Balancing flexibility and regulatory risk: AI-assisted strategies for non-GAAP disclosures in competitive markets

In this paper, I explore how firms can strategically position their non-generally accepted accounting principles (GAAP) disclosures in a competitive market and leverage artificial intelligence (AI) capabilities to navigate regulatory constraints and reduce adjustment frictions. Thus, this paper addresses three critical research questions: (1) How can firms strategically position their non-GAAP disclosures in competitive markets while managing regulatory risks? (2) What role can AI play in reducing regulatory sensitivity and adjustment costs associated with disclosure flexibility? (3) How do firms optimize their disclosure flexibility when facing both competitive pressures and regulatory constraints? By developing a differential game theory model enhanced with AI capabilities, this study provides insights into how firms can achieve optimal disclosure strategies that balance market competitiveness with regulatory compliance.

The strategic importance of non-GAAP reporting has grown significantly as firms seek to provide investors with additional perspectives on their financial performance beyond traditional GAAP metrics (Heflin, Kolev, & Whipple, 2024; Bentley, Christensen, Gee, & Whipple, 2018). However, this flexibility in reporting introduces unique challenges, particularly in competitive markets where firms must carefully balance the desire to highlight favorable performance attributes against the risk of regulatory scrutiny (Black, Christensen, Ciesielski, & Whipple, 2021). Recent studies have documented an increasing trend in the use of non-GAAP metrics, with firms often responding to their peers' disclosure strategies by adjusting their own metrics to maintain investor appeal (Christensen, Pei, Pierce, & Tan, 2019; Gomez, Heflin, & Wang, 2023). This competitive dynamic has raised concerns among regulators and researchers about the potential for an “arms race” in non-GAAP reporting flexibility (Ciesielski & Henry, 2017; Hsu, Wang, & Whipple, 2022).

The theoretical foundation of this paper draws from several streams of literature. First, signaling theory suggests that firms use non-GAAP metrics to communicate specific financial strengths that influence investor behavior and competitive positioning (Chen, Gee, & Neilson, 2021; McClure & Zakolyukina, 2024). Second, agency theory helps explain the regulatory constraints firms face when adopting non-GAAP metrics that deviate significantly from GAAP (Black, Christensen, Taylor Joo, & Schmardebeck, 2017; Christensen et al., 2019). Third, recent research on AI in corporate decision-making demonstrates how AI-enabled analytics can enhance firms' information-processing capabilities in competitive and regulatory environments (Bose, Dey, & Bhattacharjee, 2023; Estep, Griffith, & MacKenzie, 2024).

This paper develops a novel differential game theory model that captures firms' dynamic disclosure strategies within a competitive market environment characterized by regulatory and competitive pressures. The model incorporates AI-enhanced capabilities through two channels, namely reduced regulatory sensitivity (enabling firms to sustain higher flexibility) and reduced adjustment costs (enabling faster convergence to optimal disclosure levels). This approach represents a significant advancement over existing models, which are largely static and fail to capture the continuous nature of strategic disclosure decisions (Hirshleifer & Teoh, 2003; Zakolyukina, 2018).

My mathematical model reveals several key insights. First, firms' optimal disclosure strategies emerge from a dynamic balance between competitive positioning and regulatory risk management. Second, AI capabilities affect disclosure strategy through distinct channels – regulatory navigation AI increases target flexibility levels, while operational efficiency AI accelerates adjustment speed without changing the target. Third, the model demonstrates how varying levels of regulatory sensitivity and competitive pressure influence equilibrium outcomes in non-GAAP reporting choices.

The empirical implications of my model are substantial. By providing a framework for understanding how firms optimize their non-GAAP disclosures in competitive environments, the model generates testable predictions about how firms adjust their reporting strategies based on adjustment frictions and regulatory constraints. These predictions can be validated using historical disclosure data and market reactions to non-GAAP reporting choices.

This paper makes several significant contributions to the literature. First, it provides the first comprehensive model of a non-GAAP disclosure strategy that incorporates both competitive market pressures and AI-enhanced regulatory navigation and operational efficiency. Second, it extends differential game theory to the domain of financial reporting and provides new insights into how firms can optimize their disclosure strategies in dynamic competitive environments. Third, it demonstrates how AI can be utilized to enhance strategic decision-making in financial reporting, extending traditional game-theoretic analysis to account for technological sophistication in regulatory navigation and operational efficiency.

The remainder of this paper is organized as follows. Section 2 presents the theoretical framework and reviews relevant literature. Section 3 develops the mathematical model and derives optimal strategies. Section 4 presents the analytical solution and closed-form equilibrium. Section 5 presents the comparative static analysis. Section 6 discusses the empirical implications. Section 7 acknowledges limitations, and Section 8 concludes with suggestions for future research.

Non-GAAP reporting allows firms to customize financial disclosures by excluding specific items to provide an “adjusted” view of performance (Heflin et al., 2024; Bentley et al., 2018). This flexibility is particularly valuable for firms aiming to highlight growth or profitability by excluding volatile expenses such as restructuring costs or stock-based compensation (Desai & Davern, 2024; Griffin & Lont, 2021). However, as several studies suggest, this latitude in reporting introduces risks, especially in competitive markets where firms may escalate adjustments to differentiate themselves from rivals (Ciesielski & Henry, 2017; Hsu et al., 2022; King, 2017; Isidro & Marques, 2021; Young, 2014).

Empirical evidence highlights the competitive dynamics surrounding non-GAAP disclosures and shows that firms often respond to their peers' disclosure strategies by adjusting their own metrics to maintain investor appeal (Black et al., 2021; Christensen et al., 2019; Gomez et al., 2023). Non-GAAP reporting thus functions as a competitive tool, which shapes market perceptions in industries where performance comparisons are crucial. This competitive interplay aligns with signaling theory, as firms use non-GAAP metrics to communicate specific financial strengths that influence investor behavior and competitive positioning (Chen et al., 2021; McClure & Zakolyukina, 2024).

