[Sung-hyun Cho, head of technology at Databricks Korea] Digital-native companies have grown based on data and software. It is easy to think they would also be the most advanced in embedding AI into real operations. But findings in a 2026 report published by Economist Enterprise with support from Databricks, Making AI Deliver, show a somewhat different picture. The survey covered more than 1,220 global executives across 8 industries, including 150 executives at digital-native companies. The results showed that while digital-native companies lead in willingness to adopt AI and the scope of its use, they did not necessarily have the highest operational maturity in stably embedding AI across their organisations.

AI expansion is the top priority, but operational maturity is separate

Some 18 percent of executives at digital-native companies selected “embedding AI at scale across core business processes” as the most important investment priority over the next 2 years. That was the highest share among the 8 industries, about double the overall industry average of 9.8 percent. The second-highest response rate, in the energy, oil and gas industry, was 12.6 percent.

The results show that for digital-native companies, AI is becoming a core factor that goes beyond a simple work-support tool and shapes products, customer experience, operating models and revenue structures. Cost cuts and regulatory responses are also important, but a more fundamental task for these companies is to build a foundation that allows AI to operate repeatedly and reliably within core business processes. The survey also showed that digital-native companies gave AI expansion a higher priority than any other industry.

In South Korea, this shift is already under way. Tmap Mobility is advancing personalised recommendations using large-scale mobility data while also creating an environment where non-data specialists can directly explore and analyse the data they need. It also built a granular access-control and data-ownership system, strengthening both the scope of data use and governance. This shows that expanding AI use requires not only developing individual models but also a shared environment that enables people to use trusted data safely.

A scaling gap between broad adoption and “operational embedding”

Digital-native companies were using AI more broadly than the industry average across all business areas surveyed. In the survey, “at scale” includes both deploying AI across multiple work workflows and stably embedding it into operating systems. Among these, the “Fully Embedded at Scale” category, described as the “operational embedding stage”, refers to a state in which more than 100 users actually use AI, operations run under service level agreements (SLAs), and performance and business impact are continuously monitored. Broad deployment and stable operational embedding therefore need to be assessed separately.

Under this benchmark, the only function where digital-native companies ranked first was research and development and product development. Their level of operational embedding in finance ranked seventh among the 8 industries, while operations and supply chain ranked sixth. By contrast, the telecoms industry had a lower share of respondents who picked AI embedding as the top investment priority, at 7.9 percent versus 18 percent for digital-native companies. But it showed higher operational embedding levels in 5 of 8 functions, including IT, legal and compliance, finance, sales and customer service, and operations and supply chain.

This is the “AI scaling gap”. The point is not that traditional industries have overtaken digital-native companies overall. Digital-native companies still lead in the breadth of AI use and their willingness to scale it. But future competitiveness depends not on how many AI projects are started, but on how reliably adopted AI is converted into an operating system.

Why this gap matters

Digital-native companies are not failing to realise the value of AI. About 92 percent of respondents said AI return on investment (ROI) exceeds their original plans. That is higher than the overall industry average of 84 percent. The problem is that high ROI does not necessarily mean organisation-wide operational maturity. Achieving results in some projects and stably embedding AI into core work across multiple organisations are different tasks.

Much of this difference is likely to appear in operating architecture. To embed AI across an organisation, governed data access, trusted data pipelines, observability, model evaluation, SLAs, cost management, security, data lineage and a continuous feedback system must work together. A shared foundation is also needed so multiple organisations can use it and continuously monitor it in production environments.

South Korean gaming company Krafton is also an example of strengthening such a foundation. Krafton integrated its data management environment and, based on Unity Catalog, put in place consistent access policies and a data governance system. It built an environment where multiple organisations can safely share and use large-scale data. This shows that to scale AI reliably, it is necessary to have not only individual models but also trusted data and a consistent governance system that multiple organisations and AI projects can use in common.

Without a shared foundation, engineering organisations find it difficult to focus on developing new products and improving customer experience. That is because they must spend significant time and cost managing data pipelines, coordinating governance systems scattered across organisations, and repeatedly building similar systems. In the end, resources that should go into innovation are tied up in duplicated build-outs and maintenance.

This survey does not directly prove the causes of the gap. But it is necessary to examine whether the maintenance of governance systems is lagging behind the speed at which AI projects are expanding, and whether there is a sufficient foundation for multiple organisations and projects to use data and systems in common and continuously monitor operating status. The question executives should ask is clear. Does our organisation have a shared operating foundation for AI, or are we only continuing to increase individual AI projects?

What is needed now is not “more AI” but “better operations”

The scaling gap does not stem from a lack of AI value or willingness to invest. Digital-native companies are already experiencing high ROI. The solution therefore is not only to push more AI pilots or increase machine-learning engineers. Current performance must be converted into a system that is repeatable, governed and operates reliably in real production environments.

The starting point is operating architecture. Data pipelines and governance, AI workloads, models, agents and applications must be organically connected within a single operating system. Security, data lineage, monitoring and performance measurement should also be provided as a shared foundation that multiple organisations can use, rather than functions each organisation builds separately every time.

Ultimately, the companies that close the scaling gap will not be those that pursue the most AI experiments, but those that convert AI into a repeatable operating foundation. Digital-native companies already place a higher priority on AI expansion than any other industry. Now they need to move beyond continually adding AI on top of existing business and instead ensure it takes root stably in everyday operating methods and core decision-making processes.