Introduction
Research and practitioner discussions surrounding generative AI have largely focused on adoption. Questions concerning whether firms use generative AI, how frequently they use it, and what benefits they obtain have dominated both academic and managerial conversations. While these questions are important, they leave a significant gap in our understanding of how generative AI influences organisational decision-making.
Adoption alone tells us very little about the quality of managerial judgement.
Two organisations may use the same generative AI system and achieve very different outcomes. One organisation may use AI as a valuable source of insight that enhances strategic thinking. Another may become overly dependent on AI-generated recommendations, reducing critical reflection and increasing the risk of poor decisions. A third may underutilise the technology entirely, failing to benefit from its analytical capabilities.
The difference is not necessarily explained by the technology itself. Rather, it may be explained by how managers determine when, why, and to what extent they rely on AI-generated outputs.
This article argues that future research should move beyond adoption-focused explanations and place greater emphasis on trust calibration. Specifically, I propose a conceptual framework that explains how SME managers calibrate trust in generative AI during strategic decision-making under uncertainty.
The central argument is that decision quality depends not simply on AI capability but on the relationship between perceived AI competence, managerial expertise, contextual uncertainty, and organisational verification practices.
The limitations of adoption-based explanations
The literature on technological innovation has traditionally treated adoption as a key indicator of success. Models such as the Technology Acceptance Model (TAM), Diffusion of Innovations Theory, and the Unified Theory of Acceptance and Use of Technology (UTAUT) have generated valuable insights into why individuals and organisations adopt new technologies.
However, generative AI presents a challenge to these approaches.
Knowing that a manager uses ChatGPT, Claude, Gemini, or another generative AI system does not reveal how that manager uses the technology during strategic decision-making. Adoption measures cannot explain:
- whether AI recommendations are critically evaluated
- whether managers challenge AI-generated assumptions
- whether AI is treated as an advisor or an authority
- whether AI outputs influence final decisions
- whether managers recognise AI limitations
As generative AI becomes embedded in managerial work, these questions become increasingly important.
Future research therefore requires a shift from understanding technology acceptance to understanding technology reliance.
Why trust calibration matters
Trust calibration refers to the alignment between actual system capability and user reliance.
A manager demonstrates effective trust calibration when reliance on AI corresponds appropriately to the quality of the AI output and the level of risk associated with the task.
Conversely, trust miscalibration occurs when reliance becomes disproportionate.
Two forms of miscalibration are particularly important:
Over-reliance
Managers place excessive confidence in AI outputs despite uncertainty regarding their accuracy or contextual relevance.
Under-reliance
Managers reject potentially useful AI-generated insights because of excessive scepticism or lack of confidence in the technology.
Both forms of miscalibration can reduce decision quality.
The objective should therefore not be maximum trust but appropriately calibrated trust.
A Trust Calibration Framework for SME Decision-Making
I propose that trust calibration in AI-enabled strategic decision-making is influenced by four interacting dimensions.
Dimension 1: Perceived AI Competence
Managers continuously form judgements regarding AI capability.
These judgements may be based on:
- previous experiences
- perceived accuracy
- output quality
- system transparency
- task performance
When perceived competence increases, managerial reliance often increases.
However, perceived competence does not always reflect actual competence.
Generative AI systems may produce highly persuasive outputs even when underlying reasoning is weak or incomplete.
This creates a potential gap between perceived and actual capability.
Dimension 2: Managerial Expertise
The second dimension concerns the manager’s own expertise.
Experienced managers often possess:
- industry knowledge
- contextual understanding
- strategic intuition
- pattern recognition capabilities
Such expertise allows them to critically evaluate AI outputs.
Less experienced managers may struggle to distinguish between genuinely valuable insights and superficially plausible recommendations.
Consequently, managerial expertise acts as a moderating factor in trust calibration.
Higher expertise may improve calibration by strengthening critical evaluation.
Dimension 3: Environmental Uncertainty
Strategic decisions rarely occur under conditions of complete certainty.
Managers frequently face:
- volatile markets
- technological disruption
- regulatory change
- shifting customer preferences
- competitive unpredictability
Under uncertainty, confidence in human judgement may decline.
This can increase reliance on AI-generated recommendations.
Paradoxically, uncertainty often increases both the attractiveness of AI and the difficulty of evaluating its outputs.
The result is a heightened risk of trust miscalibration.
Dimension 4: Verification Mechanisms
The final dimension concerns organisational safeguards.
Verification mechanisms include:
- evidence checking
- peer review
- scenario comparison
- managerial discussion
- challenge routines
These mechanisms help managers evaluate AI-generated outputs before acting upon them.
Where verification mechanisms are weak or absent, reliance on AI may become increasingly dependent on subjective impressions rather than systematic evaluation.
Verification therefore serves as an important organisational buffer against both over-reliance and under-reliance.
The proposed process model
The framework can be summarised as a dynamic process:
AI Output → Managerial Interpretation → Trust Assessment → Verification → Strategic Action
At each stage, managerial judgement remains central.
The process is not linear or automatic.
Managers continually reassess AI recommendations in light of new information, organisational constraints, and contextual realities.
This suggests that strategic decision-making remains fundamentally human even when supported by advanced AI systems.
Implications for SMEs
The framework is particularly relevant for SMEs.
Compared with large organisations, SMEs often possess:
- fewer specialist analysts
- less formal governance
- limited decision-support infrastructure
- concentrated decision authority
These characteristics increase both the potential value and the potential risks of generative AI.
On one hand, AI can compensate for resource limitations by providing rapid access to information and analytical support.
On the other hand, concentrated decision authority means that trust miscalibration may have direct strategic consequences.
For SMEs, understanding trust calibration may therefore be more important than understanding adoption alone.
Implications for Future Research
This framework generates several research questions.
Research Question 1
How does managerial expertise influence trust calibration during AI-supported strategic decision-making?
Research Question 2
How does environmental uncertainty affect managerial reliance on AI-generated recommendations?
Research Question 3
What organisational practices improve trust calibration in SMEs?
Research Question 4
How do verification mechanisms moderate the relationship between trust and decision quality?
Research Question 5
Under what conditions does AI-supported decision-making outperform traditional decision processes?
These questions provide promising directions for future empirical investigation.
Conclusion
The growing presence of generative AI in organisations requires a shift in research focus.
The central challenge is no longer simply understanding why firms adopt AI.
Instead, the critical question concerns how managers determine when and how to rely on AI-generated outputs during strategic decision-making.
This article proposed a trust calibration framework consisting of four key dimensions: perceived AI competence, managerial expertise, environmental uncertainty, and verification mechanisms.
Together, these dimensions help explain why identical technologies may produce very different organisational outcomes.
Ultimately, the future of AI-enabled decision-making will depend not only on technological capability but also on the ability of managers to calibrate trust appropriately under conditions of uncertainty.
For SMEs in particular, trust calibration may prove to be one of the most important managerial capabilities of the AI era.
Author Note
Saba Shahzadi is an emerging researcher whose work focuses on generative AI, managerial decision-making, organisational sensemaking, and SME digital transformation. Her research explores how managers interpret, trust, and integrate AI-generated outputs when making strategic decisions under uncertainty.
shahzadisaba4z-biesm 