Photo credit: loveholidays - used with permission
George Malamidis, VP of Engineering at loveholidays
Interviewed by Matthew Skelton, co-author of Team Topologies.
“ By adopting Team Topologies, creating the BOOST Enabling team, and enhancing commercial delivery with measures of technical health, we built a resilient socio-technical architecture. When generative AI transformed software development, our organizational foundations and automated CodeScene guardrails allowed us to absorb agentic coding rapidly, driving delivery throughput to historic highs while continuously improving system code health.”
Executive summary: 5 key takeaways for transformation leaders and CIOs
At loveholidays, our mission is to open the world to everyone by offering unmatched choice, ease, and value in travel. Operating as the UK's fastest-growing travel agent requires processing trillions holiday offers daily for millions of monthly visitors. As our engineering organization expanded from 4 to 13 teams—and eventually to 130 personnel across data, product, and engineering—we recognized that scaling technical capabilities without losing velocity or code quality required an intentional socio-technical operating model. By combining Team Topologies principles with Andrew Grove's High Output Management, we established small, cross-functional stream-aligned teams alongside specialized platform teams and a dedicated enabling unit called the BOOST team. This foundation transformed our delivery metrics and served as the indispensable blueprint for safely scaling AI-native software development across our engineering estate.
5 Key Takeaways
Stream-Aligned Structure: Structuring 130 data, product, and engineering staff into small, ~5-person cross-functional mission teams backed by dedicated platform teams established aligned autonomy at scale while eliminating communication silos.
Enabling Team Catalyst: Deploying the BOOST squad as an Enabling team of principal engineers using facilitation, empathy, and shared SLAs solved the "technical excellence paradox" without top-down mandates.
Balanced Performance Measurement: Evaluating teams across a four-part performance framework (OKRs, business KPIs, technical KPIs, and team health) enabled loveholidays to double deployment frequency to 10,000 deployments per month while driving production incidents down.
Prerequisite for AI Scaling: Establishing disciplined organizational practices, clear team ownership, and strict structural guardrails provided the vital foundation needed to prevent AI adoption from compounding technical debt.
Deterministic Quality Loops: Integrating CodeScene quality metrics and MCP servers directly into agentic development loops enabled loveholidays to achieve over 80% AI-assisted commits while elevating system-wide hotspot code health above 9.75.
Problems Prior to the Adoption of Team Topologies
During our early growth phase around 2019, loveholidays faced the classic inflection point of a scaling technology company. As our organizational footprint scaled from 4 to 13 engineering teams, maintaining technical excellence across autonomous units grew increasingly complex. Teams became siloed, knowledge sharing slowed down, and establishing consistent standards across the codebase became a major operational hurdle.
Specifically, we encountered four interrelated organizational challenges prior to adopting Team Topologies:
Unaligned Autonomy is Ineffective: At loveholidays, our philosophy centers on technical excellence—building reliable, scalable, and maintainable systems end-to-end. However, team autonomy created unexpected barriers to systemic code health. Mission-oriented teams were heavily incentivized to ship visible commercial features, whereas the less visible work of maintaining architectural standards, reducing latency, and upgrading infrastructure was difficult to champion and coordinate across teams.
Communication Overhead in Growing Teams: Unchecked team growth led to communication friction, siloing, and increased cognitive load for engineers operating without clear operational boundaries.
Lack of Operational Continuity ("Feature Factory" Risk): Developers frequently built features, passed tests, and moved on without ongoing operational care. Without explicit structures for continuous stewardship, systems accumulated unmonitored errors, elevated build times, and recurring production defects.
Imbalanced Incentives: Commercial delivery naturally receives strong organizational support. Without an objective framework to measure technical health alongside strategic delivery, foundational engineering investments were frequently pushed aside for short-term commercial goals.
These challenges are present in many prowing and established organizations, not just loveholidays. Where we were different was in recognizing these problems early on and deciding to do something about them - before they turned into commercial friction.
Implementing Team Topologies and the Four-Part Performance Framework
To resolve these challenges, we turned to strategic organizational frameworks—specifically Team Topologies by Matthew Skelton and Manuel Pais, paired with Andrew Grove’s High Output Management. While I dislike rigid coding frameworks, I do find that strategic organizational toolkits like Team Topologies provide a shared language and common patterns that align cross-functional teams toward unified goals.
Small, Cross-Functional Stream-Aligned Teams
Beginning in late 2019, we restructured our department around small, cross-functional stream-aligned teams ("mission teams") composed of approximately five people. Keeping team sizes small minimized communication overhead, prevented internal silos, and accelerated context acquisition. These mission teams held end-to-end ownership of specific business domains, such as Checkout, Payments, or Flight Supply.
