I was inspired by a recent LinkedIn post by Alastair Lechler, which discussed the importance of psychological safety in the context of AI adoption. He described how companies like Microsoft, Adobe, Spotify, and Atlassian are creating "AI sandboxes" where teams can experiment without fear, resulting in a 3.5x increase in viable AI use cases compared to teams without similar environments.

The post resonated deeply with my experience as an AI strategist and educator. While leaders pour billions into AI technologies, many overlook the human element that determines whether such investments flourish or flounder. The uncomfortable truth is that AI adoption isn't primarily a technological challenge — it's a deeply human one.

A 2024 study by the American Psychological Association confirms this, finding that 41% of workers fear AI will make parts of their job redundant (and report higher workplace stress as a result). Furthermore, in environments with high psychological safety, 66% of employees were confident their employer would retrain them if AI replaced their job — 10% higher than in workplaces lacking psychological safety.

These numbers confirm what I've observed in practice: psychological safety goes much farther than establishing the right environment for experimentation. When people fear looking incompetent, making mistakes, or being replaced, their survival instincts kick in. Some "turtle up" and retreat into safe zones, nodding politely in meetings but quietly avoiding engagement. Others adopt defensive or even hostile behaviors, manifesting as corporate territorialism or outright refusal to participate.

These reactions are entirely understandable and a natural human trait. But if not properly addressed — by establishing an environment of psychological safety — innovation, technology adoption, and change of any sort becomes nearly impossible.

In this article, I'll explore:

• What psychological safety is and why it's particularly crucial for AI adoption

• Practical approaches for building psychological safety for AI transformation

• Common traps organisations fall into when addressing psychological safety

• Essential next steps for leaders navigating the intersection of people and AI

The Hidden Piece of the Puzzle: Psychological Safety

The term “psychological safety” was coined in the 2000s by renowned Harvard professor, Amy Edmondson, who defined it as a “shared belief held by members of a team that the team is safe for interpersonal risk-taking”. In other words, team members feel they have permission to speak up and be honest when discussing mistakes, problems, and anxieties, all without fearing negative consequences to their image or career.

Numerous studies have found conclusively found that teams with higher psychological safety outperform others, especially when it comes to knowledge-intensive and creative endeavours, because they freely share ideas, learn from failures and collaborative more effectively on novel challenges.

In the context of technology adoption, in particular AI, psychological safety becomes even more of a make-or-break element of success. Here’s why:

1. Automation Anxiety

The start of my career in investment banking coincided with the 2008 Global Financial Crisis, and I still keenly recall the deep-seated anxiety that gnawed at me daily as I worried the latest round of redundancies would include my team.

It's no surprise that I deeply empathise with employees today who feel the same dread about AI potentially taking their jobs. The current wave of AI innovations will impact both "lower-skilled" workers as well as highly knowledge-intensive or creative roles such as consultants, lawyers, analysts, marketers, artists, and journalists.

These fears — some valid and others potentially misplaced — often lead to technology rejection, even where it's most needed. I encountered this scenario when working with a European retailer whose Creative team, despite having the most relevant use cases for Generative AI, were initially the most resistant to adoption. Without much experimentation, they declared that AI was simply "not good enough for our work."

2. Competence Reset

AI is disrupting established expertise and career progression dynamics.

While domain knowledge remains invaluable, proficiency with AI creates a new dimension of competence. Seasoned professionals may struggle with the technology while junior, more "technology-native" colleagues adapt quickly. This shift rarely creates true "equal footing," but it can threaten professional identities built on traditional expertise.

This "competence reset" applies differently, but sometimes even more strongly, to younger employees. With AI potentially automating traditional entry-level tasks — research, first drafts, data analysis — many young professionals worry that their opportunities to apprentice and learn fundamentals are diminishing.

I experienced this firsthand when working with a private equity firm. A newly hired investment associate shared his excitement about AI while simultaneously expressing deep anxiety about how the technology could upend his fledgling career.

