Last week, Phipps Cameron had the pleasure of co-hosting a roundtable breakfast event at Franco’s on Jermyn Street in partnership with Plandek. Over a delicious breakfast, senior technology leaders gathered to discuss the profound shift happening across software engineering: "AI is rewriting the SDLC. What replaces it is the differentiator."
While AI tools have dramatically accelerated line-by-line code generation, the consensus in the room was clear; generating code faster does not automatically mean delivering commercial value faster. As AI reshapes software engineering, new friction points are emerging across the entire delivery lifecycle.
Thank you to everyone who attended and contributed to such an engaging, insightful discussion. For those unable to join us, here are the six key takeaways from the event:
1. The Bottleneck Is Moving Downstream
AI has successfully removed much of the friction from initial code generation, but that speed is exposing delays further down the line. Constraints are rapidly shifting toward testing, code review, product management, and architecture. Plandek’s research highlights this divide: lower-performing AI-enabled teams take over 35 hours to merge pull requests, compared with under 21 hours for top-performing teams.
2. Testing and Quality Assurance Are Becoming Critical
Faster code output puts unprecedented pressure on validation and QA. Google’s 2025 DORA research found that while AI adoption improves delivery throughput, it can negatively impact delivery stability. Without modernising QA practices, organisations risk letting defects, rework, and technical debt erode their hard-won productivity gains.
3. Product Management Is Emerging as a Major Bottleneck
Engineering teams can now build software faster than product organisations can define, scope, and prioritise requirements. Aligning engineering capacity with business strategy is vital. Plandek’s benchmarks reveal that high-performing teams dedicate over 41% of their total capacity to strategic roadmap delivery, compared to less than 21% among lower-performing teams.
4. Architecture and Technical Leadership Matter More Than Ever
AI-generated code still requires experienced architectural oversight, seamless system integration, and robust governance. When coding speed increases, flawed architectural decisions compound technical debt at an alarming rate. Strong engineering leadership, modern platforms, and clear standards are essential to guide AI implementation safely.
5. AI Productivity Gains Vary Significantly
The impact of AI is not uniform across organisations. Plandek’s data shows that lower-performing teams adopting AI saw lead time to value drop by nearly 50%, whereas already high-performing teams saw incremental gains of 10–15%. Crucially, organisations that proactively resolve planning, review, and quality bottlenecks achieve more than double the output per engineer compared to AI-enabled teams that leave these constraints unaddressed.
6. The Real Opportunity Is End-to-End SDLC Transformation
According to Google’s 2025 DORA research, 90% of tech professionals now use AI at work, with over 80% reporting productivity improvements. However, boosting individual developer efficiency does not automatically equal superior organisational performance. The true differentiator lies in rethinking the broader engineering operating model - realigning team structures, product leadership, and performance metrics to convert AI usage into measurable commercial impact.
Get in touch
Interested in joining future Phipps Cameron roundtable discussions or exploring leadership talent within tech and engineering? Contact our team today - get in touch with Tristan Dugmore in the first instance on Tristan.dugmore@phippscameron.com
