so delivery can scale and stay secure.

Rocky described the same approach at enterprise scale. With more than 10,000 engineers globally, ING pairs platform teams with what it calls enabling teams: early adopters who help other squads onboard onto shared standards and guardrails. 

Without that governance, he warned, thousands of engineers building their own agents become "user computing on steroids" and one team's good work never reaches another.

Data foundations: garbage in, disaster out

AI does not fix weak engineering practices. It exposes and multiplies them. As Rocky put it:

"It's garbage in, disaster out, if you're not careful."

Poor-quality data, unclear ownership and inconsistent governance do not disappear when AI arrives; they scale with it. Asked whether the data infrastructure layer is keeping up, both guests agreed compute is available if you are willing to pay for it. 

The real constraint is data quality. Reliable data, clear ownership and proper governance are the prerequisite for getting value from AI, and the risk is highest when AI output reaches customers directly.

Managing the cost of AI: FinOps for engineering teams

Assuming AI is free is the wrong assumption, and both guests treated cost as an engineering decision, not just a finance one.

• Match the model to the task. Rocky's teams are educated not to use the most complex model for the simplest job, and to optimise prompts for efficiency.

• Watch token consumption. He warned against a business process stalling because a month's token budget is consumed in a day.
• Expect prices to rise. Piotr noted the real cost of AI tooling is not yet fully passed on to customers. With a large gap between provider investment and income, both expect token prices to climb, which makes active cost management a standing discipline.

Rocky also flagged the ESG dimension: efficient use of AI is a cost and a sustainability question.

Security and ethics: guardrails, human-in-the-loop, red-teaming

Conceptually, security has not changed. What has changed is scale: more code produced means more exposure to threats. 

A weak security setup, Piotr noted, only multiplies under AI.

Rocky's rule for a regulated bank was blunt:

"Blind trust is never wise in anything you do."

The human stays in the loop, and trust, once lost, is hard to recover. 

The panel's practical approach combined old and new:

• Build guardrails and validation into the pipeline before anything reaches production, including AI agents that review the work of other agents.
• Use AI on the defensive side too: scanning images and dependencies for vulnerabilities, generating richer test data for edge cases, and running red-teaming with AI agents for internal ethical hacking.
• Keep doing the established basics, such as penetration testing and security audits, more frequently, because model capabilities move fast.

Responsibility for the outcome still sits with the engineering team. AI can accelerate delivery; it does not transfer accountability.

What engineering leaders should take away

Hear the full discussion between Rocky Woestenborghs and Piotr Krzepczak, including their views on AI-enabled software engineering, platform engineering, data governance, engineering leadership and the future of junior talent.

Watch the full recording here:

If you're hiring across DevOps, platform engineering, software engineering or AI infrastructure, speak to the Abutalib (tshah@montash.be) about building the skills your organisation needs next.