3. New Learning Paradigm

Introducing AI tools fundamentally changes workplace learning approaches. Unlike previous technologies with established playbooks, AI implementation remains largely uncharted territory, creating unique challenges:

• Experimentation is essential: While vendors provide technical documentation, the truly valuable knowledge — how to effectively integrate AI into specific workflows and where human oversight remains critical — must be discovered through firsthand trial and error. Without psychological safety, teams default to conservative applications that barely scratch the surface of AI's potential.

• Professionals must redefine their roles: When AI can handle tasks that once formed the core of one's professional identity, individuals must reconsider where their unique value lies. This self-directed evolution requires space to experiment with surrendering certain responsibilities while expanding into areas where human judgment remains paramount.

• Continuous adaptation is required: The rapid evolution of AI capabilities means even experts must regularly return to being novices. Without psychological safety, organisations may find initial AI momentum quickly stalling as teams become reluctant to repeatedly place themselves in vulnerable learning positions.

This "learning zone" with its constant experimentation can only flourish when team members feel safe to try, err, reflect, learn, and try again without fear of diminished status or career repercussions.

4. Trust and Reliability Gap

Given the extensive publicity about Generative AI's tendency to "hallucinate" or generate plausible-sounding but incorrect responses, many employees — especially those without sufficient training — are uncertain about whether and how to use such tools. Some avoid the technology altogether rather than risk their work or reputations on potentially unreliable outputs.

The opaque nature of AI systems also represents a fundamental shift from previous technologies. Unlike tools such as Excel where the relationship between inputs and outputs is transparent, most AI technologies operate as "black boxes." Users must trust processes they cannot fully verify — a significant adjustment for professionals accustomed to understanding their tools completely.

Early users of Generative AI are often bewildered by the probabilistic nature of outcomes — submitting the same prompt multiple times can produce similar but not identical outputs. This contrasts sharply with the deterministic outputs of software tools most professionals are accustomed to.

Piecing Together a People-First Approach to AI Adoption

The challenges outlined above might seem daunting — and they are. But they're far from insurmountable. In fact, recognising these human dynamics is the first step toward addressing them effectively.

Through my work guiding teams through AI transformation, I've observed a clear pattern: organisations that proactively build psychological safety and focus on the human element of change significantly outperform those that focus exclusively on the technical aspects of AI adoption.

Let's explore five key strategies that consistently create the psychological conditions for successful adoption:

1. Setting the Tone from the Top

Leaders set the cultural foundation for AI transformation by modeling the behaviours they want to encourage and by demonstrating both openness and vulnerability.

For example, when a CEO acknowledges, “I too am learning about these new systems, and I sometimes make mistakes,” it invites employees to share their concerns without fear of judgement. By actively inviting input in team meetings, “What are your thoughts, and where do you see potential challenges?”, and listening appreciatively to every contribution, leaders create an atmosphere of mutual respect and engagement.

The right tone also includes explicitly addressing the fears that naturally accompany technological shifts, such as concerns about job security or skill obsolescence. Directly speaking in public about the issue, such as in a town hall or in an open letter, is important, even when leaders themselves may not necessarily have all the answers.

I remember a promise that a functional leader at one of my clients gave to his team at an away day, “I know many of you are concerned about the potential impact of AI on your jobs. We don’t have all of the answers yet and we cannot promise that your jobs will remain unchanged. But what we can promise is that the leadership team will do everything in our power to give you the opportunity and skills to thrive in this new environment.” These words, simple yet powerful in their honesty and sincerity was an important catalyst for the organisation's AI exploration journey.

2. Reaffirming Human Value

Counterintuitively, successful AI adoption begins with reaffirming human value, not technological capability. When organisations communicate that they value people's uniquely human qualities, it creates the security needed for experimentation and more broadly, to embrace change.

For this to happen, organisations must delineate where AI ends and human judgment begins, and should articulate specific domains where human skills such as creative ideation, ethical decision-making, stakeholder management, and contextual understanding of organisational history and culture, remain essential. This clarity provides psychological anchoring, giving employees confidence that their professional identities aren't under threat, while directing teams to develop abilities that complement rather than compete with AI.