Increased scrutiny from regulatory bodies, particularly the U.S. Securities and Exchange Commission (SEC), highlights the need for firms to balance flexibility with transparency in non-GAAP disclosures (Black, Christensen, Kiosse, & Steffen, 2017; Katz & McIntosh, 2019; Henry, Weitz, & Rosenthal, 2020). Studies show that firms adopting highly flexible metrics are more likely to face SEC comment letters and regulatory intervention (Deloitte & Touche LLP, 2024; Donelson, Kartapanis, & Koutney, 2024; Michael Coglianese CPA, 2024). Regulatory oversight aims to prevent excessive deviations from generally accepted accounting principles (GAAP) and ensure that non-GAAP metrics do not mislead investors or obscure financial realities (Black et al., 2021).

Agency theory offers insight into these dynamics and suggests that firms face agency costs when adopting non-GAAP metrics that deviate significantly from GAAP (Black, Christensen, Taylor Joo, & Schmardebeck, 2017; Christensen et al., 2019). Regulatory bodies impose scrutiny costs on firms to align their reporting practices with investor interests, thus limiting opportunistic disclosure (Choi, Young, & Zhang, 2019). Consequently, firms must carefully weigh the regulatory risk associated with flexibility against the competitive benefits of adjusted metrics. In this study, I model regulatory risk through a quadratic cost function that represents the escalating cost of regulatory scrutiny as firms increase their flexibility, a concept supported by prior literature (Crocker & Slemrod, 2005; Gerakos & Kovrijnykh, 2013).

Game theory provides a valuable framework for modeling disclosure decisions within a competitive market environment, where firms continuously adjust their reporting while facing competitive and regulatory pressures. Differential game theory, specifically, allows me to capture the dynamic adjustments in non-GAAP flexibility as firms manage both competitive pressures and regulatory constraints over time (Buckdahn, Cardaliaguet, & Quincampoix, 2011; Barron, 2024; Hou, Guan, & Yu, 2020). Prior studies have demonstrated the applicability of differential games in competitive disclosure contexts, where firms iteratively adjust their strategies to optimize market outcomes (Anastasopoulos & Asteriou, 2014; Flagmeier, Müller, & Sureth-Sloane, 2023).

The model in this paper represents firms as players in a differential game where each firm's objective is to maximize market valuation by optimizing the flexibility of its non-GAAP metrics while balancing regulatory risk. The inclusion of adjustment costs discourages abrupt changes in flexibility and ensures that firms adopt gradual, calculated shifts in their disclosures, a strategy commonly observed in empirical studies (Jennings & Marques, 2011; Desai & Davern, 2024; Laurion, 2020). The competitive penalty parameter introduced in the model further aligns with empirical findings on inter-firm rivalry in financial reporting and measures the cost each firm incurs when its competitor adopts a more flexible reporting approach (Balakrishnan & Cohen, 2011; Liang, 2024; Majeed, Yan, & Tauni, 2018).

AI-enhanced analytics has reshaped corporate decision-making by allowing firms to navigate competitive and regulatory environments and adjust strategies accordingly (Bose et al., 2023; Estep et al., 2024; Ng & Alarcon, 2020). Within the context of non-GAAP disclosures, AI can analyze historical disclosure trends, regulatory responses and enforcement patterns to help firms better comply with regulatory requirements and streamline disclosure adjustment processes (Susnjak, 2024). Prior studies emphasize the impact of AI on competitive strategy, particularly through reinforcement learning and time-series forecasting, which enhance firms' operational capabilities and strategic positioning (Borges, Laurindo, Spínola, Gonçalves, & Mattos, 2021; Krakowski, Luger, & Raisch, 2023; Kemp, 2024).

Thus, by incorporating AI capabilities, the model in this paper illustrates how technological sophistication affects firms' regulatory sensitivity and operational efficiency in managing disclosure practices. This capacity enhancement is consistent with bounded rationality theory, which acknowledges firms' limited information-processing abilities and the role of AI in overcoming these constraints (Davidson, 2024; Li, Wu, Cao, & Wang, 2021). The model's AI component aligns with studies that show how technological improvements can reduce operational and regulatory costs in competitive environments and enable firms to achieve equilibrium outcomes that reflect optimal strategies for both non-GAAP reporting and regulatory compliance (Packard & Clark, 2020).

While extensive literature exists on the effects of non-GAAP reporting on market perception and regulatory risk, few studies address dynamic disclosure strategies within competitive markets. Existing models are largely static and lack the continuous adjustment mechanism that reflects dynamic strategic decision-making (Hirshleifer & Teoh, 2003; Zakolyukina, 2018). Moreover, although recent studies explore the role of AI in competitive settings, they have yet to examine AI's impact on non-GAAP disclosure choices specifically.

The model developed in this study addresses these gaps by combining an optimal control framework with AI-enhanced capabilities and examining how technological sophistication affects firms' equilibrium disclosure choices through reduced regulatory sensitivity and adjustment costs. This dynamic model provides a comprehensive view of strategic decision-making in non-GAAP reporting, incorporating competitive penalties, regulatory risks and adjustment costs in a continuous framework.

In this model, I examine how individual firms optimize their non-GAAP disclosure flexibility in an industry environment characterized by regulatory scrutiny and adjustment costs. While firms operate in the same competitive industry, the model captures competition through its effect on the regulatory environment rather than through direct strategic interaction. Specifically, heightened industry competition intensifies regulatory attention, as captured by the parameter κ in the regulatory cost function. Each firm selects non-GAAP metrics to maximize its perceived market valuation while accounting for regulatory risks and adjustment costs. The model employs an optimal control framework where each firm's disclosure trajectory can be solved analytically, yielding closed-form solutions that facilitate comparative static analysis. This approach provides clear predictions about how changes in competitive intensity (via $κ$⁠), AI-enhanced capabilities (via $δ$ or $γ$⁠) and firm characteristics affect equilibrium disclosure choices.

In constructing this model, I use a differential game framework to capture the continuous, dynamic decisions that each firm makes in adjusting its non-GAAP reporting ^[1]. This setup allows me to evaluate how each firm's choice of non-GAAP metrics evolves over time in response to both external pressures (e.g. regulatory risk) and strategic competition.

