Preprint
Review

This version is not peer-reviewed.

Artificial Intelligence and Engagement Pricing in Accounting Firms: A Literature Review and Research Agenda

Submitted:

13 October 2026

Posted:

14 October 2026

You are already at the latest version

Abstract
Accounting firms are adopting artificial intelligence (AI) quickly, but the way they plan and price engagements has changed much more slowly. This paper reviews research from four areas that rarely meet: the pricing of accounting and audit services, the productivity effects of AI on knowledge work, the use of AI in accounting and auditing, and activity-based costing and project control. The review finds that AI's effects on accounting work are large but uneven across tasks, that new review and oversight work appears alongside time savings, and that hours freed by AI are often absorbed by other work rather than appearing as lower recorded effort. Firm-level studies link AI investment to lower audit fees and to changes in staff mix, but almost no research examines how firms adjust the internal, hours-based plans on which fees are set. I identify six gaps: the measurement of activity-level AI effects, the cost of oversight, the fate of freed capacity, in-engagement monitoring, the sharing of AI gains with clients, and the position of small and mid-sized practices. I then propose a research agenda of seven questions with suggested methods and data. The paper aims to help researchers and practitioners study and manage AI-enabled engagements on the basis of evidence rather than general efficiency assumptions.
Keywords: 
;  ;  ;  ;  ;  ;  ;  

1. Introduction

Few professions are as exposed to generative AI as accounting. Studies of occupational exposure consistently place accountants and auditors among the groups whose tasks overlap most with what large language models can do (Eloundou et al., 2024; Felten et al., 2023), and early field evidence shows accountants already using AI for routine processing, analysis and drafting (Choi & Xie, 2025). Accounting firms, from global networks to small practices, are investing in AI tools on the expectation that they will reduce the time needed to deliver services.
That expectation raises a pricing question. Most accounting engagements, whether billed hourly, at a fixed fee or under a value-based arrangement, are priced from an internal plan that estimates the hours required by activity and team level and converts them into a cost and a target margin (Maister, 1993; Simunic, 1980). If AI changes the hours, the plan should change. In practice, plans are often rolled forward from prior years, and AI effects, if they are considered at all, enter as a single efficiency percentage. Whether this is adequate depends on how AI actually changes accounting work, and that is a question the literature has begun to answer only recently.
This paper brings together research from four streams that are usually studied separately: the pricing of accounting services, AI and the productivity of knowledge work, AI in accounting and auditing, and the costing and control methods used to plan and monitor professional work. The aim is to clarify what is known, what is not, and what research would most help firms price AI-enabled engagements well. The review is narrative and integrative rather than systematic. I searched Google Scholar, SSRN and the main accounting, management and economics journals for work on these topics, giving priority to peer-reviewed studies and to recent working papers from established research institutions. Table 1 summarizes the streams covered.
The rest of the paper is organized as follows. Section 2 to 5 review each stream. Section 6 brings the findings together and identifies gaps. Section 7 sets out a research agenda, Section 8 discusses implications for practice, and Section 9 concludes.

2. How Accounting Services Are Priced

2.1. Effort, Leverage and Risk

The classic account of professional service firm economics links profit to three things: the rates a firm can charge, the leverage of junior to senior staff, and the utilization of available hours (Maister, 1993). Fee research in auditing formalizes the same intuition. Simunic (1980) models audit fees as a function of the resources the auditor must use and the expected liability it bears, and decades of empirical work confirm that client size, complexity and risk explain most of the variation in fees, largely through their effect on effort (Hay et al., 2006). In this view, the fee is a price for expected hours plus a premium for risk.

2.2. Value Pricing and Its Limits

A long-running practitioner critique holds that hours are a poor measure of what clients value and that time-based billing rewards inefficiency (Baker, 2011). Value pricing proposes instead that fees be set according to the value of the service to the client and agreed in advance. The critique is especially relevant to AI, because a firm that bills by the hour loses revenue when AI makes it faster. However, even firms that adopt fixed or value-based fees need to know their expected cost to judge whether a fee is profitable. The internal plan therefore remains central, whatever price format the client sees.

2.3. The Internal Plan as the Missing Object

Much of the fee literature studies outcomes, such as fee levels and fee premiums, rather than the internal planning process that produces them. We know relatively little about how firms build engagement budgets, how detailed they are, how often they are revised during delivery, and how they respond to technology. This gap matters because AI enters the firm through its work processes, and its effect on price runs through the plan. A pricing literature that looks only at fees will see AI's effect late and in aggregate.

