SPECIAL FEATURE | AI leaders build proprietary intelligence, not pilot portfolios — Bain & Co.
Bain’s report, Proprietary Intelligence: How to Win with AI, said 82% of CEOs surveyed reported that their AI transformation was realizing only some or none of its intended results. Only 18% said they were reaching most or all of their AI ambitions.
Employees working together in collaboration and teamwork. Source: Stock photo provided by Kyndryl
Most companies are still treating artificial intelligence as a collection of pilots and productivity experiments rather than as a business transformation, creating a widening gap between organizations that are building long-term AI capabilities and those that are not, according to Bain & Company.
Bain’s report, Proprietary Intelligence: How to Win with AI, said 82% of CEOs surveyed reported that their AI transformation was realizing only some or none of its intended results. Only 18% said they were reaching most or all of their AI ambitions.
The figures came from the Bain CEO Survey 2026, which covered 100 CEOs. In the survey, 23% said results from their AI transformation or experimentation were below expectations, while 59% said some results were being realized. Another 15% said most results were being realized and 3% said results were above and beyond their original ambition.
Bain said the problem is not primarily the pace of AI development or technological limitations. Instead, companies are struggling with organizational capabilities, focus and the foundations needed to deploy AI at scale.
The most commonly cited barrier was a lack of in-house expertise and tools, identified by 43% of CEOs. Another 41% said their organizations remained focused on local pilots rather than broad transformation, while 39% said company data and platforms were not ready for AI adoption at scale.
Other concerns included unproven returns on AI investments, cited by 36%; navigating risks and legal concerns, 34%; resistance to adopting AI at the required speed, 32%; and uncertainty over how AI would affect industry and company strategy, 31%.
Only 18% cited a failure to dedicate their best leaders to AI implementation, while 15% said their organizations were not modeling the behaviors needed to scale AI’s impact. Six percent cited a lack of board or leadership support for large-scale AI adoption.
From pilots to proprietary intelligence
Bain said companies that are pulling ahead are not necessarily moving faster along the same path as their competitors. Instead, they are building what the report calls “proprietary intelligence” — a combination of unique data, encoded workflows and learning systems that can compound over time.
The report identifies three elements behind this approach: proprietary data consisting of a company’s accumulated record of customers, operations and outcomes; encoded workflows that capture institutional knowledge and enable agents to act on it at scale; and a learning architecture that creates feedback loops between people and AI systems.
The resulting cycle is intended to strengthen over time. Proprietary data improves AI agents, agents help employees, employees redesign work and encode new workflows, and those workflows generate additional data that can improve the system.
Bain cited financial operations platform Ramp as an example. After reaching 99% adoption of AI tools across the company, Ramp found that employees were beginning to plateau because its tools lacked shared infrastructure for connecting applications, propagating workflows and carrying context between sessions.
The company subsequently built Glass, an internal AI productivity layer designed to allow an employee’s breakthrough to become a broader organizational capability while maintaining persistent memory across the company. Bain said Ramp viewed internal productivity as a competitive moat rather than relying entirely on external vendors.
Seven decisions for AI transformation
Bain identified seven decisions that it said distinguish companies developing proprietary intelligence from those generating AI activity without achieving comparable transformation.
The first is posture. Bain said CEOs should commit capital to a multiyear strategic position rather than subjecting every AI initiative to an in-year return-on-investment test.
The second is domain focus. Instead of running dozens of pilots across different functions, companies should concentrate resources on three to five areas where AI can materially change the economics of the business.
The third is data. Companies need to develop proprietary data and a semantic layer as foundations for competitive advantage rather than attempting to retrofit them after AI deployments encounter problems.
The fourth is technology architecture. Bain recommends an orchestration layer that is operated in-house and connected to proprietary data and workflows, rather than placing the entire architecture under a single vendor’s platform.
The fifth is the operating model. Companies need to redesign workflows and workforce structures alongside AI deployment instead of simply adding AI tools to existing processes.
The sixth is the learning system, which should allow each AI deployment to make subsequent deployments smarter, faster and cheaper through feedback loops, shared memory and evaluations.
The seventh is governance. Bain recommends a governance structure specifically focused on changing the business, operating alongside the governance mechanisms responsible for running the existing business.
Agentic AI changes the equation
Bain argued that the emergence of agentic AI makes these decisions more urgent because AI agents can perform tasks beyond generating or summarizing information.
According to the report, agents can plan multistep tasks, act on enterprise systems through application programming interfaces, maintain state across extended interactions and operate with delegated authority on behalf of a business.
That capability creates a different competitive dynamic from previous technology transitions, Bain said. Companies that were late to cloud computing, digitization or modern data platforms could generally close the gap later by investing more money or adopting better technology.
The report argues that AI advantages can begin compounding from Day 1 because data, workflows and organizational learning can reinforce one another.
Bain also pointed to Bradesco as an example of the importance of learning from early failures. The bank initially designed its first agentic deployment around a few large, complex agents. Beta testing showed that the approach was too slow, expensive and difficult to scale safely.
Rather than forcing the architecture into production, the bank rebuilt it. The reset took roughly five months, but Bain said the resulting platform became the foundation for subsequent deployments. Bradesco is now running customer-facing AI in several core banking moments serving 22 million customers.
CEO involvement remains critical
Bain said the companies making the most progress share three leadership behaviors: personal commitment, concentration and investment in organizational learning.
The report said CEOs need to be directly involved in the transformation rather than simply approving budgets or sponsoring an AI program. They also need to focus resources on a small number of consequential areas and invest in infrastructure such as semantic layers, enterprise orchestration, shared memory and evaluation systems.
Bain cited Denis Machuel, CEO of recruitment services company Adecco, as an example of this approach. Machuel personally led the company’s AI transformation, emphasizing training and involving recruiters in designing agentic recruitment processes. The company assigned high-volume tasks to AI agents while retaining human judgment for other work.
Adecco aims to have agentic AI power 50% of its revenue by the end of 2026, according to the report.
Bain’s conclusion is that AI transformation is increasingly a question of strategic choices rather than simply technology adoption. The report argues that companies need to determine whether their investments are creating capabilities that become more valuable with each deployment — or merely adding another set of pilots to an expanding portfolio.
Full disclosure: All news articles published on the TechSabado website are written by human journalists, unless otherwise specified. Final text editing is also performed by human editors, with artificial intelligence (AI) used only to assist with additional grammar and style guide corrections..
————————————————————————-
TEN YEARS OF TECHSABADO!
PLEASE LIKE our FACEBOOK PAGE and SUBSCRIBE to OUR YOUTUBE & SPOTIFY channels.
WATCH TECHSABADO ON OUR YOUTUBE CHANNEL:
WE ARE ALSO ON SPOTIFY
WATCH OUR OTHER YOUTUBE CHANNELS:
PLEASE LIKE our FACEBOOK PAGE and SUBSCRIBE to OUR YOUTUBE CHANNEL.
