GenAI competence in accounting students rises with usefulness, readiness, and self-efficacy
Generative AI is moving into accounting classrooms quickly, but acceptance of the tool is not the same as competence with it. A new study from five Vietnamese universities focuses on GenAI-enabled accounting information systems competence rather than simple adoption, asking whether students believe they can use GenAI to process accounting information, verify outputs, detect inconsistencies, and apply judgment in AIS-related tasks.
The distinction matters because accounting work depends on evidence, documentation, internal control awareness, and professional skepticism. A fluent AI response can look persuasive while still being wrong, incomplete, or poorly grounded. In that setting, perceived competence is not a soft educational metric; it is tied to how future accountants may handle automated text, analytic support, and decision prompts in real practice.
Survey data from 471 accounting and auditing students formed the basis of the analysis. The investigators used PLS-SEM, a common latent factor diagnostic, and a Gaussian copula endogeneity assessment to test whether the observed relationships held up beyond baseline significance testing. The model linked GenAI readiness, perceived ease of use, digital learning self-efficacy, facilitating conditions, perceived usefulness, and AI ethics awareness to both GAISC and career readiness.
One result stood out immediately: perceived usefulness was the strongest baseline antecedent of GAISC.
Perceived usefulness dominated the competence model
At baseline, all nine hypothesized relationships were statistically significant and positive. The model explained 53.1% of the variance in GAISC, 45.6% in AI ethics awareness, and 52.9% in career readiness. Among the predictors, perceived usefulness had the largest association with GAISC, with β = .427 and f2 = .244. In practical terms, students who saw GenAI as useful were more likely to report that they could handle accounting information tasks with it.
That pattern is consistent with the way accounting learning often works. Students tend to engage more deeply with tools that appear to save time, clarify procedures, or improve task performance. In accounting, usefulness is not just convenience; it is tied to whether the system helps users trace numbers, interpret outputs, and move from raw data to defensible conclusions. A tool that cannot improve those steps will not feel competent-relevant, no matter how sophisticated it sounds.
GenAI readiness and perceived ease of use also showed positive associations with GAISC. Digital learning self-efficacy and facilitating conditions followed the same direction in the baseline model. The structure is familiar from technology acceptance research, but the endpoint is more specific here: students are not merely deciding whether to adopt a tool. They are judging whether they can use it in an accounting context without losing control over verification and judgment.
Ethics awareness tracked closely with accounting-system competence
The link between GAISC and AI ethics awareness was especially strong, with β = .675 and f2 = .838 in the baseline model. That relationship suggests that students who feel more capable of using GenAI for accounting tasks also report greater awareness of ethical issues around AI use.
In an accounting setting, ethics is not an abstract add-on. Students must think about source reliability, inappropriate delegation, transparency, and whether an AI-generated answer can be used without validation. Competence and ethics therefore move together: the more students understand how to interrogate AI outputs in an AIS setting, the more likely they are to recognize the ethical stakes attached to those outputs.
Career readiness also rose with GAISC in the baseline analysis. The study extends beyond generic AI comfort and places competence inside a career-development frame. For accounting students, confidence in using GenAI may translate into a sense of preparedness for workplaces where AI-supported analysis, documentation, and reporting are likely to be routine.
Endogeneity checks softened two of the strongest-looking paths
The Gaussian copula adjustment changed part of the picture. Seven relationships remained robust, but two paths lost statistical significance after adjustment: facilitating conditions to GAISC, and AI ethics awareness to career readiness. Those two were still supported in the baseline model, but they did not survive the more conservative endogeneity test.
This is the sharpest caution in the paper. Supportive environments may still help students, but the adjusted analysis suggests that the FC-GAISC relationship should not be treated as settled. The same caution applies to the ethics-to-career-readiness path. Students who are more ethically aware may also feel more career-ready, yet the data do not support a robust independent association once endogeneity is considered.
By contrast, the main competence-related associations held up well. Perceived usefulness remained the most influential baseline antecedent, and GAISC continued to relate strongly to ethics awareness. The study therefore points toward a competence model in which perceived value drives self-assessed ability, and that ability is closely tied to ethical reasoning in AI-supported accounting work.
Accounting education now needs competence, not just access
The study positions GAISC as something more demanding than broad AI literacy. It is not enough for students to know that GenAI exists or to feel willing to use it. They need to evaluate outputs, recognize limits, preserve audit-style skepticism, and keep control of the accounting logic beneath the text generation.
That emphasis fits the broader shift in accounting education toward verification, reperformance, and critical review of AI-generated work. If educators focus only on adoption, they risk mistaking tool familiarity for professional capability. If they focus only on speed or convenience, they may miss the deeper question of whether students can detect an error before it becomes an accounting decision.
The evidence comes with clear limits. GAISC and career readiness were measured through self-report, so the findings do not demonstrate objectively verified competence or employability. The design was cross-sectional, which means the relationships are associative rather than causal. Even so, the data offer a useful map of how accounting students in Vietnam are connecting GenAI usefulness, readiness, ethics awareness, and career confidence in one competence framework.
The study leaves one practical question open: whether classroom designs that force students to verify, challenge, and correct AI-generated accounting output can produce stronger competence than exposure alone.