3. AI and the Productivity of Knowledge Work

3.1. Large but Uneven Gains

Experimental and field studies report substantial productivity effects of generative AI. In a controlled experiment, Noy and Zhang (2023) found that access to ChatGPT reduced the time professionals needed for writing tasks and improved quality. In a large field setting, Brynjolfsson et al. (2025) found that an AI assistant increased the productivity of customer-support agents, with the largest gains for less experienced workers. Dell'Acqua et al. (2023) studied consultants and found strong gains on tasks within AI's capabilities but lower accuracy on a task just beyond them; they describe this as a "jagged technological frontier." The practical lesson is that AI's effect depends on the task. An engagement made up of many different tasks will not experience a single, uniform efficiency gain.

3.2. Where Does the Saved Time Go?

A second line of evidence is equally important for pricing. Using large-scale Danish data, Humlum and Vestergaard (2025) find that, although workers report productivity benefits from AI chatbots, recorded hours and earnings had not changed measurably two years after ChatGPT's release, while employers reorganized tasks and new AI-related work, such as overseeing and integrating AI output, became widespread. In accounting, Choi and Xie (2025) find that AI use shifted roughly 8.5% of accountants' time from routine data entry toward higher-value work rather than reducing total time. Together, these findings suggest that time saved by AI is often absorbed elsewhere. For an accounting firm, this means that AI's efficiency may never appear as fewer recorded hours on an engagement, and therefore may never reach the pricing plan. I have elsewhere called this unmeasured, AI-released time hidden capacity (Cetin, 2026).

3.3. Macroeconomic and Task-Based Perspectives

Economic research offers a broader frame. General purpose technologies take time to raise measured productivity because firms must first invest in complementary processes and skills (Bresnahan & Trajtenberg, 1995; Brynjolfsson et al., 2021). Automation both displaces tasks and creates new ones (Acemoglu & Restrepo, 2019; Autor, 2015), and AI in particular lowers the cost of prediction, which raises the value of the judgment needed to act on predictions (Agrawal et al., 2018). Exposure studies (Eloundou et al., 2024; Felten et al., 2023; Frey & Osborne, 2017) indicate that accounting tasks are highly exposed. These perspectives imply that AI will not simply shrink accounting engagements; it will change their composition, moving effort from preparation to review, judgment and client interaction.

4. AI in Accounting and Auditing

4.1. From Automation to Analytics

Accounting research on technology predates generative AI. Studies of robotic process automation and data analytics show that automation shifts work from routine preparation toward exception handling, review and judgment (Kokina & Davenport, 2017; Moffitt et al., 2018). Cooper et al. (2019) document that automation in public accounting firms freed staff from repetitive tasks, with time redirected to higher-value activities, and that firms did not expect large headcount reductions. Research agendas published in the same period anticipated many of today's questions, including how AI would change audit procedures, staffing and judgment (Appelbaum et al., 2017; Brown-Liburd et al., 2015; Issa et al., 2016). Austin et al. (2021) show that the adoption of data analytics depends on interactions among auditors, client managers, regulators and technology, a reminder that efficiency gains are shaped by institutional context.

4.2. Machine Learning for Accounting Judgments

Machine learning has been shown to improve specific accounting judgments. Ding et al. (2020) find that machine learning produces more accurate insurance loss estimates than managers' estimates, and Bertomeu et al. (2021) show that machine learning can help detect misstatements. These studies show what AI can do on well-defined tasks, but they say little about the time and cost of doing them inside an engagement.

4.3. Firm-Level Evidence on Fees and Staffing

Two recent studies examine AI at the level of the audit firm. Fedyk et al. (2022) use data on AI-related human capital at large audit firms and find that AI investment is associated with fewer restatements, lower audit fees and, over time, fewer accounting employees. Law and Shen (2025), studying audit offices, find that hiring AI specialists was associated with an increase in auditor jobs, mainly at junior and middle levels, suggesting that AI changes the skills firms need rather than simply replacing staff. The two sets of results are not necessarily inconsistent: they differ in samples, periods and measures. Both show that AI already affects fees and staffing. Neither, however, observes the engagement-level plans through which those effects pass.

