When AI Meets the Customer: New Study Maps the Hidden Recipes Behind Firm Performance

Every company today is told to transform itself digitally, to sprinkle artificial intelligence across its services, and to reinvent its value proposition before a competitor does it first. Yet a striking number of these transformations stall, not because the technology fails, but because the people building the digital services and the people using them see two entirely different products. A new study published in Information Systems Frontiers by Shweta Kumari Choudhary and Arpan Kumar Kar of the Indian Institute of Technology Delhi, together with Prashant Kumar of the Management Development Institute Gurgaon, tackles this disconnect head-on. Their central question is deceptively simple: which combinations of digital interaction capabilities and AI governance capabilities are actually sufficient to drive high firm performance among digital service providers, and where do providers misread what their users truly want?

The theoretical backbone of the research is Dynamic Capability Theory, a framework in strategic management that describes how firms sense opportunities, seize them, and reconfigure their resources in fast-changing environments. Rather than treating individual factors such as website quality or governance structures as independent levers, the authors argue that these capabilities operate as bundles. A firm with superb AI governance but weak digital interaction may underperform, while a competitor with a modest governance framework but outstanding interaction design may thrive. This is the logic of equifinality: there is no single golden path to high performance, but several distinct recipes, each with its own required ingredients.

To uncover these recipes, the team turned to fuzzy-set Qualitative Comparative Analysis, or fsQCA, a set-theoretic method developed by sociologist Charles Ragin that has become increasingly popular in information systems research. Unlike regression, which estimates the average effect of each variable across a sample, fsQCA asks which configurations of conditions are sufficient for an outcome to occur. Each firm and each user response is assigned membership scores between zero and one for every condition, capturing degrees of truth rather than binary categories. The method then logically reduces the data into minimal combinations of conditions that consistently lead to high firm performance. This makes fsQCA particularly well suited to the combinatorial complexity of digital transformation, where the same outcome can emerge from very different capability mixes.

The empirical foundation is unusual and methodologically ambitious: dual-survey data collected from both digital service users and the service providers themselves. By surveying the two sides of the same service relationship separately, the researchers could compare perceptions directly rather than inferring user attitudes from provider reports. This design allowed them to measure what they call the perception divide, the gap between how managers and designers interpret their digital services and how the actual users experience them. Prior research has repeatedly shown that managers often overestimate how well they understand their customers, and this study brings that concern into the era of AI-mediated service delivery.

The findings deliver a pointed message to the technology industry. Service providers, the authors report, frequently misinterpret user expectations, and they do so in two specific and consequential areas. The first is perceived website socialness, the degree to which a website feels socially present, warm, and human rather than cold and transactional. Decades of research on technology acceptance have shown that people respond to interfaces as if they were social actors, and this study confirms that the social feel of a digital service is not a cosmetic flourish but a genuine driver of service effectiveness and, ultimately, firm performance. Providers who treat their websites as mere information channels systematically underestimate how much users value this social dimension.

The second blind spot is AI interoperability, the ability of AI systems to work smoothly with other systems, platforms, and data flows. Users, it turns out, notice when an intelligent service integrates seamlessly with the rest of their digital lives, and they reward that integration through engagement and satisfaction. Providers, however, tend to underestimate its significance, perhaps because interoperability is invisible in demos and marketing materials while being painfully obvious in daily use. In an ecosystem where customers juggle dozens of apps and services, an AI feature that cannot talk to anything else may feel less like intelligence and more like friction.

The configurational analysis adds further nuance by showing that AI governance capabilities enter the performance equation in different ways depending on the pathway. AI governance, in this context, encompasses the structures, policies, and practices through which firms oversee their AI systems, including questions of fairness, transparency, accountability, and explainability. Recent scholarship has framed governance as moving AI from a black box to a glass box, and related work has linked responsible AI governance to measurable corporate performance gains. What the fsQCA approach reveals is that governance is not universally mandatory in the same dose; in some configurations it works in concert with strong interaction capabilities, while in others different combinations compensate for its absence. The practical implication is that executives should stop asking whether governance matters and start asking which configuration of governance and interaction fits their strategic position.

The study also carries a warning encoded in its methodology. Because fsQCA identifies sufficient configurations rather than average effects, it exposes the fragility of one-size-fits-all digital transformation strategies. Two firms with identical budgets and similar AI investments can land on entirely different performance outcomes depending on how their capabilities are assembled. The authors frame this as a matter of aligning users’ interpretations of digital services with those of managers and service providers, a challenge they identify as one of the most significant obstacles facing firms under escalating pressure to reconfigure their value propositions through AI-driven digital transformation. The perception divide is not a soft problem of communication; it is a structural risk that can quietly nullify millions of dollars of technology spending.

For the broader research community, the paper demonstrates the value of configurational thinking in information systems, following a line of work that has applied set-theoretic methods to digital business strategy and organizational research. It also connects to a growing literature on affective computing, algorithmic fairness, and anthropomorphism, all of which grapple with how humans perceive and trust intelligent machines. By anchoring these perceptual constructs within Dynamic Capability Theory and testing them with a method designed for causal complexity, the study offers a template for future work that wants to move beyond isolated variables and toward realistic portraits of how firms actually succeed.

The immediate takeaway for practitioners is refreshingly concrete. Before commissioning another AI feature, firms should audit whether their digital touchpoints feel socially alive to users and whether their AI systems interoperate with the platforms their customers already inhabit. They should also recognize that governance is not bureaucratic overhead but a capability that, correctly configured alongside interaction strengths, forms part of a sufficient recipe for high performance. And above all, they should close the perception divide by measuring user experience directly rather than assuming that managerial intuition captures it. In the race to become AI-driven, the firms that win may not be those with the most sophisticated models, but those that understand, with unusual precision, what their users actually experience when the algorithm meets the interface.

Subject of Research: Configurational analysis of digital interaction and AI governance capabilities influencing firm performance in digital service providers

Article Title: A Configurational Analysis of Digital Interaction and AI Governance Capabilities for Firm Performance

Article References: Choudhary, S. K., Kumar, P., & Kar, A. K. (2026). A Configurational Analysis of Digital Interaction and AI Governance Capabilities for Firm Performance. Information Systems Frontiers. https://doi.org/10.1007/s10796-026-10810-7

Image Credits: AI Generated

DOI: 10.1007/s10796-026-10810-7

Keywords: digital transformation, AI governance, firm performance, fsQCA, dynamic capability theory, website socialness, AI interoperability, perception divide, digital service providers, information systems, capability configurations, user experience