The non-GAAP disclosure level for each firm, indicated by $xA(t)$ for Firm A and $xB(t)$ for Firm B at any given time $t$⁠, represents the level of flexibility in their non-GAAP reporting. Higher values of $x$ correspond to greater deviations from GAAP standards, which indicates a more aggressive approach to non-GAAP adjustments. Each firm's state variable captures its reporting style, where greater flexibility enables a firm to present more favorable (but potentially less transparent) performance metrics. However, this comes at the potential cost of higher regulatory scrutiny and competitive pressure.

The control variables represent the strategic adjustments that each firm makes to its non-GAAP disclosure flexibility. $uA(t)$ is assumed to be the rate at which Firm A adjusts its non-GAAP disclosure flexibility at time $t$⁠, while $uB(t)$ denotes the rate at which Firm B adjusts its non-GAAP disclosure flexibility at time $t$⁠. In practice, these adjustments reflect the firms' responses to market conditions, competitor disclosures and regulatory signals. A positive control value $u(t)>0$ indicates that a firm is increasing its flexibility (e.g. moving towards a more aggressive disclosure metric), while $u(t)<0$ suggests a reduction in flexibility, which signals a more conservative approach.

The differential equations governing the evolution of each firm's disclosure strategy are expressed as follows:

These equations indicate that each firm's disclosure flexibility at time $t$ (represented by $xA(t)$ and $xB(t)$⁠) changes over time based on their respective control variables $uA(t)$ and $uB(t)$⁠. These dynamics allow me to observe how each firm's reporting strategy evolves continuously in response to both internal objectives and external pressures.

Each firm seeks to maximize its market valuation by selecting non-GAAP metrics that highlight favorable performance attributes, while also accounting for competitive responses and regulatory risks. The objective function for each firm, therefore, is designed to balance these competing factors over a finite time horizon $T$⁠.

The model assumes a zero discount rate (⁠$r=0$⁠), implying that firms value profits uniformly over the finite planning horizon. This assumption simplifies the analysis and explains the absence of discount-rate terms in both the objective function and the associated Hamiltonian.

The payoff function for Firm A, denoted by $JA$​, is given by:

Similarly, the payoff function for Firm B, denoted by $JB$​, is:

These payoff functions are structured to capture the benefits and costs associated with each firm's non-GAAP disclosure strategy. First, $αAxA(t)$ and $αBxB(t)$ represent the benefits derived from increased flexibility in non-GAAP disclosures for each firm. A higher flexibility level can improve perceived profitability or growth, which potentially increases market valuation. The coefficients $αA$ and $αB$​ quantify the relative impact of disclosure flexibility on each firm's valuation.

Second, $βAxB(t)$ and $βBxA(t)$ capture the competitive penalties that arise when a rival firm employs high flexibility in its non-GAAP reporting. For example, if Firm B adopts a highly flexible disclosure, Firm A may face market pressure to adjust its own disclosures to maintain competitiveness. The parameters $βA$​ and $βB$​ indicate the extent to which each firm's market position is impacted by the competitor's non-GAAP strategy.

Third, $γAuA(t)$ and $γBuB(t)$ represent the cost of adjusting disclosure levels. However, these costs are setup as quadratic terms to penalize rapid or large adjustments, thereby encouraging each firm to adjust its disclosure levels gradually and avoid erratic shifts that could raise regulatory concerns or lead to excessive scrutiny (Gomez et al., 2023; Jo & Yang, 2020). Firms would incur additional costs from making abrupt or large changes in their non-GAAP metrics due to factors such as investor skepticism, operational disruptions or increased scrutiny. The parameters $γA$ and $γB$ measure the relative sensitivity of each firm to the cost of making such adjustments. By using quadratic terms, the adjustment costs increase faster with larger adjustments This approach also provides a natural penalty for large adjustments without needing hard constraints or additional differential equations.

Finally, $δAR(xA(t))$ and $δBR(xB(t))$ account for the regulatory risk associated with increased non-GAAP flexibility. More flexible disclosures may attract greater regulatory scrutiny, especially if they significantly deviate from GAAP standards. The parameters $δA$ and $δB$​ represent each firm's sensitivity to regulatory risk.

The function $R(x)$ is defined as follows:

It models the regulatory risk as a quadratic function of the disclosure flexibility level $x$⁠. This functional form reflects the increasing scrutiny that firms face as they adopt more aggressive (i.e. highly flexible) non-GAAP metrics. The constant $κ$ is a regulatory sensitivity factor, which measures the degree to which flexibility incurs regulatory costs. This quadratic form implies that small increases in flexibility result in marginal increases in risk, but as flexibility grows, the regulatory risk escalates more rapidly. It also effectively creates a “scaled” penalty effect that measures increasing costs of flexibility without inflating them excessively. This approach is more analytically tractable and aligns with standard practices in control theory.

The payoff functions for Firm A and Firm B are modeled as integrals in consideration of the continuous, time-dependent nature of each firm's strategy and the resulting payoff over a finite time horizon T. In other words, each firm's decision to adjust its non-GAAP disclosure level is not a one-time action but rather a dynamic process that evolves continuously over time. Since firms are adjusting their non-GAAP disclosures over time, the model needs to account for the cumulative effect of these adjustments at each instant within the time interval $[0,T].$ An integral allows for the aggregation of these time-dependent payoffs over the entire time horizon. This helps to measure the accumulation of benefits and costs from each firm's decisions at each point in time.

For example, the expression $αAxA(t)−βAxB(t)−γA2uA(t)2−δAR(xA(t))$ represents the instantaneous payoff for Firm A at any given time $t$⁠. This instantaneous payoff combines the benefits of increased flexibility, the costs from competitive penalty, adjustment costs and regulatory risks, which are all evaluated at time $t$⁠. The integral $∫0T$​ then sums up these instantaneous payoffs over the entire period $t=0 to t=T$ (i.e. calculates the area under the curve), which measures the total accumulated payoff over time. In a differential game framework, players make decisions continuously over time, and their payoffs depend not only on their own actions but also on the ongoing actions of competitors. Modeling the payoff as an integral over time allows me to incorporate this interdependent, time-sensitive nature of the game. It also provides a complete picture of how each firm's strategy impacts its cumulative payoff from start to finish.

