Companies that explain how they are actually deploying artificial intelligence may be getting more business value from AI than organizations that simply increase spending or issue broad AI commitments. A new study conducted by Larridin and researchers at Carnegie Mellon University found that publicly traded U.S. companies with the most concrete AI disclosures in regulatory filings recorded an 8% revenue-growth advantage over companies providing the least detail.

The AI boom has produced no shortage of corporate announcements. Companies are investing in copilots, generative AI platforms, automation and increasingly autonomous systems. But a more difficult question remains: Which organizations are turning those investments into measurable business growth?

New research from Larridin, conducted with researchers at Carnegie Mellon University, suggests one potentially useful signal is hiding in plain sight: how specifically companies describe their AI deployments.

The study, Do AI Adoption Signals Predict Company Performance?, examined more than 500 publicly traded U.S. companies, excluding the five largest AI chipmakers. Researchers compared seven indicators of AI adoption with subsequent revenue growth, operating margins and stock returns.

The most notable result was an 8% revenue-growth advantage among companies whose 10-K filings contained the most specific AI disclosures compared with those providing the least detail.

The researchers call this characteristic “narrative concreteness.”

In practical terms, that means distinguishing between a company saying it is investing in artificial intelligence and one explaining which AI systems it has deployed, where they are being used and what measurable outcomes they have produced.

That distinction could become increasingly important as enterprises move from AI experimentation toward production deployment.

AI spending alone is a weak measure of progress

The findings challenge a common assumption in enterprise AI: that larger investments necessarily indicate greater AI maturity.

According to the study, all seven AI-related signals examined showed a relationship with revenue growth. But several broader adoption and composite measures became less predictive after researchers controlled for industry, company size and previous growth momentum.

Narrative concreteness remained informative after those controls.

That does not mean a detailed 10-K automatically causes revenue growth. The study identifies a correlation, not proof that better disclosure itself produces better financial performance.

The distinction matters.

Companies that can clearly explain their AI deployments may simply be organizations with stronger operational discipline. They may have better measurement systems, clearer AI strategies or more mature approaches to connecting technology investments with business outcomes.

Still, the finding gives executives and investors another way to think about AI maturity.

Rather than asking only How much is this company spending on AI?, stakeholders can ask: Where is AI being deployed, who is using it, what has changed and can the organization measure the result?

The enterprise AI maturity problem

The research arrives as companies across industries attempt to industrialize generative AI.

Microsoft, Google, Amazon, Salesforce and Adobe, among other major technology vendors, have embedded generative AI and AI agents into enterprise software. Businesses are consequently gaining access to AI capabilities across customer service, software development, sales, marketing, analytics and operations.

But access does not equal adoption.

An enterprise might have thousands of employees with access to an AI assistant while only a fraction use it regularly. Even active users may not have the skills to apply AI effectively to high-value workflows.

That makes AI adoption, workforce proficiency, business impact and organizational maturity increasingly important measures for CIOs and chief data and AI officers.

Larridin’s research attempts to quantify some of these dimensions alongside signals derived from SEC filings and hiring activity.

The seven measures include adoption, proficiency, impact and maturity scores, as well as narrative concreteness, investment intensity and an AI-hiring builder rate.

The approach reflects a broader shift in enterprise AI measurement: from counting licenses and infrastructure spending toward measuring actual organizational behavior.

AI looks more like a growth tool than a cost-cutting engine

One of the study’s more consequential findings concerns the debate over AI and employment.

The research did not directly measure layoffs, so it cannot establish that AI has no effect on headcount. But researchers found no corresponding relationship between AI adoption and operating-margin expansion.

That weakens the idea that enterprise AI is already functioning primarily as a large-scale labor-cost reduction mechanism.

Instead, the evidence points toward a different early-stage pattern.

Companies appear to be using AI to expand capabilities, improve customer experiences and pursue new revenue opportunities, rather than immediately converting AI adoption into broad cost savings.

Carnegie Mellon University professor Shixiang (Woody) Zhu, who led the research team, characterized the measurable impact as appearing more clearly in revenue growth than operating margins or stock performance.

That distinction is important for enterprise technology leaders.

AI projects are frequently evaluated against productivity and cost-reduction targets. But if the technology’s earliest measurable returns are appearing through revenue expansion, companies may need broader performance frameworks covering sales productivity, customer retention, product development and new business creation.

What enterprise leaders should take from the research

For CIOs and AI leaders, the study offers a relatively straightforward lesson: make AI deployments measurable and specific.

An enterprise AI program should be able to identify which business process is changing, which employees or customers are affected, what capability the AI system provides and which metric should move as a result.

That could mean measuring software-development throughput, customer-service resolution rates, sales conversion, content-production capacity or the time required to complete an operational process.

It also changes how AI programs should be communicated.

A vague statement about “accelerating digital transformation with AI” tells executives little. A description of a deployed system, its user base, workflow integration and measured business outcome provides considerably more evidence.

The study therefore points toward an emerging definition of AI maturity that goes beyond technology adoption.

The most advanced organizations may not necessarily be those spending the most on AI. They could be those that understand where AI works, know how employees use it, measure the consequences and continuously connect those results to business performance.

For an industry still struggling to separate AI experimentation from durable enterprise value, that may be a more useful benchmark than spending alone.

Market Landscape

Enterprise AI is entering a measurement phase.

The first wave centered on access to large language models, generative AI applications and cloud infrastructure. The next challenge is demonstrating that those technologies produce sustainable business outcomes.

The Larridin-Carnegie Mellon research highlights an emerging gap between AI adoption and AI value realization. Companies can deploy AI broadly without necessarily generating measurable margin improvements or investor returns.

That creates opportunities for AI governance, workforce analytics, observability and enterprise AI measurement platforms.

The major cloud and software ecosystems—including Microsoft Azure, Amazon Web Services, Google Cloud, Salesforce and Adobe—are increasingly embedding AI into existing enterprise workflows. As those capabilities become standard features, measuring actual utilization and business impact could become more important than simply purchasing AI-enabled software.

For enterprise buyers, the competitive question is consequently shifting from Which AI platform should we buy? to How will we prove that employees, customers and business processes are benefiting from it?

Top Insights

• Companies with concrete AI disclosures showed an 8% revenue-growth advantage, suggesting measurable deployments may signal stronger enterprise AI execution and business value.
• Carnegie Mellon and Larridin studied more than 500 public companies, comparing AI adoption, proficiency, hiring and disclosure signals against financial performance.
• AI adoption was associated with revenue growth but not margin expansion, challenging the assumption that enterprise AI is already primarily a labor-cost reduction strategy.
• Narrative concreteness remained informative after controls, making specific descriptions of deployed AI systems a potentially useful maturity and execution signal.
• Enterprise AI leaders should measure outcomes, connecting deployments to workforce behavior, customer benefits, operational changes and quantifiable revenue opportunities.

Power Tomorrow’s Intelligence — Build It with TechEdgeAI