Equally important is consistent messaging about AI as an augmentation tool rather than a replacement. Organisations that successfully build psychological safety emphasise how AI handles routine tasks to free human talent for higher-order thinking. A practical way of doing so is for leaders to share concrete examples of how AI enhances human work, transforming employee anxiety into curiosity.

Finally, domain expertise remains particularly critical in the AI era. While AI systems can process vast amounts of information, they lack the nuanced understanding that comes from years of immersion in a field. The most effective organisations position subject matter experts as essential for validating AI outputs, providing contextual interpretation, and identifying subtle patterns that algorithms might miss.

A financial services firm I worked with exemplified this by mapping which aspects of an analysts role might shift to AI and which dimensions — client relationships, strategic insight, risk oversight — would become more central to their roles. This exercise allowed team members to envision their evolving value rather than fixating on potential losses.

3. Normalising Learning Loops

As discussed above, AI adoption inherently involves experimentation, which by nature includes the risk of failure. In psychologically safe environments, mistakes aren't viewed as failures but as vital learning opportunities. Teams with high psychological safety treat these missteps as valuable data points that reveal insights for improvement, as opposed to blemishes on their record.

Here are some ways in which organisations can normalise learning loops:

• Rituals that normalise constructive discussion of mistakes: One of my favourite techniques, that I use frequently with clients, is to conduct "premortems" and "postmortems" for AI initiatives. Conduct "premortems" before initiatives launch, imagining failure and working backward to identify potential issues. Follow with "postmortems" after implementation, reviewing actual outcomes without blame and capturing lessons. These bookend practices create a continuous learning cycle normalising failure discussion at every project stage.

• "Failure spotlights" or awards for lessons learned: Publicly celebrate valuable insights gained through unsuccessful attempts. For instance, the Tata Group's "Dare to Try" Award honors courageous failures and learning from worthwhile but unsuccessful innovative endeavours. These recognition mechanisms actively remove the stigma from the word "failure" and reinforce the notion that "errors are data" — valuable signals for improvement rather than reasons for punishment.

• Incentives that are aligned with these behaviours: Behaviour follows incentives. When evaluations and compensation structures punish experimentation or reward only flawless execution, employees avoid risks regardless of leadership rhetoric about embracing mistakes. Microsoft under Satya Nadella exemplifies effective alignment by holistically evaluating employees on individual impact, contribution to others' success, how they leveraged others' work, and on their willingness to learn and experiment.

• Low-stakes environments foe : The creation of low-stakes environments for experimentation — such as the AI sandboxes or innovation labs that were noted at the start of this article — can also be powerful mechanisms for supporting teams to safely experimentation with new approaches.

4. Nurturing Human Capabilities

Successful AI adoption ultimately depends on the people using the technology. While tools and platforms matter, the human capabilities gap often determines whether AI initiatives flourish or falter.

As a university and corporate educator, I've discovered that both effective learning and psychological safety begin with showing, not telling. When teams see AI addressing their specific challenges, the technology becomes less threatening and more accessible. For marketers, demonstrating how AI can optimise email campaigns they currently spend hours crafting creates a bridge between present skills and future possibilities. For financial analysts, showing how AI can identify patterns in quarterly data they manually review builds confidence that the technology enhances rather than eliminates their analytical judgment.

Creating psychologically safe learning environments also means embracing playfulness and normalising experimentation. In workshops, I intentionally incorporate activities like having teams compete to generate absurd AI product descriptions, resulting in shared laughter over "ergonomic water" and "quantum-infused sticky notes." These moments break hierarchical barriers and create shared vulnerability, both important cornerstones of psychological safety.

Psychological safety in training also means acknowledging different adoption paces — while some team members eagerly experiment, others need more time to build comfort. Here are some ways in which varying learning needs and preferences can be accommodated:

• Multi-speed training programmes: In my client engagements, I've found success blending in-person workshops with on-demand content that participants can revisit independently, which has proven valuable for more gradual learners. Differentiated cohorts, progressing at a pace suitable for each group, can be formed based on initial surveys about comfort and experience levels with AI.