To sum it up, the differential game model presented in this paper presents the essential trade-offs in non-GAAP disclosure strategy. Each firm seeks to maximize market valuation through flexible non-GAAP metrics, but doing so increases the risk of regulatory scrutiny and competitive pressure. The model's payoff functions account for these dynamics, which allows each firm to strategically balance these competing factors. The use of control variables $uA(t)$ and $uB(t)$ enables each firm to dynamically adjust its reporting strategy over time. The differential equations show how disclosure levels evolve in response to both the firm's internal objectives and external factors, mainly regulatory and market pressures. This continuous adjustment reflects real-world dynamics. The regulatory risk function $R(x)$ highlights the increasing costs of aggressive non-GAAP reporting. Firms face quadratic growth in risk with higher levels of flexibility, which indicates the diminishing returns from flexibility as scrutiny intensifies. This constraint serves as a check against excessive deviations from GAAP standards.

The incorporation of AI into this model indicates its capacity to improve the firms' operational efficiency in managing non-GAAP disclosures, rather than predicting competitor behavior. AI capabilities affect firm strategy through two primary channels. The first channel is through reduced regulatory sensitivity (⁠$δ$⁠) and the second channel is through reduced adjustment costs (⁠$γ$⁠).

AI systems can help firms better navigate the regulatory environment by monitoring SEC comment letter patterns, analyzing enforcement trends and identifying disclosure practices that minimize regulatory attention (Bose et al., 2023; Estep et al., 2024). This improved regulatory navigation is parameterized as a reduction in the regulatory sensitivity parameter $δi$ for firm i. Firms with more advanced AI capabilities effectively face a lower perceived regulatory cost for any given flexibility level.

AI can also reduce the costs associated with adjusting non-GAAP disclosures by automating the analysis of how metric changes affect investor perceptions, streamlining internal approval processes and providing real-time feedback on market reactions (Gupta, Modgil, Bhattacharyya, & Bose, 2022; Susnjak, 2024). This operational efficiency is parameterized as a reduction in the adjustment cost coefficient $γi$⁠. Formally, let $δiAI=δi(1−θiδ)$ and $γiAI=γi(1−θiγ)$⁠, where $θiδ,θiγ∈[0,1)$ represent firm i's AI capability in regulatory navigation and operational efficiency, respectively. Higher values indicate more advanced AI implementation. The effect of AI on optimal disclosure strategies can be analyzed through comparative statics on the closed-form solution, as developed in Section 5.

The optimal control problem for each firm admits a closed-form solution. For Firm A, the Hamiltonian is as follows ^[2]:

Applying the maximum principle, the first-order condition yields:

The costate equation is as follows:

Given the initial condition $xA(0)$⁠, the state equation $x˙A(t)=uA(t)$ and the transversality condition $λA(T)=0$⁠, the resulting closed-form solution is:

where,

The solution for Firm B is analogous, with subscripts changed accordingly.

The closed-form solution reveals several important properties. First, each firm's optimal disclosure trajectory converges toward a long-run target flexibility level of $αiδiκ$⁠. This target increases with the transparency benefit (⁠$αi$⁠) and decreases with regulatory sensitivity (⁠$δi$⁠) and industry-wide regulatory pressure (⁠$κ$⁠). Second, the speed of adjustment toward this target is governed by $ψi=δiκγi$⁠. Firms with higher adjustment costs (⁠$γi$⁠) adjust more slowly, while firms with higher regulatory sensitivity (⁠$δi$⁠) adjust more rapidly. Third, while each firm's optimization problem is solved independently, the competitive environment affects both firms through the common parameter $κ$⁠, which captures industry-wide regulatory scrutiny that intensifies with competitive disclosure practices.

The competitive penalty term (⁠$−βAxB$ in Firm A's payoff) captures the market pressure each firm experiences when competitors adopt flexible non-GAAP metrics. While this term affects each firm's total payoff, it does not influence the optimal strategy because neither firm can control the other's disclosure trajectory. Each firm therefore optimizes its own disclosure path by taking the competitor's trajectory as exogenous. This structure reflects a realistic feature of non-GAAP disclosure competition – firms respond to industry-wide regulatory attention (captured by $κ$⁠) rather than engaging in direct strategic interaction. The model thus captures competitive pressure through its effect on the regulatory environment rather than through game-theoretic best responses.

The closed-form solution enables rigorous analysis of how key parameters affect optimal disclosure strategies. In this section, the effects of competition (⁠$κ$⁠), AI capabilities (⁠$δ,γ$⁠) and firm characteristics on equilibrium outcomes are analyzed.

Increased competition intensifies regulatory scrutiny, captured by higher $κ$⁠. The effect on the long-run target flexibility level is as follows:

Thus, this result suggests that higher competitive intensity (κ) reduces firms' long-run target flexibility levels. As industry-wide competition for investor attention through aggressive non-GAAP metric increases, regulatory bodies intensify scrutiny, which induces all firms to adopt more conservative disclosure approaches.

The effect on adjustment speed is given by

This result suggests that higher competitive intensity accelerates convergence to the (lower) target flexibility level. Firms respond to heightened regulatory pressure by adjusting their disclosures more rapidly.

The parameter $κ$ captures the intensity of industry-wide competitive pressure, which in this framework operates primarily through its effect on regulatory scrutiny. When firms within an industry compete aggressively for investor attention by adopting increasingly flexible non-GAAP metrics, regulators respond by heightening enforcement activity across the sector. This dynamic is well-documented in the accounting literature – the SEC has historically intensified scrutiny of non-GAAP reporting following periods of perceived abuse, issuing updated guidance and increasing comment letter frequency industry wide (Black et al., 2021; Gomez et al., 2023). The model formalizes this regulatory response through the parameter $κ$⁠, which scales the quadratic cost function and thereby amplifies the penalty for flexibility as competitive intensity rises. The comparative static $∂(αiδiκ)∂κ<0$ reveals that all firms in a more competitive industry converge to lower equilibrium flexibility levels, which indicates the heightened regulatory cost of aggressive disclosure. Simultaneously, the positive effect of $κ$ on adjustment speed (⁠$ψi$⁠) indicates that firms in high-competition industries not only target more conservative disclosures but also move toward these targets more rapidly.

