Unpacking the AI Transformation: The Impact of AI Strategies on Firm Performance from the Dynamic Capabilities Perspective

 

As artificial intelligence becomes central to business strategy, firms must move beyond mere awareness to active implementation. Our research shows that specific AI strategies drive performance differently: AI Product Strategy boosts value in volatile markets, while AI Process Strategy enhances sales in stable environments, suggesting firms must align their AI strategy with market conditions.

Understanding AI Strategic Orientation

Drawing on the dynamic capabilities framework, we conceptualize AI strategic orientation as a capability that helps firms sense opportunities, seize them, and reconfigure internal resources for technological change. We categorize firms’ AI orientation into three dimensions:

• AI Awareness, reflecting the firm’s ability to sense technological opportunities.
• AI Product Orientation, focusing on functional improvements to products and offerings.
• AI Process Orientation, focusing on operational improvements through AI-driven processes.

Measuring AI Strategy Using LLMs

We develop a novel measure of firm AI strategic orientation using large language models (LLMs) to analyze Item 1 business descriptions in 10-K filings from 2010 to 2022. After identifying AI-related disclosures, we classify them into the three orientation types and merge these measures with Compustat financial data.

To capture the long-term impact of technological investment, we employ long-difference models that reflect the gradual realization of AI-related changes. We examine how seven-year changes in AI orientation correspond to changes in operating expenses, capital expenditure, sales, and firm value.

What We Find

AI strategic orientation is associated with higher costs, but also with meaningful gains in firm value.

A one-standard-deviation increase in AI orientation over seven years corresponds to:

• 4.6% higher firm value,
• 3% higher operating expenses, and
• 5.7% higher capital expenditure.

These results suggest that investors reward firms for articulating and executing AI strategies, even when such strategies involve substantial cost increases.

Different Strategies, Different Outcomes

Each component of AI strategic orientation contributes uniquely to firm outcomes:

• AI Awareness predicts long-term increases in operating expenses and firm value.
• AI Product Orientation is linked to short-term cost increases and long-term value creation.
• AI Process Orientation is associated with higher long-term costs and rising sales.

These patterns show that different AI strategies affect different aspects of firm outcomes.

When the Environment Matters

We also study how environmental dynamism shapes the effectiveness of AI strategies. The results reveal that the "right" AI strategy depends on the business environment.

• In Volatile Environments, AI Product Orientation is the superior strategy. It is positively associated with firm value, despite higher expenses. In rapidly changing markets, using AI to seize new functional opportunities is key.
• In Static Environments, AI Process Orientation performs best. In stable markets, using AI to reconfigure operations leads to better long-term sales and cost management.

This suggests that the value of AI strategy depends heavily on the external conditions firms face.

Implications for Managers

For practitioners, the message is clear: firms that articulate a coherent AI strategy stand to benefit in the long run. AI product and process orientations require different cost structures and investment timelines, and managers should anticipate these differences when budgeting and designing AI initiatives.

Overall, our findings highlight the strategic importance of AI orientation, showing how firms can leverage AI not only to adapt but also to thrive in evolving competitive landscapes.

Authors: Jaecheol Park (Nanyang Technological University), Myunghwan Lee (Chinese University of Hong Kong), J. Frank Li (University of British Columbia), Gene Moo Lee (University of British Columbia)

Presented at: INFORMS International Meeting, July 2025