To maintain economies of scale and prevent teams from reinventing infrastructure, we balanced these stream-aligned units with specialized Platform teams, such as Cloud Infrastructure and Data Engineering, allowing us to create some economies of scale across the organization.
The Four-Part Performance Framework
Inspired by Andrew Grove’s focus on measuring both quantity outputs and quality metrics, we created a homegrown performance framework evaluating every team across four responsibilities:
Strategic Delivery (OKRs): Assessing progress against core commercial goals and product innovations.
Domain Business KPIs: Optimizing domain-specific metrics unique to each stream-aligned team.
Functional Technical KPIs: Standardizing cross-cutting technical vitals across all teams cutting code, including error counts, production incidents, build success rates, and DORA metrics.
Qualitative Team Health: Running frequent team sentiment checks to ensure developer satisfaction and sustainable working rhythms.
Correlating metrics across these four pillars enabled leadership to identify when a team was burning capacity firefighting production incidents at the expense of strategic innovation. The metrics act as a “guiding light” to help teams stay aligned while executing with autonomy.
Deep Focus: The BOOST Enabling Team Pattern
Ensuring that autonomous teams prioritized technical excellence required an explicit organizational mechanism. To fulfill this role, we created BOOST (Boost: Observe, Optimize, Support, Transform), composed of our principal software engineers. BOOST operates as a Team Topologies style Enabling team mandated to champion technical excellence and elevate engineering standards.
Detecting Cross-Cutting Problems and Scaling Innovations
Rather than operating as a centralized command body or a firefighting squad, BOOST acts as a cross-pollinator across all 130 data, product, and engineering staff. BOOST identifies breakthrough engineering patterns in isolated teams and propagates them across the department. For example, after the Event Sourcing pattern proved effective in our booking systems, BOOST principal engineers worked hands-on alongside the Checkout team to implement the exact same Event Sourcing architecture pattern, creating a reusable blueprint across loveholidays.
Operational Mode: Facilitation, Empathy, and Shared SLAs
BOOST principal engineers engage with delivery teams through facilitation and coaching rather than dictatorial mandates. Principal engineers act as force multipliers whose primary metric is organizational impact. While BOOST operates under Service Level Agreements (SLAs) for technical vitals, when metrics slip, BOOST engineers do not take over the codebase; instead, they facilitate workshops and pair with team members to resolve root causes, preserving team autonomy and accountability.
Scoreboards and Shared Accountability
Applying principles from The 4 Disciplines of Execution, we converted our technical KPI dashboard into a public "scoreboard" tracking lead metrics—such as daily application error reductions and build pipeline speeds—that are actionable by engineers and predictive of system stability. BOOST principal engineers share direct accountability for these targets: if technical KPI targets are missed, BOOST and team leadership share the performance impact. This alignment transforms technical quality into a collaborative competition across teams.
Photo credit: loveholidays - used with permission
Outcomes Provided by Team Topologies
Implementing Team Topologies, supported by BOOST and our performance framework, delivered dramatic improvements in delivery velocity and operational reliability:
Doubled Deployment Frequency: Deployment frequency increased to over 10,000 deployments per month across ~100 delivery personnel (including 70 software engineers and product managers who deploy routinely).
Reduction in Application Errors: Focusing on daily error reduction as a lead metric reduced application errors to fewer than 20,000 per month across applications serving over 40 million sessions monthly.
Systematic Decrease in Production Incidents: Tracking lead metrics directly drove down lag metrics. Severe production incidents dropped systematically, creating a sustained "zero surprises" engineering culture.
Optimized Build Pipelines: BOOST spearheaded targeted optimizations that made build pipelines 3x faster even as total build volume expanded 4.4x year-over-year.
Long-Term System Stewardship: Aligning with our principle to "leave things better than you found them," teams developed a culture of ongoing care ("you build it, you run it"), treating code as a living asset requiring continuous stewardship.
Deployments at loveholidays between July 2025 and September 2026, showing a 2x increase.
Application errors at loveholidays between July 2025 and September 2026, showing a significant decrease as AI commits increased.
From my perspective, Team Topologies provides the scaffolding that enables a healthy culture of accountability and stewardship to grow. The improvements in operational metrics clearly link back to the ways that teams are aligned around value flows and therefore outcomes.
Team Topologies as the Foundation for AI-Native Software Development with CodeScene
When generative AI and agentic coding tools emerged (around 2023), many organizations rushed to adopt automated assistants without updating management or architectural guardrails. At loveholidays, our established Team Topologies foundation allowed us to capitalize on AI rapidly and safely.