The intuition is straightforward – when regulatory penalties are severe, the cost of deviating from the optimal disclosure level, even temporarily, becomes prohibitive, compelling firms to adjust swiftly. This suggests that regulatory interventions targeting industry-wide practices, such as the SEC's 2010 guidance on non-GAAP measures, should induce both a level shift (lower average flexibility) and a speed effect (faster convergence to new norms) across affected firms.

Superior AI capabilities for regulatory navigation reduce effective $δ$⁠. The effects are as follows:

This result suggests that better AI regulatory navigation capabilities (lower $δi$⁠) increase a firm's long-run target flexibility level while slowing the speed of adjustment. AI-enabled firms can sustain higher flexibility because they face lower perceived regulatory costs, but they adjust more gradually because the penalty for being away from the target is less severe.

The regulatory sensitivity parameter $δi$ indicates the degree to which a firm internalizes the cost of regulatory scrutiny when selecting its non-GAAP disclosure strategy. Firms with advanced AI capabilities can effectively reduce this parameter by deploying machine learning algorithms to analyze patterns in SEC enforcement actions, parse the textual content of comment letters and identify disclosure practices that historically correlate with regulatory intervention (Bose et al., 2023; Estep et al., 2024). Such systems enable firms to calibrate their non-GAAP adjustments with greater precision, avoiding red flags while still capturing the valuation benefits of flexible reporting. The comparative static $∂(αiδiκ)∂δi<0$ confirms that a reduction in regulatory sensitivity elevates the firm's long-run target flexibility level. Intuitively, when the perceived cost of regulatory scrutiny diminishes, the firm optimally shifts toward more aggressive non-GAAP metrics, as the marginal benefit of flexibility now exceeds its marginal cost at higher disclosure levels. The accompanying result on adjustment speed, $∂ψi∂δi>0$⁠, reveals a nuanced trade-off – while AI-enabled firms target higher flexibility, they converge to this target more gradually. The economic logic is that lower regulatory sensitivity flattens the cost curve around the optimum, which reduces the urgency to correct deviations from the target flexibility level. This slower adjustment should not be interpreted as a weakness; rather, it suggests a more forgiving optimization landscape in which temporary departures from equilibrium are less costly.

Superior AI capabilities for operational efficiency reduce $γ$⁠. The effects are as follows:

This result suggests that better AI operational efficiency (lower $γi$⁠) does not affect the long-run target flexibility level but increases the speed of adjustment. AI enables firms to reach their optimal disclosure level more quickly without changing what that optimal level is.

The adjustment cost parameter $γi$ captures the friction’s firm encounters when modifying its non-GAAP disclosure practices. These frictions arise from multiple sources – internal governance processes requiring approval for changes to reported metrics, the need to educate investors and analysts about revised definitions, potential restatement of historical comparatives and the risk that abrupt changes signal opportunistic reporting (Jennings & Marques, 2011; Laurion, 2020). AI can mitigate these costs by automating the analysis of how proposed metric changes would affect investor perceptions, streamlining internal review workflows and providing real-time feedback on market reactions to disclosure modifications (Gupta et al., 2022). The comparative static $∂(αiδiκ)∂γi=0$ indicates that adjustment costs do not influence the firm's long-run target flexibility level.

This result follows directly from the structure of the equilibrium as the target $αiδiκ$ is determined solely by the trade-off between transparency benefits and regulatory costs, independent of how costly it is to reach that target. However, the negative comparative static $∂ψi∂γi<0$ demonstrates that lower adjustment costs accelerate convergence to this target. The intuition is that when changing disclosure practices is less costly, firms can more readily correct deviations from the optimal flexibility level, whether those deviations arise from historical legacy practices, recent market shocks or strategic experimentation.

The preceding analysis establishes that the two AI channels, regulatory navigation (captured by $δi$⁠) and operational efficiency (captured by $γi$⁠), exert distinct influences on disclosure strategy. A reduction in $δi$ raises the firm's long-run target flexibility level by diminishing the perceived regulatory cost of aggressive non-GAAP reporting, while a reduction in $γi$ accelerates the firm's convergence toward whatever target it pursues. Table 1 summarizes these comparative static results.

While the two AI channels have distinct primary effects, their interaction is not fully separable in the adjustment dynamics. Specifically, the adjustment speed is given by $ψ=δκγ$⁠, which depends jointly on both parameters. The cross partial derivative,

indicates that reductions in regulatory sensitivity (⁠$δ$⁠) have a larger impact on adjustment speed when adjustment costs (⁠$γ$⁠) are already low. The total effect on adjustment speed when both AI capabilities are improved is captured by the second-order Taylor expansion:

This result provides a formal support for a form of joint dynamics: firms that invest in both regulatory navigation and operational efficiency AI not only target higher flexibility levels but also, depending on the relative improvements in operational efficiency (lower $γ$⁠), can transition to those targets more effectively. Thus, while the channels are analytically distinct in their direct effects, their combined impact on transition dynamics generates nuanced interactions.

A firm that invests solely in regulatory navigation AI (lowering $δi$⁠) will identify opportunities for higher flexibility but may adjust toward this new optimum sluggishly, foregoing potential valuation benefits during the transition. Conversely, a firm that invests only in operational efficiency AI (lowering $γi$⁠) can adjust rapidly but lacks the insight to know whether a higher flexibility target is sustainable. The combination of both capabilities has the potential to yield a strategic advantage – the firm not only perceives a more favorable regulatory path but can also reposition its disclosures swiftly to exploit this advantage, provided the net speed gains from reduced adjustment cost (⁠$γ$⁠) are positive. This suggests interdependent benefits to investing in AI capabilities in the disclosure domain; partial adoption captures only a fraction of the potential benefit.

From an industry dynamics perspective, firms that achieve early leadership in both AI channels may establish disclosure practices that competitors struggle to replicate, particularly if the transition to higher flexibility requires both regulatory sophistication and operational agility.

