AI coding assistants have become ubiquitous across enterprise development teams, and the latest model releases have crossed a capability threshold, fundamentally changing how experienced engineers generate production code with AI. While AI speeds up coding, the greatest opportunity to accelerate innovation lies in improving quality, security, and speed across the entire software lifecycle, not just the coding stage. In this interview, GitLab Chief Product & Marketing Officer Manav Khurana explains why businesses still aren’t seeing the efficiency impacts of developer productivity, and why intelligent orchestration is the solution.
Q: AI coding assistants are everywhere now, and companies are reporting impressive productivity gains. Yet many leaders still aren't seeing those improvements translate into faster delivery. Why is this?
A: Organisations are broadly reporting developer productivity gains of 20-50% when using AI. The problem is that coding accounts for only about 52 minutes per day of the software delivery process. Accelerating just that one stage creates a challenge for everything that follows, and bottlenecks shift downstream: code review queues lengthen, test suites become burdensome, and security scans surface more vulnerabilities than teams can remediate. This is the "AI Paradox."
Amdahl's Law explains why. When you make 20% of a process 10x faster but leave the remaining 80% unchanged, you only get a 1.25x overall improvement. Leaders measure end-to-end delivery - time from concept to production - and organisations that adopted AI only for coding see disappointing results. Developers feel more productive, but the business sees minimal improvement in time-to-market. The real opportunity isn't making coding faster. It's rethinking how the entire software delivery process operates with AI.
Q: Are there specific challenges that come with traditional software development workflows that are compounded when you introduce AI?
A: Traditional software delivery follows a linear path: plan, code, test, secure, deploy, operate. Each stage has multiple tasks, and handoffs between tasks and stages introduce delays and coordination overhead. When AI accelerates only coding, each transition between stages destroys the velocity gains.
A developer uses AI to write code in an hour that previously took a day. But then they wait for code reviews, for testing to complete, for security scans to surface vulnerabilities they need to triage, for the next release window to open. At each handoff, they may lose critical context or the "why" behind their decisions, the constraints that shaped their implementation, and the business requirements that drove the feature. As AI helps developers generate more code, they're pushing higher volumes through a system designed for much slower throughput.
This is why intelligent orchestration, which enables asynchronous collaboration between developers and AI agents across the entire software development lifecycle, matters.
Q: Walk us through what intelligent orchestration is and how it works.
A: Intelligent orchestration replaces sequential stages with continuous execution loops. Instead of "code, then test, then secure," teams continuously generate, test, secure, deploy, and verify work in parallel. Agents execute autonomously while humans orchestrate from above, setting direction and governance.
Intelligent orchestration is built on three foundational pillars:
Workflows: Multiple agents collaborate on complex tasks like issue-to-merge-request flows, security analysis, code reviews, and CI/CD operations. One agent serves numerous developers. For developers, this means fewer handoffs and less waiting. They can keep coding while agents handle testing, security scans, and deployment preparations in parallel. For CTOs, this translates to higher throughput without adding headcount, as one agent can serve an entire team of developers. Humans steer and guide rather than micromanage. Teams define which context to rely on, workflows to streamline, and compliance rules to enforce.
Context: A unified data model maintains continuous execution across stages. Instead of losing context at handoffs, you have complete visibility into code, requirements, history, security implications, deployment constraints, and operational feedback. Developers no longer waste time explaining their decisions in handoff meetings or digging through Slack threads to reconstruct the reasons behind their choices. The context travels with their work. CTOs gain real-time visibility into the entire delivery pipeline, understanding exactly where bottlenecks occur and what's blocking teams without requiring status meetings. What used to be tickets, waiting, handoffs, and remediation sprints becomes continuous generation, continuous compliance, and continuous improvement.
Guardrails: Governance and compliance are built into the workflow, not bolted on afterward. Agents assess risk and recommend appropriate levels of autonomy. Low-risk changes proceed autonomously. Higher-risk changes trigger review workflows. For developers, this means security and compliance happen automatically in the background rather than becoming last-minute blockers before release. For CTOs, this provides the confidence to move faster without increasing risk, as governance is consistent and auditable across all teams and projects. Policy gates enforce rules automatically, maintaining velocity without sacrificing security or compliance.
Q: If AI is writing most of the code today, are there different or new engineering skills that are becoming valuable?
A: GitLab's research shows 82% of DevSecOps professionals in Australia believe AI will create more engineers, not fewer. What's changing is the nature of that engineering work. Developers must evolve from writing every line of code to orchestrating systems and guiding AI agents - shifting from logistics-driven to vision-driven work. Organisations that help engineers make this transition will thrive.
Skills that used to be expected of senior engineers become baseline requirements, such as breaking work down into well-defined pieces, making sound architectural decisions, taking a product-minded approach, mastering automated testing and observability, and tracking tech debt. Human skills such as creativity, strategic vision, judgment, and architectural thinking become more critical.
Q: Can you give me some real-world examples of companies that are getting this right?
A: At Southwest Airlines, Managing Director of Technology Grant Morris and his team are transforming how the airline builds and delivers software. By standardizing CI/CD pipelines via a unified platform, Southwest has brought consistency to metrics, security, and code quality across its enterprise. This enables their team to ship mission-critical software faster while maintaining the reliability and resilience required for 24/7 operations across 4,000 daily flights.
As a second example, Ericsson manages enterprise software deployments across 300+ global communications service providers. After adopting intelligent orchestration, they achieved 50% faster deployments and saved 130,000 hours over six months. They're now delivering updates in weeks rather than months. Daniel Costa Soares, Head of Software Automation and Support, Ericsson OSS/BSS, reported, “The faster deployment capabilities…directly contribute to our business growth. Communications service providers choose us because we can deliver innovation at the speed their markets demand.”
Additionally, CERN, where 10,000 scientists from over 100 countries collaborate on particle physics research, achieved 90x faster job startups through unified orchestration. The lab’s Tech Lead and Engineering Manager, Ismael Posada Trobo, says his team relies on a single platform to build, update, and host the software tools powering their research.
These are stories about intelligent orchestration enabling teams to maintain velocity at enterprise scale. It’s from eliminating handoffs, maintaining context across stages, and enabling continuous execution. The pattern repeats across industries and company sizes. The common thread is replacing fragmented toolchains with unified platforms designed for the entire software lifecycle.
Q: What’s your advice for technology leaders who are stuck in the AI paradox?
- Don't throw more tools at the problem. The AI paradox isn't just a technology challenge. It's a people, process, and technology problem that requires rethinking all three.
Start with your people. Are your developers spending their time on creative problem-solving and strategic thinking, or are they just triaging a larger backlog?
Then examine your processes. Sequential workflows - code, then test, then secure, then deploy - create friction that no amount of AI tooling can overcome. When you enable continuous execution where testing, security, and deployment happen in parallel rather than in sequence, context stops getting lost and governance becomes automatic.
Then evaluate your technology. Enterprise leaders must determine whether their platform enables teams to work in flow or forces them to coordinate across disconnected tools. The hundreds of steps in software development - ensuring quality, security, compliance, maintainability - need to accelerate together, not individually.
When you align people, process, and technology around intelligent orchestration, teams shift from being logistics-driven to vision-driven.
