One-Page Summary
Since ChatGPT launched in November 2022, hundreds of millions of people have learned to converse with language models, and the quality of that conversation depends on a human skill that barely existed before: decomposing an intention into a clear instruction, choosing words the model can act on, and repairing the request when the answer is wrong. This report asks what that skill, called prompt literacy, consists of, how it can be measured, what it is worth in jobs and productivity, and whether accounting, long called the language of business, can become the structured vocabulary of the AI era. It separates sourced evidence from interpretation, labels every survey wave, and treats the accounting thesis as a hypothesis to be tested rather than a conclusion.
Prompt design measurably changes what models produce. Chain-of-thought prompting raised GSM8K accuracy from 17.9 to 58.1 percent, and controlled experiments found gains from 40 percent faster writing to 15 percent higher customer support productivity. No validated test of an individual's prompt literacy exists yet.
The economic value is real but bounded. Job postings for prompt engineering rose from 1,400 to nearly 6,300 between 2023 and 2024, and 49 percent of occupations now use AI in at least a quarter of their tasks. Outside the AI capability frontier, however, consultants with AI access were 19 percentage points less likely to produce correct solutions.
The accounting-as-language thesis has strong institutional momentum and untested skill claims. XBRL's 2025 taxonomy modernisation is explicitly designed to accelerate AI integration, and IAASB adopted its Technology Position in September 2024, but no published study shows that accountants prompt or verify better than other professionals.
The report tests this question in three parts with different answers. Structured reporting infrastructure is already being rebuilt for AI: XBRL International's March 2025 modernised taxonomy specification is designed to directly facilitate and accelerate AI integration, and the UK's XBRL programmes have produced more than six million structured documents. AI assurance is being claimed as an accounting-led function: IAASB adopted its Technology Position in September 2024, IFAC and IESBA published generative AI guidance in June 2024, and Deloitte, EY, and PwC were reported preparing AI assurance services in June 2025. The third part, whether accounting vocabulary makes accountants better at prompting or verifying AI outputs, has no published evidence yet. The thesis remains open: institutions are moving, and the decisive experiments have not been run.
For workers, prompt literacy may be the difference between benefiting from AI access and being displaced by it, especially as the skill gap shows up in hiring. For firms, the evidence says the return on AI depends on training, verification, and knowing where the capability frontier ends. For scholars, the report maps eight research gaps, the most important being a validated prompt literacy assessment, evidence on whether productivity gains persist, and controlled experiments testing whether accounting's structured vocabulary improves human-AI communication. All data, charts, and the replication script are open access.