While Sections 5.2 and 5.3 establish how individual parameters affect optimal disclosure strategies, this section focuses on cross-sectional variation across firms. Specifically, it examines how differences in transparency benefits (⁠$α$⁠) and AI capabilities (⁠$δ$⁠, $γ$⁠) jointly shape equilibrium disclosure levels and adjustment dynamics.

The parameter α captures firm-specific transparency benefits from non-GAAP disclosures, which indicates factors such as growth opportunities, business complexity or investor demand for adjusted metrics. From the closed-form solution, the steady state target flexibility is given by $x*=αδκ$⁠, which implies that α directly scales the equilibrium disclosure level. However, α does not affect the adjustment speed $ψ=δκγ$⁠, which is governed solely by regulatory sensitivity and adjustment costs. This separation implies that cross-firm variation in α generates heterogeneity in equilibrium disclosure levels but not in transition dynamics, which provides a clear empirical distinction between level effects (α, δ) and speed effects (δ, γ).

The model generates cross-sectional predictions for variation in disclosure behavior across firms. Consider two firms, A and B, operating within the same industry and thus facing identical competitive pressure (⁠$κ$⁠). Differences in equilibrium disclosure strategies arise from firm-specific characteristics, mainly transparency benefits (⁠$α$⁠) and AI capabilities affecting regulatory sensitivity (⁠$δ$⁠). From the closed-form solution, equilibrium flexibility is given by $x*=αδκ$⁠. Thus, a firm with superior AI capabilities (lower $δ$⁠) will, all else equal, select a higher level of disclosure flexibility as follows:

This result implies that AI-enabled firms will exhibit systematically higher non-GAAP flexibility in cross-sectional data. Importantly, this behavior reflects optimization rather than opportunism. Firms with superior AI capabilities face lower marginal regulatory costs and can therefore sustain higher flexibility levels in equilibrium.

An important implication of the model concerns the interpretation of AI's effect on regulatory outcomes. While improved regulatory navigation (lower δ) reduces the marginal cost of disclosure flexibility, it also induces firms to optimally select higher equilibrium flexibility levels. Substituting the equilibrium disclosure level $x*=αδκ$ into the regulatory cost function yields an equilibrium penalty of $α22δκ$⁠, which is decreasing in $δ$⁠. This implies that firms with superior AI capabilities may, in equilibrium, incur higher total regulatory penalties despite facing lower marginal regulatory sensitivity.

This distinction highlights the importance of separating equilibrium effects from marginal effects when interpreting empirical evidence. AI reduces the sensitivity of regulatory costs to incremental increases in disclosure flexibility, but it does not necessarily reduce total regulatory exposure.

The comparative statics derived in Section 5 provide clear predictions that can be examined using firm-level disclosure and regulatory data. Table 2 summarizes the relationship between each theoretical result, its empirical prediction, appropriate observable proxies and potential identification strategies. The table serves as a roadmap for translating the model's analytical results into testable hypotheses while distinguishing between equilibrium effects, adjustment dynamics and the strategic roles of AI in regulatory navigation and operational efficiency.

The model predicts that greater industry-wide regulatory pressure (⁠$κ$⁠) leads firms to adopt lower levels of non-GAAP disclosure flexibility while adjusting more rapidly toward these more conservative disclosure strategies. Empirically, this prediction can be examined by studying disclosure behavior before and after industry-specific regulatory actions, such as SEC guidance updates or major enforcement initiatives. Using SEC comment letter frequency or enforcement activity as proxies for regulatory pressure (⁠$κ$⁠), researchers can test whether industries experiencing heightened scrutiny exhibit both lower non-GAAP adjustments (measured as the difference between GAAP and non-GAAP earnings) and faster convergence toward more conservative reporting practices. Consistent with the model, industries subjected to more intense regulatory scrutiny should exhibit both larger reductions in disclosure flexibility and steeper adjustment paths following regulatory interventions.

The model predicts that firms with stronger AI capabilities for regulatory navigation (lower $δ$⁠) will adopt higher equilibrium levels of non-GAAP disclosure flexibility because AI reduces the perceived marginal cost of regulatory scrutiny. Consequently, firms adopting regulatory technology (RegTech) solutions should exhibit gradual increases in non-GAAP flexibility following implementation, with disclosure practices converging smoothly toward higher equilibrium levels over multiple reporting periods rather than through abrupt changes. Researchers can exploit variation in the timing of AI adoption across firms to examine whether post-adoption disclosure trajectories follow the gradual convergence pattern predicted by the model.

The model also predicts that AI moderates the relationship between disclosure flexibility and regulatory intervention. Specifically, firms with stronger AI capabilities face a smaller marginal increase in regulatory scrutiny for a given increase in disclosure flexibility. However, this should not be interpreted as implying uniformly lower regulatory intervention. Because AI-enabled firms optimally choose higher equilibrium flexibility levels, their total regulatory exposure is predicted to increase in equilibrium despite lower marginal regulatory sensitivity.

Empirically, identifying this moderating effect requires research designs that generate exogenous variation in disclosure choices so that researchers can separate equilibrium disclosure decisions from the marginal effect of AI on regulatory outcomes. Quasi-experimental settings, regulatory shocks or other exogenous changes affecting disclosure practices provide promising opportunities to test whether AI weakens the relationship between disclosure flexibility and regulatory intervention while simultaneously encouraging firms to maintain higher equilibrium levels of non-GAAP flexibility.

The model predicts that AI capabilities improving operational efficiency (lower $γ$⁠) affect the speed with which firms adjust their non-GAAP disclosure strategies rather than their long-run equilibrium flexibility. Specifically, firms adopting AI-driven disclosure automation tools should exhibit faster convergence toward their target disclosure levels following changes in the regulatory or competitive environment, even though their long-run average non-GAAP flexibility remains unchanged through this channel.

This prediction differs fundamentally from the regulatory-navigation channel. Whereas reductions in regulatory sensitivity (⁠$δ$⁠) influence both equilibrium disclosure flexibility and adjustment dynamics, reductions in adjustment costs (⁠$γ$⁠) affect only the transition path. Consequently, researchers should observe faster mean reversion in firms' non-GAAP flexibility following industry shocks, SEC guidance updates or other exogenous changes that require firms to adjust their disclosure practices. Proxies for lower adjustment costs may include AI-enabled disclosure automation systems, investments in reporting technologies or other indicators of streamlined disclosure processes.

