There's no doubt AI tools can optimize workflows to accelerate delivery. But we are only beginning to understand the long-term effects of AI on organizational culture and engineering management.
In organizations moving toward an AI-first model, teams may indeed become more "efficient" in the short term, but it will be harder for individuals to internalize new practices, evolve culture, and adapt under pressure.
In my daily interactions with engineers, the trade-off is already visible: greater efficiency is hindering ownership, weakening accountability, and fading support systems that people once relied on for growth and development.
My intention here is to share some insights from day-to-day experience on how AI is reshaping collaboration inside organizations. I also want to offer practical advice to engineering leadership on how to continue accelerating business impact while preserving the culture and human aspects that attract, grow, and retain great talent.
The Social Network Effect on Performance and Collaboration
In 2021, when enterprises were still adjusting to a fully remote setup, I conducted research on collaboration and social networks within communities of practice. The goal was to understand which levers could strengthen knowledge sharing and foster relationships in a global enterprise that had gone fully remote during the pandemic.
We interviewed approximately 200 senior leaders from different teams across multiple countries. We were curious to learn about their collaboration ties: the people they contacted to collaborate in the past three months, the reason for collaboration, the strength of those relationships, and the level of trust they had in each other.
The findings revealed that the highly connected individuals who bridged teams and geographies were consistently seen as top performers. Their network position had a direct effect on both perceived and actual performance, perhaps because tacit knowledge and visibility are products of dense collaboration.
Recently, I applied this lens to engineering teams and simulated collaboration networks, using pull request (PR) reviews as an example.
PR review interactions are more than a workflow checkpoint. They are powerful mechanisms for collaboration, knowledge sharing, mentorship, and the reinforcement of shared norms.
But these interactions are slow. They hurt sprint velocity metrics.
And the AI tools that are rapidly becoming available to teams promise faster reviews, fewer bottlenecks. A great relief to a big pain point.
The trade-off is that teams gradually replace human-to-human exchanges with AI interactions. The reinforcement loop breaks. What Damon Centola calls "complex contagion" (change and trust spreading through multiple trusted peers) is replaced by a single validating node.
The Network Shift
The effects are visible in the simulation:
Pre-AI: A dense graph, multiple engineers at the center, resilient knowledge flow.
Post-AI: One "AI Assistant" node absorbs most interactions. The graph collapses into a hub-and-spoke model, with AI at the center and humans at the periphery.
When AI centralizes influence, organizations become weaker at adapting, learning, and embedding change.
Closing Thoughts
Collaboration networks are the invisible infrastructure of organizational culture. Every micro-interaction is a brick in that culture.
When unchecked, AI can weaken the ties that make organizations strong. The job of leadership isn't just integrating AI into workflows, but protecting and designing the human networks that keep culture alive.
Actions for Leaders
- Use AI to amplify impact, not replace human connection.
- Partner with HR to review talent management practices so performance isn't defined by speed alone, but by quality collaboration and knowledge sharing.
- Invest in onboarding with strong human support, not just AI assistants.
- Actively encourage mentorship, coaching, and communities of practice.
- Track not only velocity, but the health of collaboration networks.