The Winning Formula Behind Faster AI Adoption at Companies

A small number of companies have figured out how to make artificial intelligence actually pay, and they are doing it the same way: by pairing their own internal AI use with structured outside help, either specialized training or an implementation partner, rather than trying to teach themselves, based on research by Caribbean Business.

The gap between those companies and everyone else is not closing as the technology gets easier. It seems to be compounding.

An MIT-affiliated study of enterprise AI published last year found that organizations blending internal AI specialists with external vendors or consultants succeeded at a 67% rate, while those building entirely in-house succeeded just 22% of the time.

That’s three times the odds, for the same technology, decided by whether a company brought in help. The study’s authors attribute the difference to what they call a learning gap, not a technology gap, or off-the-shelf tools that perform well in a demo and then fail to adapt to a company’s actual workflows and institutional knowledge unless someone does the sustained work of getting them there.

The price of skipping that work is now measurable. Boston Consulting Group’s 2026 research found AI leaders achieve three times the cost reduction, 1.6 times the EBIT margins and 2.7 times the return on invested capital of their peers, with revenue per employee growing four percentage points faster on an industry-adjusted basis.

PwC’s 2026 AI Performance Study found that 20% of companies now capture roughly three-quarters of all the economic value AI generates.

Research published this year in Revista Caribeña de Psicología in Puerto Rico found fear to be the dominant emotion among local workers facing AI integration in their organizations. Training addresses the how of adoption; enforceable guardrails address the whether anyone trusts it.

Into that environment, where the learning gap is widening globally, and where fear remains a barrier locally, arrives the latest accelerant aimed at Puerto Rico: an expanded AWS Entrena Puerto Rico, the free Spanish-language training platform Gov. Jenniffer González Colón announced this week, now built out to five learning routes with the University of Puerto Rico as academic partner.

The expanded curriculum introduces specialized tracks such as IA sin límites, Agentes de IA sin código, AI Practitioner, Fundamentos de Nube, and De Cero a Builder, offering more than 40 hours of structured content. The program is designed to serve a broad audience—students, professionals, entrepreneurs, small businesses, and educators—reflecting the island’s urgent need for digital upskilling as AI adoption accelerates across industries.

AWS executives framed the initiative as part of a broader effort to democratize access to AI knowledge. “Just as we democratized access to the cloud 20 years ago, today we want to do the same with artificial intelligence,” said Américo de Paula, AWS director for northern Latin America and the Caribbean.

The University of Puerto Rico (UPR) has joined as an academic partner, adding institutional reach across its 11 campuses and integrating AWS micro-credentials into its educational offerings.

UPR President Zayira Jordán Conde said the collaboration strengthens the university’s responsibility to expose students and faculty to the competencies reshaping the labor market.

The push for widespread AI literacy comes as new national safety and privacy standards emerge to govern how AI is deployed in schools. In a landmark agreement announced by the American Federation of Teachers (AFT), the United Federation of Teachers (UFT), and Microsoft, U.S. school districts will now be able to incorporate legally enforceable protections into their Microsoft contracts.

The agreement prohibits tech companies from using student data to train AI models, bans student tracking, requires human oversight for AI-driven decisions, and mandates plain-language transparency for parents and educators. AFT President Randi Weingarten described the framework as a response to the absence of federal or state rules governing AI in education: “We have forged a hard-fought, iron-clad privacy agreement with real teeth that protects students and families.”

These guardrails, along with platforms like AWS Entrena, signal a maturing ecosystem where both opportunity and protection develop together.

How to accelerate adoption

The MIT findings, published as “The GenAI Divide: State of AI in Business 2025” by the university’s NANDA initiative, came from 300 public AI deployments and interviews with more than 150 business leaders.

The successful deployments in the study shared three traits: tightly scoped initiatives rather than enterprise-wide transformations, a focus on a specific business domain instead of general-purpose tooling, and outside partners brought in early.

Expertise gaps, data-security concerns and limited organizational knowledge are what now prevent companies from scaling AI effectively, precisely the constraints an outside partner or a structured training program exists to remove.

Companies that concentrated spending on back-office automation, document processing, compliance, internal workflows, saw better returns than those that put budget into the more visible sales and marketing pilots, where more than half of corporate AI money went.

McKinsey researchers describe the resulting dynamic as a flywheel rather than a head start. Faster experimenters generate more proprietary data, which improves their models and decisions, which lets them experiment faster still.

The firm has measured the spread in digital and AI maturity between leaders and laggards widening by roughly 60% over a recent multi-year period. That is the compounding most executives have been watching from the outside, and it is why the learning curve matters more than the license.

Where Puerto Rico’s companies actually stand

So where does Puerto Rico stand against these benchmarks? Local data now tracks the same divide. V2A Consulting’s “State of AI in Puerto Rico 2025,” the second annual edition of its survey and the most current local benchmark available, found 86% of Puerto Rico organizations using AI, up from 84% a year earlier, across more than 100 organizations in 14 industries.

More telling than the adoption rate is that 94% say they plan to invest in AI, a shift the consultancy reads as the conversation moving from whether to adopt to how to scale, and more than 40% already run dedicated AI agents, mostly in content creation, customer interaction and administrative workflows.

The report is explicit that the obstacles are no longer technological. Expertise gaps, data-security concerns and limited organizational knowledge are what now prevent companies from scaling AI effectively, precisely the constraints an outside partner or a structured training program exists to remove.

“Puerto Rico is at a decisive moment. The data reflect broad adoption, but also a significant gap between organizations using AI as a tool and those developing it as a strategic capability. That difference will determine competitiveness in the coming years,” said Xavier Diví, a director at V2A.

His colleague Manuel Calderón, the firm’s founder, put the emphasis on the same distinction: beyond the 86%, what matters is that organizations are approaching AI with more focus, discipline and clarity about what real impact requires.

A separate IDC survey conducted in late 2025 and presented locally this spring corroborates this finding but flags a distinct priority: the survey put Puerto Rico’s enterprise AI adoption at 65%, using a different methodology, and found 97% of companies planning to increase AI investment over the following 12 months, with spending growing 14%. IDC identified data structure, whether a company’s information is organized well enough for AI to use, as the island’s single biggest inhibitor, distinct from Latin America, where security and user experience lead.

Its Puerto Rico analyst was clear about the implication: companies need technology partners with the expertise and portfolio to turn AI use cases into working systems.