The distinction between the δ and γ channels provides an opportunity to disentangle the two mechanisms through which AI influences disclosure strategy. Researchers can examine whether AI adoption is associated with higher equilibrium disclosure flexibility (the δ channel), faster adjustment toward equilibrium (the γ channel) or both. The model predicts that firms with comprehensive AI implementations supporting both regulatory navigation and operational efficiency should exhibit both higher equilibrium flexibility and faster convergence to their optimal disclosure strategies when the operational efficiency gains are sufficiently large.

The model predicts that firms adopting both regulatory navigation AI and operational efficiency AI should exhibit stronger equilibrium disclosure outcomes than firms investing in only one capability. Specifically, comprehensive AI implementations are expected to produce both higher equilibrium non-GAAP flexibility through the regulatory navigation channel and faster convergence toward that equilibrium through the operational efficiency channel if efficiency gains are sufficient. When this is the case, firms deploying both RegTech solutions for compliance monitoring and AI-enabled disclosure automation systems could exhibit both higher average disclosure flexibility and more rapid adjustment following changes in the competitive or regulatory environment.

The model further predicts systematic cross-sectional variation in disclosure behavior arising from differences in transparency benefits (α) and AI capabilities (δ and γ). Firms with greater transparency benefits or stronger AI capabilities are expected to maintain higher equilibrium flexibility than otherwise similar firms operating in the same industry. These differences reflect optimal strategic responses to firm-specific characteristics rather than opportunistic reporting behavior, which suggests that observed heterogeneity in non-GAAP disclosure practices should be interpreted within the context of firms' underlying economic incentives and technological capabilities.

Empirically, these predictions can be examined by using fully saturated models that jointly capture regulatory navigation AI, operational efficiency AI and their interaction effect. Because firms endogenously choose their equilibrium disclosure flexibility and adjustment speed, identification of these effects should ideally rely on settings that generate exogenous variation in disclosure choices, such as regulatory shocks or other quasi-experimental settings. Such designs would provide cleaner evidence on how AI jointly shapes equilibrium disclosure behavior and adjustment dynamics.

While the model generates clear empirical predictions, testing these predictions requires careful measurement and identification. The key theoretical parameters, namely regulatory sensitivity (δ), adjustment costs (γ), competitive intensity (κ) and transparency benefits (α), are not directly observable and must therefore be approximated using empirical proxies. For example, regulatory sensitivity may be proxied by firms' AI-enabled compliance capabilities or historical exposure to SEC scrutiny, adjustment costs by indicators of disclosure automation or reporting technologies, competitive intensity by industry-level measures of regulatory attention or competition and transparency benefits by firm characteristics such as growth opportunities, business complexity or investor demand for supplemental disclosures.

A second challenge is that firms endogenously choose their disclosure strategies based on their underlying economic environment and technological capabilities. Consequently, simple cross-sectional associations between AI adoption, non-GAAP flexibility and regulatory outcomes may not identify the causal mechanisms implied by the model. Research designs exploiting exogenous variation in AI adoption, regulatory environments or disclosure incentives, such as regulatory changes, staggered technology adoption or other quasi-experimental settings, are therefore likely to provide more credible tests of the model's predictions.

Finally, the model highlights the importance of distinguishing among three empirically distinct mechanisms. Changes in regulatory navigation AI (δ) affect both equilibrium disclosure flexibility and adjustment dynamics, changes in operational efficiency AI (γ) affect only the speed of adjustment and the joint adoption of both capabilities generates interactive effects. Empirical studies that separately identify these channels will provide a more complete understanding of how AI influences firms' strategic non-GAAP disclosure decisions.

The model developed in this paper, while robust in capturing strategic non-GAAP disclosure dynamics, presents several limitations that are inherent to its assumptions, structure and underlying data constraints.

First, the model uses a quadratic function to represent regulatory risk associated with non-GAAP flexibility, assuming that regulatory costs increase at an increasing rate with higher levels of flexibility. However, in reality, regulatory responses to non-GAAP disclosures are complex and influenced by various factors beyond flexibility alone, such as industry practices and the evolving regulatory environment. While the quadratic function provides a useful approximation, additional factors could be integrated in future research to create a more dynamic risk model. For example, a multi-stage regulatory model could adjust for differences in industry scrutiny and incorporate variables such as historical firm behavior or sector-specific guidelines. This would create a more nuanced regulatory cost function.

Second, the model assumes that the key parameters, including transparency benefits (α), regulatory sensitivity (δ), adjustment costs (γ) and competitive intensity (κ), remain constant over the finite planning horizon. This assumption overlooks potential changes in firm's priorities, market dynamics or investor sentiment that could affect these parameters over time. Although the model treats these parameters as constant over the finite planning horizon, the comparative static analysis demonstrates how changes in transparency benefits, regulatory sensitivity, adjustment costs and competitive intensity affect equilibrium disclosure strategies. Future research could extend the framework by allowing these parameters to evolve over time in response to changes in the regulatory environment, market conditions or investor preferences.

Third, the model assumes that AI capabilities translate directly into reduced regulatory sensitivity (δ) and adjustment costs (γ). In practice, the effectiveness of AI systems in helping with regulatory requirements depends on data quality, algorithm sophistication and the evolving nature of regulatory enforcement patterns. Firms may overestimate or underestimate the regulatory cost reductions achievable through AI adoption, which may lead to suboptimal disclosure strategies. Future research could incorporate uncertainty in AI effectiveness by modeling δ and γ as stochastic parameters or by including a learning component where firms update their perceived AI capabilities based on observed regulatory outcomes.

Fourth, the differential-game framework models competition between two firms, which limits the applicability of the model in industries with more than two significant competitors or in complex market structures where multiple players interact with varying degrees of influence. To generalize the model, future research could extend the framework to an n-player game, which can provide more intuitive competitive dynamics across a larger set of firms. While computationally more demanding, this approach would offer broader relevance, particularly in industries where market perception is shaped by multiple competitors engaging in non-GAAP disclosures.