Why Team Topologies Was an Essential Prerequisite
As noted in the DORA 2025 State of AI-Assisted Software Development report, AI acts as an amplifier of an organization's existing culture: strong practices get multiplied, but poor practices and messy architecture get multiplied just as fast. Without clear team boundaries, ownership, and quality standards, agentic coding quickly generates "unguarded complexity" and compounds technical debt.
Because Team Topologies had established clear domain boundaries, small team footprints, and an active Enabling team in BOOST, loveholidays possessed the socio-technical infrastructure needed to govern AI adoption effectively.
Integrating CodeScene and Defining "AI-Ready" Code Health
To manage AI rollout across all 130 personnel, we adopted the DX AI framework focusing on Adoption, Cost, and Impact, tracking co-authored AI commits rather than vanity lines of code. However, early unguided AI adoption initially caused code health in hotspot areas to dip. Standard prompt engineering alone in CLAUDE.md files proved insufficient to prevent AI agents from committing structural technical debt.
To solve this, loveholidays partnered with CodeScene to embed automated CodeHealth guardrails directly into the agentic development loop:
Deterministic Quality Bar: We established a CodeScene CodeHealth score of 9.8 as the minimum bar for code to be considered "AI-Ready". LLM agents are significantly more successful and consume fewer tokens when operating on codebases with health scores above 9.75.
Commit-Hook Enforcement: We embedded CodeScene CLI checks as automated commit hooks. If an AI agent generated code dropping repository health below 10.0, the hook rejected the change and forced the agent into an automated refactoring loop.
CodeHealth MCP Server: We integrated CodeScene's Model Context Protocol (MCP) server directly into developer IDEs, providing real-time code health feedback to agents while drafting code.
Our approach to AI-native engineering was industry-leading, and we’ve since learned that many other organizations are adopting a similar discipline following a loveholidays case study published by CodeScene that goes into more detail.
Evolving Team Constraints and Context Acquisition
AI-native engineering is now reshaping traditional Team Topologies constraints. Historically, stream-aligned team sizes were capped at five people because acquiring mental context to run complex systems was expensive. AI agents have made context acquisition cheaper and faster. Supported by runbooks and AI context agents, loveholidays is transitioning toward broader departmental support networks where engineers can safely support neighboring systems, reducing on-call fatigue while maintaining high-trust stewardship.
Results of AI-Native Scaling
Within twelve months of deploying CodeScene guardrails backed by our BOOST enabling architecture:
Over 80% of all committed code across loveholidays became AI-assisted.
86% of engineering teams achieved "AI-Ready" hotspot code health (>9.75), with 100% rated as healthy (>9.0), placing loveholidays in the top 5% of the software industry.
AI Build Success Rates: AI-assisted build success rates surpassed 90%, while the DORA change failure rate remained rock-solid at 1–2%.
P90 Lead Time to Deploy: Lead time to deploy changes dropped from 138 minutes to 44 minutes.
The CodeHealth™ metric (from CodeScene) at loveholidays between July 2025 and September 2026, showing shift towards 100% health.
Incidents at loveholidays between July 2025 and September 2026, showing a decrease in incidents following the use of CodeHealth™ guardrails
The combination of a strong “stewardship scaffolding” (provided by Team Topologies), balanced delivery metrics, and the BOOST Enabling team helped us to adopt and sustain an AI-native engineering approach that results in higher quality code than without AI. But this is not “vibe coding”; this is disciplined engineering with clear guardrails.
Conclusion: Systems Thinking and Sustainable Innovation
Scaling loveholidays into an AI-first travel platform handling trillions of offer calculations proved that technical excellence cannot be left to chance or forced through top-down mandates. By adopting Team Topologies, creating the BOOST Enabling team, and enhancing commercial delivery with measures of technical health, we built a resilient socio-technical architecture. When generative AI transformed software development, our organizational foundations and automated CodeScene guardrails allowed us to absorb agentic coding rapidly, driving delivery throughput to historic highs while continuously improving system code health.
Next steps: applying Team Topologies at scale
Request a keynote talk to inspire your teams and kickstart a transformation initiative
Order 50 or more copies Team Topologies Second Edition now in bulk - with free worldwide shipping
Buy an Enterprise Content License with access to a huge library of official Team Topologies workshop and training materials to catalyse large-scale transformation
Get an online consultation on Team Topologies with one of the co-authors of the book
About the author:
George Malamidis, VP of Engineering at loveholidays
George Malamidis is the Vice President of Engineering at loveholidays. Over his seven-year tenure, George has led the evolution of loveholidays' engineering operating model, pioneering the adoption of Team Topologies, continuous delivery performance frameworks, and AI-native software development practices. Before loveholidays, George held engineering positions at DRW, ThoughtWorks and others.