4.4. Generative AI and the Human in the Loop

Eulerich et al. (2024) find that ChatGPT performs well on many accounting and auditing exam questions, especially with improved prompting, and argue that it has significant disruptive potential for the profession. Choi and Xie (2025) provide some of the first field evidence on accountants' use of AI in daily work, reporting faster month-end closes and more granular general ledgers among firms using AI, alongside concerns about AI errors and a tendency to follow uncertain AI suggestions. Experimental studies show that professionals' reliance on AI is not straightforward: auditors may rely less on AI-generated evidence than on human-generated evidence in some settings (Commerford et al., 2022), and managers' responses to auditor AI depend on whether their own company also uses AI (Estep et al., 2024). For pricing, the implication is that AI introduces oversight and review work whose size is uncertain and must be planned.

5. Planning, Costing and Control of Professional Work

5.1. Activity-Level Costing

Activity-based costing was developed because broad cost averages distort decisions (Cooper & Kaplan, 1988). Time-driven activity-based costing simplified the method by asking for only two estimates per activity: the time required per unit of output and the volume of output (Kaplan & Anderson, 2004, 2007). This structure fits accounting engagements well, since most of their activities have natural output units, such as accounts reconciled or invoices processed. It also fits the evidence reviewed above: if AI's effect differs by task, it is most naturally measured as a change in time per unit for each activity.

5.2. Estimation Bias and Re-Forecasting

Effort estimates are known to be optimistic. The planning fallacy describes a persistent tendency to underestimate completion times (Buehler et al., 1994), and research on projects documents widespread optimism bias in cost and effort forecasts (Flyvbjerg, 2006). New technologies are likely to amplify this bias, because planners lack historical data on how they perform. Project control offers a remedy: earned value management compares progress with plan and re-forecasts the remaining effort through the estimate to complete (ETC) and the estimate at completion (EAC) (Anbari, 2003; Fleming & Koppelman, 2010; Project Management Institute, 2021). These tools are widely used in engineering and IT projects but are less commonly discussed in the context of recurring accounting engagements, and I found no studies that use them specifically to test whether planned AI efficiency is being realized.

6. Synthesis: What We Know and What We Do Not

Taken together, the literature supports four conclusions. First, AI can reduce the time required for many accounting tasks, sometimes substantially. Second, the effect varies widely by task and can be negative where AI output needs heavy review. Third, time saved is frequently absorbed by other work rather than appearing as lower recorded effort. Fourth, at the firm level, AI investment is already associated with changes in fees and staffing. What the literature does not yet explain is the link between these findings: how the effects of AI on individual tasks are translated, or fail to be translated, into engagement plans and prices. Six gaps stand out.
  • • Measurement of activity-level effects. Most productivity evidence comes from outside accounting or from aggregate measures. There is little systematic evidence on time savings for specific accounting activities.
  • • The cost of oversight. Studies note that AI output must be reviewed, but few measure the time this takes or how it should be priced.
  • • The fate of freed capacity. Evidence that saved time is absorbed elsewhere (Humlum & Vestergaard, 2025; Choi & Xie, 2025) has not been examined at the engagement level, where it determines whether pricing reflects AI at all.
  • • In-engagement monitoring. Research on fees looks at outcomes; research on project control rarely studies accounting. We know little about whether monitoring plans during delivery improves margin accuracy in AI-enabled engagements.
  • • Sharing AI gains with clients. Fedyk et al. (2022) find lower audit fees associated with AI investment, but how firms decide how much of a saving to pass on, and how clients respond, remains largely unstudied.
  • • Small and mid-sized practices. Most firm-level evidence concerns large audit firms. Smaller practices, which serve a large share of businesses, may have fewer resources to measure AI effects and less room for pricing errors.
One recent proposal addresses several of these gaps by combining activity-level pricing plans, with explicit parameters for AI efficiency, adoption and oversight, with monthly ETC reviews (Cetin, 2026). Such frameworks are a starting point, but they need empirical testing.

7. A Research Agenda

Table 2 sets out seven research questions that follow from the gaps above, with suggested methods and data. The questions are intended to be answerable with data that firms already collect, such as time records, engagement budgets and fee data, combined with surveys and field experiments.
Two methodological points apply across the agenda. First, because recorded hours may not reflect actual effort when AI is used, studies should combine time records with output measures (for example, transactions processed or accounts reconciled) wherever possible. Second, because AI tools and practices change rapidly, longitudinal designs that follow the same engagements over several cycles will be more informative than single snapshots.

