Is AI scaling your worst habits? 

1. Codify your values

Translate your values into observable, trainable behaviors so people know what “good” looks like. 

Install:

• Behavioral scorecards in performance reviews (e.g., how well someone shares knowledge) 
• Values-based interview prompts (e.g., “Tell me about a time you disagreed and still helped your colleague succeed.”) 
• Embedded behaviors in rituals (e.g., shout-outs for cross-functional wins in stand-ups) 

Best practice: 

Netflix famously links values such as “candor” and “judgment” to specific behaviors in its performance framework. These show up in hiring, firing, and promotion decisions. 

2. Make feedback a system

Feedback should be a normal, recurring structure designed to surface dissent early and make improvements safe. 

Install:

• Weekly feedback loops
• Anonymous feedback tools
• Dissent dashboards that track how often feedback is raised, acted on, and closed 
• Retrospectives that include behavior-based KPIs (e.g., “How well did we live our value of transparency this sprint?”) 

Best practice: 

Bridgewater Associates runs radical transparency through structured feedback tools and recorded meetings. People rate each other in real time, and feedback is built into daily operations. 

3. Build human-AI decision protocols 

AI should operate under explicit decision logic that’s visible to teams. 

Install:

• AI-human decision matrices (e.g., customer support: AI auto-responds to tier 1, with human escalation for tone or nuance) 
• Escalation paths for ethical flags (e.g., bias spotted in hiring recommendations) 
• Audit trails for every high-stakes AI-influenced decision (e.g., budget forecasts, hiring, terminations) 

Best practice: 

Salesforce’s “Office of Ethical and Humane Use” has protocols indicating when humans must step in, including reviewing hiring algorithms and marketing AI output for equity issues. 

4. Audit culture

Culture must have its own audit loop to surface dissonance before it calcifies. 

Install: 

• Values-to-behavior scans (e.g., values actually showing up in daily actions) 
• Dissent and silence metrics (e.g., how often people challenge ideas or hold back) 
• Cultural heatmaps by team (e.g., who reinforces culture vs. who quietly drifts) 

Best practice: 

Allianz regularly captures feedback on leadership trust, decision clarity, and ethical pressure. Its Allianz Engagement Survey feeds into business unit planning and leadership KPIs. If a red flag shows up, leaders are expected to act immediately. 

5. Deploy leaders like models 

Leaders encode norms, so how they handle disagreement, make trade-offs, or admit fault is mimicked by teams and systems. 

Install:

• Decision journals (leaders publicly write their assumptions and learnings) 
• Trade-off visibility in meetings (“Here’s what we’re prioritizing and what we’re giving up.”) 
• Walkthroughs of reasoning aloud to teams (so logic becomes shared, not just followed) 

Best practice:

Amazon’s Narratives Over Slides policy forces leaders to document the thinking behind decisions, creating institutional memory and accountability.