• One-to-one or small group coaching: These facilitate psychological safety through confidential spaces for vulnerability. This model acknowledges that not everyone is immediately ready to display their learning curve publicly. In personalised mentoring sessions I’ve provide, I've witnessed profound shifts as employees move from defensive resistance to cautious experimentation.

• Peer support networks: Establish "AI champions" or mentors within teams as accessible resources for colleagues to ask questions without fear of judgment. These champions — often early adopters rather than technical experts — provide both practical guidance and emotional reassurance, helping reframe AI as opportunity rather than threat.

By designing learning pathways that allow for different speeds, organisations create space for everyone to succeed, preventing those moving more deliberately from feeling discouraged.

5. Establishing Communication Channels

Even with supportive leadership, psychological safety requires structured pathways for communication. Here are some approaches I’ve used:

• Organisational baselining: Begin with a comprehensive organisational survey that assesses not just technical readiness, but also motivational levels and emotional responses to AI, making sure to documenting concerns, anxieties, hopes, and expectations. This provides a sanctioned space to voice apprehensions while giving leadership crucial insights into the psychological landscape.

• Regular pulse checks: Continue emotional pulse checks throughout the transformation journey. Emotions fluctuate over time, and momentum built in early stages can dissipate if not closely monitored and maintained.

• Anonymous feedback mechanisms: Provide ongoing opportunities for candid input through multiple channels that respect different comfort levels with transparency. Consider a variety of modalities such as blinded digital suggestion boxes, physical "concern cards" available in meeting rooms that can be submitted anonymously, dedicated Slack channels with anonymising features.

• Employee representation: Offer employees across different levels representation in the organisation’s AI implementation and ethics committees, and signal that their voices matter in key decisions, thereby reinforcing psychological safety. Diverse perspectives are in any case essential when implementing AI, and junior team members — whose feet are closer to the ground — may pick up on important implications that might be missed with executive-only input.

Effective communication structures recognise that psychological safety requires both vertical and horizontal pathways, ensuring that the inevitable uncertainties of AI adoption have channels for expression rather than festering as unaddressed anxieties.

Avoiding Psychological Safety Traps in AI Transformation

While many organisations recognise — on the surface at least — the value of psychological safety in AI transformation, many fall into predictable traps that undermine their efforts. Here are the pitfalls you need to avoid:

1. Making Symbolic Gestures Without Substance

Perhaps the most damaging trap of all is the disconnect between what organisations say about their intentions and what they actually do. When employees notice such disconnects, they very quickly become doubtful and cynical. I’ve seen this misalignment manifests in several ways:

• Empty slogans and promises: Leaders often outwardly claim to promote an “experimentation-led” transformation where “failure is celebrated”. Yet many often react with impatience or disappointment when early AI implementations don't immediately deliver perfect results. Employees quickly become cynical when they see this gap between the espoused values of experimentation and the actual responses to the messy reality of adoption.

• Inconsistent leadership signals: Even minor leadership signals can dramatically impact psychological safety for AI adoption. In one leadership workshop I facilitated, the CEO waded into a spirted discussion with comments like, "Where are we going to find the money for that?” or “How does that make sense?”. The effect was chilling and immediate — the rest of the leadership team instantly became hesitant to propose further ideas. Such micro-reactions, seemingly minor, can have an incredibly deleterious effect on morale and psychological safety.

• One-sided investment priorities: When organisations describe AI as a tool for augmentation and enhancement but only approve projects focused on headcount reduction, employees immediately detect the incongruence. I've observed how organisations that pursue a balanced portfolio of AI initiatives — combining efficiency goals with innovation and employee experience improvements — build more credible transformation narratives.

2. Confusing Psychological Safety with Being Nice

A frequent misunderstanding is equating psychological safety with making everyone comfortable or being "nice" all the time.

In fact, a psychologically safe culture does not equate to a superficial "feel-good" atmosphere and instead encourages candid debate and even uncomfortable conversations, so long as they are respectful and constructive. It's not about avoiding conflict or never challenging each other, but instead about removing the fear of punishment or recrimination for doing so.

This is especially critical in the context of AI transformations due to the multitude of tough questions that need to be answered such as which tasks should be automated versus augmented, where human oversight remains critical, the "competence reset" that AI triggers, and how roles must evolve in response to new technology, not to mention the risks, biases, and operational challenges arising from AI implementation.