8. Implications for Practice

Although the evidence base is still developing, the literature already suggests several practical steps for accounting firms. Firms should plan AI effects at the level of activities, not as a single percentage, because effects differ sharply by task. They should budget explicitly for the review of AI output, since oversight is real work. They should measure performance per unit of output, not only total hours, so that freed capacity becomes visible. And they should re-forecast engagements during delivery, for example monthly, so that over- or under-performance against AI assumptions is detected while there is still time to act. None of these steps requires specialized software; they require discipline in planning and in reviewing the numbers. For small and mid-sized practices in particular, simple spreadsheet-based tools may be enough to avoid the most costly pricing errors.

9. Conclusion

Research on AI in accounting has grown quickly, but it has mostly focused on what AI can do and on outcomes at the level of the firm. The engagement plan, where hours are estimated, costs are built and fees are set, has received little attention, even though it is where AI's effects must be captured if pricing is to reflect them. This review has shown that AI's effects are uneven, that they create new oversight work, and that freed time is easily absorbed and lost from view. These findings point to a clear need for research on activity-level measurement, oversight costs, the fate of freed capacity, in-engagement monitoring, the sharing of gains with clients and the situation of smaller practices. Work on these questions would help accounting firms of every size price AI-enabled services on the basis of evidence.

References

  1. Acemoglu, D.; Restrepo, P. Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives 2019, 33(2), 3–30. [Google Scholar] [CrossRef]
  2. Agrawal, A.; Gans, J.; Goldfarb, A. Prediction machines: The simple economics of artificial intelligence; Harvard Business Review Press, 2018. [Google Scholar]
  3. Anbari, F. T. Earned value project management method and extensions. Project Management Journal 2003, 34(4), 12–23. [Google Scholar] [CrossRef]
  4. Appelbaum, D.; Kogan, A.; Vasarhelyi, M. A. Big data and analytics in the modern audit engagement: Research needs. Auditing: A Journal of Practice & Theory 2017, 36(4), 1–27. [Google Scholar] [CrossRef]
  5. Austin, A. A.; Carpenter, T. D.; Christ, M. H.; Nielson, C. S. The data analytics journey: Interactions among auditors, managers, regulation, and technology. Contemporary Accounting Research 2021, 38(3), 1888–1924. [Google Scholar] [CrossRef]
  6. Autor, D. H. Why are there still so many jobs? The history and future of workplace automation. Journal of Economic Perspectives 2015, 29(3), 3–30. [Google Scholar] [CrossRef]
  7. Baker, R. J. Implementing value pricing: A radical business model for professional firms; John Wiley & Sons, 2011. [Google Scholar]
  8. Bertomeu, J.; Cheynel, E.; Floyd, E.; Pan, W. Using machine learning to detect misstatements. Review of Accounting Studies 2021, 26(2), 468–519. [Google Scholar] [CrossRef]
  9. Bresnahan, T. F.; Trajtenberg, M. General purpose technologies: "Engines of growth"? Journal of Econometrics 1995, 65(1), 83–108. [Google Scholar] [CrossRef]
  10. Brown-Liburd, H.; Issa, H.; Lombardi, D. Behavioral implications of big data's impact on audit judgment and decision making and future research directions. Accounting Horizons 2015, 29(2), 451–468. [Google Scholar] [CrossRef]
  11. Brynjolfsson, E.; Li, D.; Raymond, L. Generative AI at work. Quarterly Journal of Economics 2025, 140(2), 889–942. [Google Scholar] [CrossRef]
  12. Brynjolfsson, E.; Rock, D.; Syverson, C. The productivity J-curve: How intangibles complement general purpose technologies. American Economic Journal: Macroeconomics 2021, 13(1), 333–372. [Google Scholar] [CrossRef]
  13. Buehler, R.; Griffin, D.; Ross, M. Exploring the "planning fallacy": Why people underestimate their task completion times. Journal of Personality and Social Psychology 1994, 67(3), 366–381. [Google Scholar] [CrossRef]
  14. Cetin, A. Capturing hidden capacity: Activity-level pricing plans and monthly estimate-to-complete monitoring for AI-enabled accounting services; (SSRN Working Paper); 2026. [Google Scholar]
