AI speeds coding but not enterprise software delivery

ClearRoute has released its State of the Route to Live 2026 report on enterprise software delivery. It argues that while AI has sped up coding, it has not improved the broader process of getting software into production.

The consultancy drew its findings from four years of software delivery assessments across financial services, retail, healthcare, media and technology. It found that many large organisations still face delays in testing, security, compliance, governance and release management, even as AI tools help developers write code more quickly.

According to the report, median lead time to production in enterprises remains 30 to 45 days. In one assessed company, a business-critical feature took 266 days to go live, while fixing a critical production bug took 158 days.

The figures highlight a gap between faster software creation and actual delivery. The main constraint, ClearRoute argues, is not coding itself but the operational processes around it, particularly where teams still rely on manual approvals, fragmented environments and release models built for control rather than continuous change.

Manual processes also appear linked to higher failure rates. Organisations with high levels of manual testing and release management record change failure rates of 10% to 20%, compared with less than 5% for what the report describes as elite teams.

Productivity gap

The commercial effect can be significant. ClearRoute cited one global financial institution where an annual engineering investment of GBP £2.76 million yielded only 5% of its value in delivered features. Another organisation needed between three and six months for new engineers to make their first contribution.

This suggests AI may amplify existing strengths and weaknesses rather than resolve them. Teams with mature delivery processes can move faster, but those with weak foundations may simply produce more code without increasing the rate at which software reaches users.

In comments accompanying the report, James Jarvis, Chief Executive Officer of ClearRoute, said: "AI has changed the speed of software creation, but not the speed of enterprise delivery. For many organisations, the bottleneck was never writing code, but the manual approvals, brittle test suites, fragmented environments, governance processes and release constraints surrounding them.

"The organisations that win from AI will not be the ones with the most coding assistants. Speed is the new competitive advantage, which means the next competitive frontier is operationalising AI safely across the full Route to Live and moving beyond fragmented tools to platform engineering as a core operational capability. This will close the gap between decision and delivery, getting change live faster, safer and at scale."

The report presents this as an organisational problem as much as a technical one. Faster code generation does little to shorten delivery cycles if testing remains manual, release approvals are slow and development teams work across disconnected systems.

Shift to agents

ClearRoute also points to a change in how businesses are using AI in software development. Many organisations, it says, are moving beyond code completion tools towards systems connected to software environments, including agent-led code review, autonomous testing and more automated delivery pipelines.

That shift raises new questions about control and oversight. Businesses adopting AI agents without stronger identity, access and governance structures risk replacing one form of tool sprawl with what the consultancy calls "agent sprawl".

Sarndeep Nijjar said this next phase would depend on tighter technical boundaries. "The next phase of AI in software delivery will be defined by control. Enterprises are moving from individual copilots to AI agents that can interact with pipelines, environments and production systems. That creates huge opportunity, but only if those agents operate inside clear technical boundaries. Without that foundation, enterprises will not scale AI safely; they will simply create another layer of complexity.

"This is where platform engineering becomes critical. It gives organisations the governed pathways, automated guardrails and reusable delivery foundations needed to move from AI experimentation to AI operations."

ClearRoute works with large organisations across industries including banking, retail and technology. Its report suggests the pressure on those companies is no longer just to write software faster, but to reduce the time between deciding to make a change and getting that change live in production.

While AI tools have changed developer workflows, the report argues that the broader software lifecycle remains the harder problem. In that view, enterprises that fail to modernise testing, governance and release processes may find that faster code generation exposes delivery weaknesses rather than solving them.