Your AI Code Tools Are Only As Good As Your Foundations

Authored by David Sugden, Head of Engineering

With AI code assistants everywhere, are organisations realising their full potential?

The 2025 DORA report landed on a finding that should give every engineering leader pause: AI is a mirror and a multiplier. In high-performing teams, it accelerates delivery and improves quality. In struggling teams, it amplifies existing weaknesses. Even with over 90% of software developers reportedly using AI in the workplace, many organisations are failing to see productivity gains translate into business outcomes.

Our latest AI White paper reports that the differentiator is not having access to the latest AI tooling, but everything around it.

Speed without safety is just faster failure

The promise of AI coding assistants is compelling – who doesn’t want faster development, automated testing, and a reduction in repetitive toil? And at the individual level, the gains are real. Developers report saving several hours per week, and teams using AI heavily are shipping significantly more pull requests than those who are not.

But speed without the right controls creates problems that compound over time.

At Axiologik, we have worked with several organisations over the last 12 months where AI-generated code was duplicating existing logic and increasing code complexity. Our observations are supported by research by GitClear who found that duplicated code blocks increased 8x in the 18 months to late 2024, driven largely by AI assistance without adequate guardrails.

And yet, we suggest these are not only AI problems. They are engineering excellence hygiene problems that AI makes visible faster. Our own work confirmed the absence of quality checks and blocking gates that was allowing defects introduced by both humans and AI to reach production undetected.

Governance must scale with capability

One principle underpins everything we see in successful AI-enabled engineering organisations: governance must keep pace with capability. Giving a team access to agentic AI workflows while their CI pipeline runs often-ignored, advisory-only scans and their test coverage sits below 30% is not innovation. It is a risk that will eventually surface as technical debt, security incidents, or both.

DORA’s research reinforces this. Teams that adopted AI without strong platform engineering foundations actually experienced negative performance impacts. At the same time, the organisations pulling ahead are those treating AI adoption as a change management challenge, aligning access to demonstrated engineering maturity rather than simply rolling out the latest tools to all their development teams.

Assessing where you stand

Axiologik’s “AI in Engineering” maturity framework benchmarks organisations across categories such as core development practices, tooling, infrastructure, quality, process, governance, operations, sustainability, and developer experience. It is designed to give engineering leaders an honest, evidence-based view of where AI can deliver value today and what foundational work needs to be planned out first.

As part of every AxioIntelligence engagement, our experts will benchmark your engineering practice, identify the gaps between current and target maturity, and build a phased adoption roadmap that ties AI enablement to measurable controls. The result is a clear path from experimentation to confident, scalable AI-assisted delivery.

Foundations first, then velocity

The organisations that will gain the most from AI in engineering over the next 12 months are not the ones rushing to enable agentic workflows. They are the ones whose leaders are investing in secure foundations, enforced quality gates, structured training, and measurable controls, who are then accelerating from a position of confidence.

AI does not fix broken processes. It makes them break faster. Get the foundations right, and the returns follow.

Ready to benchmark your engineering AI maturity? Talk to us about an AxioIntelligence engagement.