Access to AI tools does not guarantee adoption from your people. If employees don’t know where their accountability ends and the algorithm’s begins, they can freeze. Dr. Ravinder Tulsiani argues that the key to AI culture isn’t just upskilling and blanket training, it’s establishing the psychological safety of clear decision rights.
Walk into any organisation that has recently rolled out a generative AI pilot, and you will likely find two distinct groups of people. The first group is enthusiastically experimenting, generating code, emails, and strategies with abandon. The second group, often the majority, is hesitant. They have taken the mandatory “AI Fundamentals” course. They have the license. Yet, they remain reluctant to integrate the tool into their core workflow.
The issue is rarely a lack of skill. It is a lack of safety.
It is easy to label this second group as resistant to change or lacking in digital agility. But in my work diagnosing organisational constraints, I have found that the issue is rarely a lack of skill. It is a lack of safety.
We are asking employees to collaborate with a non-human entity that hallucinates facts, bypasses traditional creative processes, and operates as a “black box.” Simultaneously, we are telling them that “you remain responsible for the output”.
Without explicit boundaries, this creates a state of decision paralysis. Employees are unsure if relying on AI makes them efficient or lazy. They are unsure if an AI error will be treated as a system glitch or a personal failure of judgment.
Decision rights as psychological safety
To fix this, L&D and HR leaders need to look beyond the technical curriculum and address the psychological contract of work. We need to reframe “Decision Rights” not just as a compliance necessity, but as a tool for psychological safety.
In the Domino Map™ framework, a diagnostic operating system I use to help organisations structure learning, we focus heavily on mapping the “Decision Rights Grid.” This exercise is less about policing usage and more about liberating the human worker.
When an employee knows exactly which decisions are “Human-Only” and which are “AI-Supported,” the anxiety dissipates and clarity allows them to move.
The three zones of work
To build a healthy human-AI culture, leaders must explicitly categorise work into clear zones. This transparency signals to the workforce that their human judgment is still valued and protected.
1. The human-only zone (The Zone of Judgment)
There must be explicit areas where AI is kept out. These are decisions involving nuance, high-stakes ethics, leadership empathy, and complex personnel matters. When organisations fail to define this zone, employees fear that “everything” is up for automation, leading to defensive behaviour. By drawing a circle around these tasks, leaders say: “Your humanity is the value-add here.”
2. The AI-supported zone (The Zone of Collaboration)
This is the messiest area, and the source of most anxiety. Here, AI generates the draft, the code snippet, or the plan, but the human decides. The danger here is “automation bias”, the tendency to blindly trust the machine to avoid the cognitive load of checking it. Culturally, we must train people to see scepticism as a skill, not a sign of slowness. The metric of success here isn’t speed; it’s the quality of the edit.
3. The AI-automated zone (The Zone of Relief)
These are low-risk, repetitive tasks where we explicitly want the human to step back. Defining this zone is an act of care, it removes the drudgery that leads to burnout. But even here, the “escalation trigger” must be clear. When the machine fails, how does the human step back in without fear of blame?
Moving from “prompt engineering” to “decision engineering”
The current obsession with “prompt engineering” is a distraction. The enduring skill of the future is “decision engineering”, the ability to know when to delegate to a machine and when to intervene.
For L&D practitioners, this means our role is shifting. We are no longer just content creators; we are architects of workflow. We cannot simply throw tools at a team and hope for the best. We must first diagnose the emotional and structural constraints of the environment.
I recently worked with a team that was struggling to adopt an AI planning tool. The issue wasn’t the interface; it was the “Escalation Rule.” No one had defined what would happen if the AI’s forecast was wrong. Once we established a clear protocol, defining exactly when the human planner should override the system, adoption skyrocketed. The employees didn’t need a better tutorial; they required permission to trust their own judgment over the machine’s.
AI safety
The organisations that will thrive in this next era won’t necessarily be the ones with the most advanced models. They will be the ones with the most clarity.
By establishing a diagnostic-first culture that defines outcomes and decision rights before assigning training we do more than improve efficiency. We build a workplace where humans can use powerful tools without the paralysing fear of undefined accountability.
We don’t just need AI-literate employees. We need employees who feel safe enough to lead the machines they have been given.
Dr. Ravinder Tulsiani is Principal, Interim Enterprise Capability & L&D Advisor at Catalyst Learning
