How trust works in teams and why the old models are breaking

Executive overview

Trust is not declining — it is redistributing. Across history, trust has moved through three phases: local (proximity-based), institutional (entity-based), and distributed (network-based). Each shift changed who and what people trusted, and how quickly.

The current crisis is a design problem. Most institutions were built as pyramids; modern trust flows sideways. In remote and hybrid workplaces, leaders who don't adapt lose trust faster than they realise.

The trust pause — the moment of slowing down to ask whether someone is worthy of your trust — has been designed out of modern life, and that absence is dangerous.

The three chapters of trust

• Local trust: proximity and reputation-based; worked when communities were small and stable
• Institutional trust: arose with cities and trade; innovations like contracts, insurance, and brands allowed trust through entities rather than individuals
• Distributed trust: enabled by digital platforms; for the first time, strangers could be trusted at scale, mediated by technology (ratings, reviews, marketplaces)
• A fourth phase is emerging — auto sapient trust — where we trust AI not just to do things, but to decide and create

Why we give trust away too easily

• 80% of people are naturally trusting; the default is to extend, not withhold, trust
• People seek the illusion of control — information from influencers feels like control over outcomes (diet, fitness, parenting)
• Platforms are designed for speed and efficiency; friction — including the trust pause — has been removed
• Trust decisions are now driven by emotional response: how information makes someone feel, not who is saying it or whether it's factual
• Context is stripped out: an influencer's advice doesn't account for your specific circumstances

The three phases of distrust on teams

1. Defensiveness — the person is still engaged; this is the critical intervention point
2. Disengagement — withdrawal from meetings, effort, and reliability; tracked via absence and inconsistency
3. Disenchantment — active attempts to turn others against the leader; at this stage, exit is usually required

• The cycle from defensiveness to disenchantment accelerates in virtual environments — physical absence causes people to fill in gaps negatively
• 80% of leaders regret how long they kept a disenchanted person on the team
• Avoidance is the main reason: discomfort, fear of a gap, hope it resolves itself

How to catch distrust early

• Intervene at the defensiveness phase; the conversation is uncomfortable but far less so than later phases
• Name the behaviour without accusation: "I noticed you were very invested but it came across as defensive — where's that coming from?"
• Listening to the explanation often creates a deep trust moment
• Disengagement requires hard, data-backed feedback on reliability and presence
• Disenchantment requires removal; leaving someone in that state causes ongoing damage and mental drain for everyone

Building trust in hybrid and remote teams

• Proximity no longer generates trust; leaders must replace it deliberately
• Two traits distinguish effective remote leaders: consistent expectation-setting and consistent presence
• Clarify what you will and won't do; erratic behaviour — odd-hour messages, unpredictable responses — erodes trust fast
• When teams meet in person, protect time for unstructured conversation; don't fill it with agenda
• Intentional in-person time carries virtual collaboration for weeks afterward

Trusting AI: a framework for leaders

• Trust has two dimensions: capability (competent, reliable, does what it says) and character (integrity, empathy, humility)
• Previous technology was trusted for capability; AI is increasingly being trusted for character — a fundamentally different and riskier proposition
• Map AI use cases on a 2x2: high trustworthiness + high trust (good), high trustworthiness + low trust (design problem), low trustworthiness + high trust (danger zone), low trustworthiness + low trust (low impact)
• The danger zone — where people over-trust AI in areas it isn't ready for — is the key risk to manage
• AI companies' core challenge: identify and fix misalignment between their interests and the interests of users and contributors; unresolved misalignment compounds over time (see social media)