  15. Choi, J. H.; Xie, C. Human + AI in accounting: Early evidence from the field, (SSRN Working Paper No. 5240924). 2025.
  16. Commerford, B. P.; Dennis, S. A.; Joe, J. R.; Ulla, J. W. Man versus machine: Complex estimates and auditor reliance on artificial intelligence. Journal of Accounting Research 2022, 60(1), 171–201. [Google Scholar] [CrossRef]
  17. Cooper, L. A.; Holderness, D. K.; Sorensen, T. L.; Wood, D. A. Robotic process automation in public accounting. Accounting Horizons 2019, 33(4), 15–35. [Google Scholar] [CrossRef]
  18. Cooper, R.; Kaplan, R. S. Measure costs right: Make the right decisions. Harvard Business Review 1988, 66(5), 96–103. [Google Scholar]
  19. Dell'Acqua, F.; McFowland, E.; Mollick, E. R.; Lifshitz-Assaf, H.; Kellogg, K.; Rajendran, S.; Krayer, L.; Candelon, F.; Lakhani, K. R. Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality; (Harvard Business School Working Paper No. 24-013); 2023. [Google Scholar]
  20. Ding, K.; Lev, B.; Peng, X.; Sun, T.; Vasarhelyi, M. A. Machine learning improves accounting estimates: Evidence from insurance payments. Review of Accounting Studies 2020, 25(3), 1098–1134. [Google Scholar] [CrossRef]
  21. Eloundou, T.; Manning, S.; Mishkin, P.; Rock, D. GPTs are GPTs: Labor market impact potential of LLMs. Science 2024, 384(6702), 1306–1308. [Google Scholar] [CrossRef] [PubMed]
  22. Estep, C.; Griffith, E. E.; MacKenzie, N. L. How do financial executives respond to the use of artificial intelligence in financial reporting and auditing? Review of Accounting Studies 2024, 29(3), 2798–2831. [Google Scholar] [CrossRef]
  23. Eulerich, M.; Sanatizadeh, A.; Vakilzadeh, H.; Wood, D. A. Is it all hype? ChatGPT's performance and disruptive potential in the accounting and auditing industries. Review of Accounting Studies 2024, 29(3), 2318–2349. [Google Scholar] [CrossRef]
  24. Fedyk, A.; Hodson, J.; Khimich, N.; Fedyk, T. Is artificial intelligence improving the audit process? Review of Accounting Studies 2022, 27(3), 938–985. [Google Scholar] [CrossRef]
  25. Felten, E.; Raj, M.; Seamans, R. How will language modelers like ChatGPT affect occupations and industries? arXiv 2023, arXiv:2303.01157. [Google Scholar]
  26. Fleming, Q. W.; Koppelman, J. M. Earned value project management, 4th ed.; Project Management Institute, 2010. [Google Scholar]
  27. Flyvbjerg, B. From Nobel Prize to project management: Getting risks right. Project Management Journal 2006, 37(3), 5–15. [Google Scholar] [CrossRef]
  28. Frey, C. B.; Osborne, M. A. The future of employment: How susceptible are jobs to computerisation? Technological Forecasting and Social Change 2017, 114, 254–280. [Google Scholar] [CrossRef]
  29. Hay, D. C.; Knechel, W. R.; Wong, N. Audit fees: A meta-analysis of the effect of supply and demand attributes. Contemporary Accounting Research 2006, 23(1), 141–191. [Google Scholar] [CrossRef]
  30. Humlum, A.; Vestergaard, E. Still waters, rapid currents: Early labor market transformation under generative AI; (NBER Working Paper No. 33777); National Bureau of Economic Research, 2025. [Google Scholar]
  31. Issa, H.; Sun, T.; Vasarhelyi, M. A. Research ideas for artificial intelligence in auditing: The formalization of audit and workforce supplementation. Journal of Emerging Technologies in Accounting 2016, 13(2), 1–20. [Google Scholar] [CrossRef]
  32. Kaplan, R. S.; Anderson, S. R. Time-driven activity-based costing. Harvard Business Review 2004, 82(11), 131–138. [Google Scholar] [PubMed]
  33. Kaplan, R. S.; Anderson, S. R. Time-driven activity-based costing: A simpler and more powerful path to higher profits; Harvard Business School Press, 2007. [Google Scholar]
  34. Kokina, J.; Davenport, T. H. The emergence of artificial intelligence: How automation is changing auditing. Journal of Emerging Technologies in Accounting 2017, 14(1), 115–122. [Google Scholar] [CrossRef]
  35. Law, K. K. F.; Shen, M. How does artificial intelligence shape audit firms? Management Science 2025, 71(5), 3641–3666. [Google Scholar] [CrossRef]
  36. Maister, D. H. Managing the professional service firm; Free Press, 1993. [Google Scholar]
  37. Moffitt, K. C.; Rozario, A. M.; Vasarhelyi, M. A. Robotic process automation for auditing. Journal of Emerging Technologies in Accounting 2018, 15(1), 1–10. [Google Scholar] [CrossRef]
