Part 4: Rewards. What We Measure and Celebrate

Process defined how work flows; Rewards determines whether people are paid, praised, and promoted for keeping it flowing or for working against it. How people are measured and rewarded influences how they carry out their work each day. The challenge is to motivate and reinforce the behaviours that add value to the organisation.

What we reward and celebrate:

• Peaceful excellence over stressful heroism: sustainable delivery, not late-night saves. This names what we are replacing as well as what we are building. Teams that have been delivering through heroics need explicit permission to stop.
• Adaptiveness and learning over predictability and control
• Collaboration and synergy over individual heroics
• Systemic thinking: asking “what about the system made this likely?” rather than “who made the mistake?”
• Teams that complete their planned work and maintain flow over teams that produce “record outputs” at the expense of others
• Raising concerns and admitting mistakes. Silence is the most dangerous signal.

What we do not reward:

• Busyness. High utilisation without throughput is waste; as Goldratt puts it, activating a resource and utilising it are not the same thing.
• “Record outputs” that create logjams for other teams
• Hoarding knowledge or creating single points of failure
• Local optimisation at the expense of the whole

The research on burnout reinforces the “peaceful excellence” choice. The Accelerate dataset found the factors most correlated with burnout are environmental, not personal: organisational culture, deployment pain, leader effectiveness, and organisational investment, echoing Maslach’s finding that most organisations try to fix the person and ignore the work environment, even though changing the environment succeeds more often (Forsgren, Humble & Kim, Accelerate, 2018, citing Leiter & Maslach). A rewards system that celebrates heroics is an environmental burnout factor wearing a medal.

Growth Maps and Performance Assessment

Our growth maps are built around a Builder Core + Specialisation Wings model. Every role (Engineering, Product, Management) develops a shared set of Builder Core capabilities: AI Orchestration, Problem Discovery, Rapid Prototyping, Quality Thinking, Communication, Domain Insight, Collaboration. Each person also develops capabilities specific to their specialisation wing.

Performance assessments split into Builder Core (~50%) and Specialisation (~50%). As the Builder Core grows over time, accelerated by AI making specialised skills more accessible, its weight increases. This makes the convergence toward full-stack “builders” gradual and measurable.

Breadth is the primary growth axis; depth is the amplifier. The default path to seniority is end-to-end capability: the ability to take a feature from user need to production release. Specialisation depth makes you more effective but is not an alternative track to seniority. Someone who is world-class in one area but cannot operate outside it is not senior in this framework.

L-levels reflect orchestration scope, not just expertise depth:

• L1–L2: Learning to use AI tools; task-level contribution
• L3–L4: Directing multi-agent workflows; feature-level ownership
• L5–L6: Designing organisational operating models around AI-augmented teams

What this means for hiring: Every job description includes Builder Core capabilities. We stop hiring pure coders who can’t talk to users, pure PMs who can’t prototype, and managers who lack AI fluency. This is not a preference; it is a refusal to build the wrong capability for what is coming. The bar for Builder Core competency is non-negotiable at every level and every specialisation.

What this means for compensation: Existing pay bands hold initially, mapped from Builder titles to market benchmarks (e.g., Senior Builder in Engineering benchmarked against Senior Software Engineer). Over time, we introduce cross-track recognition and may move toward unified builder bands with specialisation premiums.

Rewards alignment is critical. If we say we value collaboration but measure individual output, the system sends a contradictory signal. You get what you measure.