Beyond Critical Thinking: 13 Durable Skills Driving AI Adoption

Recent research by Gerlich (2025) found a significant negative correlation between frequent AI tool usage and critical thinking abilities. This was particularly evident in younger participants, who showed higher dependence on AI tools and scored lower on critical thinking assessments compared to older participants. The study attributes this decline to ‘cognitive offloading’, the delegation of thinking tasks to machines, which appears to undermine our capacity for independent analysis.

However, Multiverse recognises that as the world of work evolves, so too will our conceptualisation of intelligence and the skills required for effective AI interaction. Beyond just critical thinking, there exists a whole new set of durable skills that individuals must master to harness AI’s potential.

Our research aimed to investigate the specific durable (soft) and cognitive skills that enable successful AI adoption and integration in the workplace.

We had 3 research questions:

We used the following definitions of durable and cognitive skills:

Durable (soft) skills refer to personal attributes and social abilities like communication, adaptability, and ethical awareness that enable effective human interaction and collaboration, representing uniquely human competencies that cannot be algorithmically replaced (Amann & Stachowicz-Stanusch, 2020; Kumar, 2023).

Cognitive skills refer to the mental abilities and processes fundamental to acquiring knowledge and understanding, including analysing, applying, creating, and reasoning, which are essential for learning, decision-making, and critical evaluation of AI outputs (Zhai et al., 2024; Gerlich, 2025).

To ensure the authentic representation of these human skills, we employed a Grounded Theory approach. This is a data led, iterative process that builds theoretical frameworks directly from data, rather than testing pre-existing hypotheses. This allowed us to observe human behaviour in an AI environment, extract and pinpoint core skills from this raw data.

We conducted this observational research using Think Aloud Protocol Analysis (TAP; Ericsson & Simon, 1993), a research method which gathers verbal reports as data. The participants, 20 of Multiverse’s AI users ranging from beginner to expert level, verbalised their thoughts and decisions as they carried out daily tasks using AI. This was paired with follow-up interviews to understand participants’ perceptions of the way they interacted with AI.

After collecting our initial data, we conducted thematic analysis which highlighted a set of 13 skills with examples of how each skill optimises AI use in the workplace.

These address research question 1, ‘what specific durable and cognitive skills are essential for successful and effective AI use in the workplace, and why?’, and research question 2, ‘how is task performance using AI affected when the relevant durable and cognitive skills are not present?’

Below, you can see an example skill that was evidenced in our research, ‘Tailoring Communication’. As alluded to above, this example shows how grounded theory research was used to identify specific skills. We analysed the raw data and grouped themes together, undergoing a process of iteration and refinement which eventually led to our final skillset of 13.

1. Tailoring communication: Discerning whether AI output has the desired tone for a particular audience or situation, and refining prompts if it is not.

This skill was observed as participants reviewed AI outputs to ensure a match with their desired tone, to sound like the human user, or to be appropriate for a particular audience. In the TAP analysis, a participant talked about understanding their environment in relation to AI’s outputs:

Another participant reflected on the key soft and cognitive skills they employed in their AI interactions:

We also captured evidence addressing research question 2, as participants reflected on the consequences of not tailoring their communication when using AI:

Whilst this participant candidly explains:

A note on critical thinking...

Interestingly, the evidence we captured for cognitive skills when using AI echoes established research demonstrating that when people anticipate future access to information, they exhibit lower rates of information recall but enhanced recall for information location and access methods (Sparrow, Liu & Wegner, 2011). This suggests that memory storage is being relocated rather than diminished, prompting us to reconsider which cognitive abilities are most valuable when working alongside AI systems. Our research supports this phenomenon, suggesting that the challenge lies not in cognitive decline as Gerlich’s research concluded, but in determining which skills to prioritise in an AI-augmented work environment.

Addressing research question 1, the full set of our 13 critical skills for AI adoption is listed below, along with their groupings:

1. Analytical reasoning: Breaking down complex information for AI to more effectively deliver its instructions; recognising tasks that AI is or is not suitable for.

2. Creativity: Pushing the boundaries of AI use and experimenting with new approaches to drive innovation.

3. Systems thinking: Identifying patterns in AI performance to predict how AI will respond to a task.

4. AI ethics: Spotting bias and recognising how it affects AI outcomes; using AI outputs in an ethically sound way to inform business recommendations.

5. Cultural sensitivity: Identifying when AI outputs lack sufficient geographic or cultural awareness.

6. Curiosity: Examining the broader context and requirements of a task to augment AI outputs.

7. Self-regulated learning: Reflecting on the success of a chosen AI approach; partnering with AI to self-assess its outputs.

8. Detail orientation: Fact checking AI for hallucinations and errors; using one’s own domain expertise to ensure accuracy.

9. Adaptability: Iterating and refining one’s approach to interacting with AI based on the quality of outputs.

10. Determination: Patience and willingness to continue trialling new approaches with AI, even during unsuccessful AI interactions.

11. Empathetic interaction: Treating AI as an extension of one’s own mind and thoughts; anthropomorphising AI to create more thoughtful, receptive, and intentional dialogue.

12. Tailoring communication: Discerning whether AI output has the desired tone for a particular audience or situation, and refining prompts if it is not.

13. Exchanging feedback: Using AI to proactively seek feedback on work.

Finally, addressing research question 3, our research also revealed that participants at four different AI experience levels exhibited distinct characteristics.

Interestingly, we found that female participants consistently underestimated their AI competency in self-assessments, requiring upward adjustments to a higher experience rating based on observed performance - highlighting important implications for how AI confidence is perceived across demographics.

In addition to answering our research questions, we have addressed a critical gap in the literature by conducting bottom up, grounded theory based research. Almost every piece of research or articles written about durable (soft) skills relies on pre-existing definitions of durable and cognitive skills. Our inductive research, on the other hand, observes how these skills naturally emerge and manifest in real workplace contexts - allowing us to discover authentic skill categories which reflect how humans behave in relation to AI.

Multiverse has already recognised the importance of these soft skills and successfully mapped them onto our existing learning programmes. For example, in our AI for Business Value programme, the technical requirement to ‘model business processes using relevant techniques, standards, notation and software tools’, directly connects with the durable skill of ‘Creative Thinking: being confident enough in one’s own AI abilities to push the boundaries of AI use’, demonstrating how durable skills are essential for mastering technical skills.

Additionally, being able to identify these skills allows us to progress towards being able to assess them and measure them, helping employee’s develop deeper and more sustainable AI capabilities beyond more basic AI awareness and technical skills.

There are several key takeaways for leaders from this research:

Make strategic AI investments: Rather than pursuing blanket AI adoption that can reach billions in expenditure, leaders should evaluate tools based on their specific use cases and longevity, and whether they will unlock your company’s potential or hinder progress. Consider reframing your company’s skill development priorities towards transferrable soft and cognitive skills which in turn enhance any technical competency.

Crucially, focus on investing in learning as much as the tools themselves - creating the time, space and resources for deep and lasting AI adoption is as critical an investment as purchasing the technologies.

Map existing training: If your organisation has existing AI that requires technical training but you aren’t seeing progress in AI adoption, consider mapping that training against our newly identified durable skills. This approach may increase adoption and learning of your already-invested AI technologies. Leaders can also identify where relevant AI durable skills naturally align with technical competencies and integrate them, rather than treating them as separate initiatives.

Normalise cognitive offloading: Help your teams understand that relying on AI for certain tasks isn’t cognitive laziness, but strategic resource allocation that exercises an entirely new set of cognitive capabilities. Leaders can model and encourage when it is appropriate to use AI, while still valuing uniquely human contributions.