  38. Noy, S.; Zhang, W. Experimental evidence on the productivity effects of generative artificial intelligence. Science 2023, 381(6654), 187–192. [Google Scholar] [CrossRef] [PubMed]
  39. Project Management Institute. A guide to the project management body of knowledge (PMBOK guide), 7th ed.; Project Management Institute, 2021. [Google Scholar]
  40. Simunic, D. A. The pricing of audit services: Theory and evidence. Journal of Accounting Research 1980, 18(1), 161–190. [Google Scholar] [CrossRef]
Table 1. Research streams reviewed and their relevance to engagement pricing.
Table 1. Research streams reviewed and their relevance to engagement pricing.
Research stream Representative studies Main insight for engagement pricing
Pricing of accounting services Simunic (1980); Maister (1993); Hay et al. (2006); Baker (2011) Fees are anchored in expected effort, team leverage and risk; even fixed or value-based fees rely on an internal hours-based plan.
AI and knowledge-work productivity Noy & Zhang (2023); Dell'Acqua et al. (2023); Brynjolfsson et al. (2025); Humlum & Vestergaard (2025) Gains are large on some tasks, small or negative on others, and freed time is often absorbed by new tasks rather than showing up as fewer hours.
AI and exposure of occupations Frey & Osborne (2017); Felten et al. (2023); Eloundou et al. (2024) Accounting is among the occupations most exposed to AI, so pricing assumptions are likely to change faster than in many other services.
AI in accounting and auditing Kokina & Davenport (2017); Cooper et al. (2019); Fedyk et al. (2022); Eulerich et al. (2024); Law & Shen (2025); Choi & Xie (2025) AI shifts effort from preparation to review and judgment; firm-level AI investment is associated with lower audit fees and changing staff mix.
Human-AI interaction and oversight Commerford et al. (2022); Estep et al. (2024); Choi & Xie (2025) Reliance on AI output is uneven and requires review effort that must be planned and priced.
Planning, costing and project control Cooper & Kaplan (1988); Kaplan & Anderson (2004, 2007); Buehler et al. (1994); Flyvbjerg (2006); Fleming & Koppelman (2010) Activity-level costing and in-flight re-forecasting (estimate to complete) offer tools to measure uneven AI effects, but they have rarely been applied to AI-enabled engagements.
Table 2. A research agenda on AI and engagement pricing in accounting firms.
Table 2. A research agenda on AI and engagement pricing in accounting firms.
# Research question Possible methods and data
RQ1 How large are AI time savings for specific accounting activities (e.g., reconciliations, invoice processing, account analysis, financial statement drafting), and how much additional review do they require? Time-and-motion studies; analysis of activity-coded time records before and after AI adoption; field experiments within firms.
RQ2 How do accounting firms currently adjust engagement budgets and fees for AI, and how common are aggregate (single-percentage) adjustments? Surveys and interviews with partners, managers and pricing teams in firms of different sizes.
RQ3 What happens to the hours released by AI: are they redeployed, absorbed by unplanned work, or left unrecorded? Comparison of planned, recorded and estimated hours; case studies; linking time data with output data.
RQ4 Do activity-level plans and monthly estimate-to-complete reviews improve margin accuracy and capacity use compared with traditional budgeting? Quasi-experimental designs within firms; pre/post comparisons; simulation using historical engagement data.
RQ5 How are AI efficiency gains shared between firms and clients, and how do clients respond to transparent fee adjustments? Contract and fee data; client surveys; experiments on fee communication.
RQ6 How do small and mid-sized practices differ from large firms in their ability to measure and price AI effects? Surveys of small practices; studies using practice-management software data.
RQ7 How does the oversight effort required for AI output relate to quality outcomes such as errors, restatements or review findings? Archival quality measures; experiments on reviewer behavior; analysis of review notes.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.