Organisations that swing the pendulum too far toward artificial harmony will ultimately face having the numerous challenges associated with AI transformation being swept under the rug, resulting in more anxiety rather than less.

3. Treating Psychological Safety as a One-Time Event

A critical mistake many organisations make is treating psychological safety for AI adoption as a finite project rather than an ongoing commitment. This fundamental misunderstanding leads to short-lived initiatives that ultimately fail.

The hard truth is that psychological safety may require deep seated organisational and cultural change that cannot be achieved through isolated events. Away days, team-building exercises, or AI hackathons might generate temporary enthusiasm, but without consistent reinforcement, this momentum inevitably dissipates.

Creating psychological safety demands deliberate, ongoing investment with clear responsibility, specific strategies, and allocated resources. The most successful organisations approach it as core infrastructure requiring regular maintenance — particularly as AI capabilities evolve.

Rather than viewing psychological safety as a box to check, effective organisations integrate it as an essential component of their transformation approach, recognising that sustainable change happens through consistent reinforcement at all levels.

Building Your Psychological Safety Roadmap

Now that we've explored both the essential approaches for building psychological safety and the common traps to avoid, the question becomes: how do you begin this journey in your own organisation?

Creating a psychologically safe environment for AI adoption doesn't happen by chance — it requires intentional action and sustained commitment. Let's look at three practical steps to start building your psychological safety roadmap.

1. Assess the Current State

Look honestly at your organisation's readiness by asking questions (and soliciting honest feedback) such as:

• When someone feels uncomfortable about AI risks, do they feel comfortable speaking up about it?

• If an AI programme doesn't go as planned, does it tend to be discussed openly or quietly shelved?

• Would a junior team member feel safe raising concerns about an AI project that senior leaders are enthusiastic about?

• Are people given the time and space to try new approaches (or as I like to call it “messing around”!) with AI, even if some don't pan out?

2. Start Small and Adapt

Rather than attempting organisation-wide transformation immediately, start small by selecting a team with a meaningful AI use case to experiment with different techniques and discover which practices resonate most with your unique organisational culture.

This living laboratory becomes both proving ground and adaptation space, where practices aren't simply imported but organically evolved to fit your company's DNA. When approaches are tailored to your specific context rather than generically applied, they gain authenticity and sustainability that generic best practices often lack.

3. Lead by Example

And last but perhaps most importantly, start by committing yourselves as leaders to this journey.

Effective leaders model vulnerability by acknowledging their own AI learning curves and sharing missteps. They avoid the "empty slogans" trap by matching public statements about experimentation with genuine patience when implementations struggle.

By protecting learning spaces, rewarding thoughtful risk-taking, and demonstrating that expertise now includes comfort with uncertainty, leaders create permission throughout the organisation to approach AI with curiosity rather than fear.

No pressure there at all!

Conclusion

Too often we forget that the competitive advantage doesn't come from AI itself, but from how effectively humans can adapt to and alongside it.

This is why the organisations outperforming in AI adoption aren't necessarily those with the most advanced technology or largest investments but are instead those creating environments where people feel secure enough to embrace change.

The journey toward psychological safety isn't a one-time initiative but an ongoing commitment to cultivating an environment where teams can navigate technological change with confidence rather than fear. As AI continues to evolve, this human foundation will become even more crucial to organisational success.

If you're seeking to build this foundation in your organisation, I invite you to connect with me to discuss how these principles might be applied to your specific context.

Justin Tan is passionate about supporting organisations to navigate disruptive change and towards sustainable and robust growth. He founded Evolutio Consulting in 2021 to help senior leaders to upskill and accelerate adoption of AI within their organisation through AI literacy and proficiency training, and also works with his clients to design and build bespoke AI solutions that drive growth and productivity for their businesses. Alongside his consultancy work, he is an Adjunct Assistant Professor at University College London, where he lectures on digital health and AI. If you're pondering how to harness these technologies in your business, or simply fancy a chat about the latest developments in AI, why not reach out?