Artificial Intelligence Capital and Business Innovation
Most attempts to measure how businesses use artificial intelligence count things: patents filed, software licensed, tools switched on. Nick Drydakis, an economist at Anglia Ruskin University and the University of Cambridge, argues that these counts miss the part that matters. A firm can own AI tools without knowing what to do with them. His study of small and medium-sized enterprises (SMEs) in England asks a more human question: does a business's AI Capital—the AI-related knowledge, skills, and capabilities of its people—go hand in hand with innovation?
What AI Capital measures
Drydakis builds AI Capital on human capital theory. It has three components. Knowledge is understanding what AI is, what it can do in customer management, finance, and marketing, and where its risks lie in privacy, security, and ethics. Skills are the hands-on use of AI tools in communication, sales, financial operations, workflows, and human resources. Capabilities are the organizational ability to build AI into long-term strategy, governance, and decision-making.
He draws the boundary deliberately. AI Capital begins where generic digital literacy ends: at the point where staff must interpret a model's predictions, understand its errors and uncertainty, manage its risks, and redesign work around it. Knowing how to use a spreadsheet does not count.
To measure it, he developed and validated a new instrument, the AI Capital of Business scale. A pool of eighty-three candidate items was reviewed by sixteen AI, IT, and business experts and cut to forty-five—fifteen for each component—each rated on a five-point scale. Innovation is measured on a separate twenty-five-item scale across five dimensions: product and service innovation, process innovation, technology adoption, market and customer engagement, and organizational culture and strategy.
The data
The study follows the same firms across two survey waves a year apart, in early 2024 and early 2025, drawn from twelve sectors ranging from manufacturing to hospitality. The first wave reached 287 SMEs; 252 of them (88 percent) answered again, giving 504 paired observations. Surveys were routed to the people closest to AI decisions: managers, IT staff, innovation leads, and R&D personnel.
What it finds
The association is strong and remarkably stable. Across pooled, random-effects, and fixed-effects regressions—with controls for revenue growth, size, and age—each additional point of AI Capital is associated with roughly a third of a point more innovation (coefficients of 0.315 to 0.322). In plainer terms, a firm one standard deviation higher in AI Capital scores about 15 points higher on the innovation scale, around 17 percent above the average firm. The fixed-effects estimate, which compares each firm only with its own earlier self, is the most telling: when a business's AI Capital rises, its innovation tends to rise with it.
The link holds across every dimension of innovation, and it is strongest where firms meet their markets:
| Innovation dimension | Coefficient |
|---|---|
| Product and service innovation | 0.066 |
| Market and customer engagement | 0.064 |
| Organizational culture and strategy | 0.063 |
| Technology adoption | 0.061 |
| Process innovation | 0.059 |
Of the three components, knowledge carries the largest association (0.700, against 0.613 for skills and 0.622 for capabilities). Drydakis reads this as knowledge being a foundation: a firm has to understand what AI can do before its skills and structures can turn that into something new. The relationship also holds for small and large firms, young and established ones, and slow- and fast-growing ones, though it is somewhat stronger for firms growing faster than 5 percent a year (0.408 versus 0.306).
How far to trust it
The study is careful, and it is candid about its limits. Both scales are self-reported, often by the same respondents, so firms that think well of their AI prowess may think well of their innovation too. The two scales correlate at 0.948, which is striking for measures meant to capture different things; diagnostic tests suggest common-method bias does not drive the results, but it cannot be ruled out. Average AI Capital barely moved between the two waves, so the within-firm estimates rest on modest changes over a single year. The sample is English SMEs only, and the design establishes association, not causation.
Why it matters
The study's practical message is that buying AI is the easy part. For most small firms, Drydakis suggests, the priority is not large investments in hardware or bespoke systems but upskilling the people they already have to use off-the-shelf tools well—and building simple governance, such as an internal AI champion or a small cross-functional team, around them. For policymakers, it implies that grants for technology alone are unlikely to be enough without training and advisory support alongside them.
It also echoes an older argument. Drydakis frames AI as a general-purpose technology, and his findings rhyme with Paul David's account of the dynamo and the computer: a transformative technology pays off only once the organizations around it have been rebuilt to use it. AI Capital is, in effect, an attempt to measure that rebuilding while it is still underway.
Source
Drydakis, Nick. “Artificial Intelligence Capital and Business Innovation.” Journal of Strategy & Innovation 37 (2026): 200570. https://doi.org/10.1016/j.jsinno.2026.200570.