Companies have consistently underinvested in unified data architecture, and talent data fragmentation has become a fundamental barrier to enterprise HR’s capacity to make strategic workforce decisions.
According to the 2025 MuleSoft Connectivity Benchmark Report, even while enterprises run an average of 897 applications, just 34% offer an integrated user experience across all channels.
According to a 2024 survey by the Association for Talent Development, firms spend an average of $1,283 per person annually on training; nevertheless, 92% of corporate learning programs are unable to link their expenses to quantifiable outcomes.
Editorial Director Matthew DeMello recently spoke with Raúl Monroig, People Organization Vice President for the Intercon Region at Bristol Myers Squibb, on the podcast “AI in Business.”
Monroig highlighted the necessity for HR directors to focus their skill-building efforts on the competencies that actually generate business value, as well as the role AI plays in combining disparate HR data into actionable talent intelligence.
We look at two important takeaways for HR directors in the analysis of their discussion that follows:
- Integrating disparate talent data ecosystems: Combining employee data from various HR technology stacks into reliable, validated data models that do away with the need for self-assessed capabilities and allow for accurate talent mobility and development decisions.
- Disciplined skill focus and scientific outcome measurement: Programs that build 15–20 skills at once should be replaced with concentrated investments in just two to five essential capabilities, as determined by whether or not they truly increase ROI and business success.
Bringing Together Disjointed Talent Data Systems
The data used to make talent decisions isn’t trustworthy enough to support enterprise HR teams, which is a fundamental issue that most AI solutions don’t address. The problem is not a lack of data, but rather the fragmented and incomplete nature of the information HR professionals use.
In other words, HR teams operate across vast technology stacks, such as Workday, dashboards, Eightfold, and dozens of specialized tools, but these systems exist in isolation. Monroig succinctly summarizes this dilemma: “We work with only part of the data that we should have available, but we don’t.”
The information is dispersed throughout. The majority of us don’t have a framework that enables us to put everything together and deal with it from a shared perspective. I believe it’s essential for how we begin working on developing individuals and creating teams.
Raul points out that disjointed systems are only one aspect of the quality issue. HR teams start with a weak foundation—self-assessment—when they try to map capabilities, create pipelines, or create growth plans. The same incentives that influence public professional profiles also affect employees who evaluate their own abilities: a propensity to emphasize strengths and downplay weaknesses.
According to Monroig, self-reported competency data is structurally unreliable since people typically believe they are more capable than their peers.
Any self-reported data-based skill-building program is essentially constructed on sand. The problem is not just methodological; it also stems from a more fundamental structural problem with the way businesses handle talent intelligence.
When companies try to integrate AI-driven mobility or skill-matching workflows, the trust imbalance gets worse. The flaws in their inputs are inherited by these systems. It is a strategic failure to invest in integrated insight rather than a technical annoyance to have fragmented data dispersed across platforms.
The problem can be solved because there are already substitute data sources available in:
- Performance indicators
- Peer evaluation
- Results of the project, and
- Signals of behavior
Ground truth regarding what humans can truly do can be found in these sites. Self-assessed abilities and a compartmentalized system of record data are still considered adequate by HR platforms.
Instead of using generic, off-the-shelf AI solutions, Monroig contends that providers should be chosen and set up specifically for the goal of unifying scattered data. By ingesting many data sources, machine learning systems are able to construct unified personnel profiles that reveal trends in siloed systems that people are unable to notice. Organizations can utilize AI as a consolidation layer, converting disparate signals into coherent talent intelligence, rather than waiting for flawless integration.
According to Monroig, HR can use this kind of data consolidation as a model for the larger company, using AI as a fundamental engine to make disparate systems cohesive and useful.
Measuring Scientific Outcomes And Focusing On Disciplined Skills
According to Monroig, HR directors frequently fail due to overburden rather than malice. Companies regularly implement extensive skill initiatives, such as:
- AI knowledge
- Development of Leadership
- Management of people
- Digital proficiency, and
- Programs for legacy compliance
Without a well-defined hierarchy of the competencies that genuinely provide corporate value. Raul observes that the outcome is foreseeable: “We end up trying to develop skill-building systems that, frankly, cost a fortune, and we don’t know if they even deliver.”
The larger problem is that when the underlying data is fragmented, HR is unable to assess impact. Attribution is impossible without a cohesive understanding of worker performance, advancement, and results.
HR cannot prioritize which capabilities are important if it cannot link talent investments to revenue growth, retention, customer satisfaction, or team success.
According to Monroig, because the data foundation cannot support disciplined decision-making, this measurement gap pushes HR toward comprehensiveness—building everything and optimizing nothing.
He contends that there are three repercussions when HR lacks a cohesive, reliable data architecture:
- No attribution: HR is unable to determine which abilities are associated with business results.
- No prioritization: Since all skills seem to be equally vital, funding is provided for them all.
- Lack of strategic focus: Although skill portfolios grow, their impact is still unknown.
According to Monroig’s framing, fragmented data makes concentrated skill planning difficult; architecture rather than ambition is the primary problem.
The result is more than just squandered money. Large-scale development systems are frequently constructed by HR organizations without ever verifying if the skills they target have an impact on business outcomes. Simple inquiries are not tested:
- Does this ability affect performance on the bottom line?
- Does it increase retention or revenue?
- Does it lead to better customer outcomes?
These issues are still unresolved in the majority of businesses, and Monroig points out that the pharmaceutical industry exacerbates this blind spot.
“We should pay close attention to the two, three, or even five skills that will help us grow our company in the years to come. In order to develop those talents at the appropriate level with the appropriate individuals, [we should be unrelenting and extremely rigorous. We want to develop leadership and people management skills,” he contends.
Monroig also highlights a more fundamental ambiguity: historically, HR has not been based on a proven scientific understanding of skill development or behavioral modification. For decades, development programs and reward systems have been implemented without thorough testing.
In the future, Monroig believes that three skills are crucial for businesses managing the use of AI:
- Curiosity
- Agility as well as
- A customer-focused approach
According to his definition, curiosity is the readiness to try things even in the absence of certain results. He talks about a Latin American employee who created a ChatGPT agent without any formal AI training, cutting the time needed for brand-launch preparation from three weeks to thirty minutes.
Agility, or the capacity to accept new tools as they become available, is reflected in this desire to test, iterate, and adapt. Additionally, the customer-service mindset guarantees that any AI solution is grounded in business requirements rather than innovation.