Fifth, the model assumes continuous adjustment of non-GAAP flexibility levels, yet in practice, firms may face delays in implementing disclosure strategy changes due to internal governance processes, external regulatory review or operational constraints. Future work could explore a discrete-time version of the model, in which adjustments are made at specific intervals. This would better mirror real-world decision-making cycles, which allows for a phased approach to adjustments that reflects practical constraints in disclosure timing.

Sixth, I model the interaction between the two firms as an open-loop differential game, in which each firm determines its disclosure trajectory at $t=0$ and commits to this path over the finite horizon. Strategies are thus functions of time only and do not adjust to contemporaneous changes in the state vector (⁠$xA(t),xB(t)$⁠). Within this structure, AI affects the firm's own cost parameters (δ and γ) rather than facilitating real-time strategic responses to competitor behavior. I explored a closed-loop (feedback) formulation, in which strategies would depend on both time and the evolving state, solved via coupled Hamilton–Jacobi–Bellman (HJB) equations. However, introducing a closed-loop AI-enabled prediction channel creates non-separable functional dependencies across the two players' value functions that eliminate tractable closed-form solutions and disrupt the smooth adjustment dynamics central to my comparative statics. I also tested a hybrid receding-horizon approach, but this added complexity without materially altering the qualitative insights. For these reasons, I retain the open-loop structure for analytical tractability and interpret AI as a time-zero enhancement to regulatory navigation and operational decision-making. A fully closed-loop extension remains a promising avenue for future research.

The model captures competition through its effect on industry-wide regulatory pressure (κ) rather than through direct strategic interaction between firms. While this approach yields analytical tractability and closed-form solutions, it abstracts from potential strategic considerations such as sequential disclosure decisions or explicit coordination. Future research could explore models with coupled state dynamics where one firm's disclosure directly affects the other's optimization problem, though such extensions would likely sacrifice analytical tractability.

Finally, the model does not fully address ethical considerations and potential reputational risks associated with AI-driven competitive disclosure strategies. Over time, stakeholders may perceive aggressive AI-enhanced non-GAAP disclosures as manipulative, which potentially harm a firm's credibility. Future models could work towards accounting for ethical risks.

This paper set out to explore how firms can strategically optimize their non-GAAP disclosure choices in competitive markets while balancing regulatory risks and examines how AI-enhanced capabilities for regulatory navigation and operational efficiency affect equilibrium disclosure strategies. Through the development of a differential game theory model incorporating AI-enhanced regulatory navigation and operational efficiency capabilities, the research demonstrates how firms can achieve optimal disclosure strategies that balance market competitiveness with regulatory compliance.

The key findings of this study are threefold. First, the differential game model reveals that firms' optimal non-GAAP disclosure strategies emerge from a dynamic interplay between competitive positioning and regulatory risk management. The incorporation of quadratic regulatory cost functions effectively captures the escalating scrutiny that firms face as they adopt increasingly flexible reporting approaches. Second, the research demonstrates that AI capabilities affect disclosure strategies through two distinct channels. Regulatory navigation AI (reducing δ) helps firms to sustain higher flexibility levels by lowering perceived regulatory costs, while operational efficiency AI (reducing γ) accelerates convergence to target flexibility without changing the target itself. Firms investing in both capabilities achieve combined benefits and reach higher flexibility levels more rapidly than competitors with narrower AI adoption, provided that operational efficiency gains are sufficient. Third, the comparative statics analysis reveals that equilibrium disclosure strategies are determined by transparency benefits (α), regulatory sensitivity (δ), adjustment costs (γ) and industry-wide competitive intensity (κ). While competitive penalties (β) affect total payoffs, they do not influence optimal disclosure trajectories because firms cannot control competitors' choices. Higher regulatory sensitivity and adjustment costs generally lead firms to adopt more conservative disclosure strategies, while competitive pressures generally reduce equilibrium disclosure flexibility by increasing the expected regulatory cost of aggressive non-GAAP reporting.

These findings have significant implications for both corporate practice and regulatory policy. For practitioners, the model provides a framework for evaluating the trade-offs between the competitive advantage and regulatory risk in non-GAAP reporting decisions. The research suggests that firms can achieve optimal outcomes by carefully balancing these competing factors while utilizing AI capabilities to enhance regulatory navigation and operational efficiency. For regulators, the findings highlight how competitive pressures can influence firms' reporting choices, which suggests the need for regulatory frameworks that acknowledge the strategic nature of non-GAAP disclosures while maintaining appropriate oversight.

The model's insights extend beyond non-GAAP reporting to broader questions of corporate disclosure strategy in competitive markets. The framework developed here could be adapted to analyze other forms of voluntary disclosure where firms must balance competitive positioning with regulatory compliance. Moreover, the integration of AI capabilities into strategic decision-making models represents a novel approach that could be applied to other areas of corporate strategy and financial reporting.

While this research provides valuable insights, several limitations should be acknowledged. The model assumes that AI capabilities translate directly into parameter reductions (lower δ and γ), which may not fully reflect the complexity of real-world AI implementation and its effectiveness in regulatory navigation. Additionally, the two-player framework, while analytically tractable, may not capture the full complexity of markets with multiple competitors. The static nature of certain model parameters also represents a simplification of dynamic market conditions.

Future research could extend this work in several directions. First, researchers could explore how the model's predictions align with empirical patterns in non-GAAP reporting, particularly in industries with varying levels of competition and regulatory scrutiny. Second, the framework could be expanded to incorporate multiple competitors and more complex competitive dynamics. Third, future studies might investigate how different AI architectures and implementation approaches affect strategic outcomes in non-GAAP reporting. Finally, researchers could examine how the model's insights apply to other forms of voluntary disclosure and strategic communication. As markets become increasingly competitive and AI capabilities continue to evolve, the framework developed here provides a foundation for understanding how firms can optimize their disclosure strategies while managing regulatory constraints.

During the preparation of this work, the author used Anthropic's Claude AI in order to check for spelling and grammar errors in the write-up and make corrections. After using this tool/service, the author reviewed and edited the content as needed, and takes full responsibility for the content of the publication.