Written by Fabio Panella
Technology is usually where HR transformation becomes visible. It’s the platform people log into every morning, the dashboard leaders pull up before a meeting, the AI assistant summarizing workforce signals, the skills marketplace suggesting a next move, the analytics layer that’s supposed to turn scattered people data into something strategic.
In the metaphor we’ve been building across this series, technology is the leaves of the tree. Culture is the soil, it decides whether there’s enough trust and appetite for learning to let transformation take root at all. Structure is the growth system, channeling that cultural energy into roles, governance, skills, and the everyday routines that decisions run through. Technology is the visible part, the layer that produces energy. Like leaves, it takes in something external (data) and turns it into something usable: intelligence.
And that’s exactly where the paradox starts. Organizations have never owned more HR systems, more data, more dashboards. Yet ask most of them a few basic questions, such as, which skills are going obsolete, where critical talent is at risk, which teams are quietly losing productivity, which workforce investments are actually paying off; and the answers get vague fast.
The problem isn’t a lack of technology. It’s a lack of intelligence.
A 2024 study of Swiss organizations by Gerber et al. backs this up. Most companies, they found, are still leaning heavily on backward-looking, descriptive reporting. There’s a persistent gap between what leaders want from HR analytics and what actually gets delivered, and data quality and analytical skill shortages sit at the center of that gap (Gerber et al., 2024).
Which brings us back to a principle from the previous article: technology doesn’t fix weak culture or poor structure. It amplifies whatever is already there. Buying an integrated platform doesn’t make a fragmented organization integrated. Rolling out AI doesn’t make a low-trust culture transparent. Launching a skills marketplace doesn’t make a rigid job architecture adaptive. Leaves only produce energy when the tree underneath them is healthy enough to use it.
And leaves are seasonal. Tools get replaced every cycle: platforms swapped out, algorithms updated, vendors acquired, interfaces redesigned. An organization isn’t defined by which tools it happens to be using this year. It’s defined by whatever system decides whether those tools ever create real value.
From Systems of Record to Systems of Intelligence
For decades, HR technology existed to do one thing: store information. Employee records, payroll, absence history, performance scores, job titles, comp bands, learning completions, succession plans. These systems brought order. They cut down on manual work and made HR administration run more smoothly.
But efficient isn’t the same as intelligent. A system of record tells you what already happened. A system of intelligence tells you what’s happening now, why it’s happening, what’s likely to happen next, and what to do about it.
That distinction matters more than it used to, because the questions HR is being asked have changed.
It’s no longer “how many people do we have,” but “do we actually have the capabilities to execute our strategy.”
It’s no longer “what was turnover last quarter,” but “where are we going to lose critical skills before the business even notices.”
It’s no longer “how many people finished the training,” but “did that training move the needle on performance, mobility, or retention.”
That’s the shift from HR reporting to workforce intelligence, at least on paper. In practice, a lot of organizations are still stuck at stage one. Gerber et al. break HR analytics maturity into reporting, diagnosis, and forecasting, and their data shows reporting still dominates by a wide margin, with predictive and forward-looking analytics still relatively rare (Gerber et al., 2024).
This matters because descriptive analytics can create a false sense of momentum. A dashboard can look impressive while it’s still just describing the past. A people analytics team can produce genuinely good insight and still have no real say over decisions. A platform can be technically integrated while business, finance, operations, and HR remain just as disconnected as ever.
That’s the context for a statistic we opened this series with: fewer than 10% of organizations can reliably tie people data to business performance. That’s not really a data problem, it’s a systems problem. People data only becomes strategic once it’s connected to outcomes: hiring data to productivity, skills data to workforce planning, engagement data to retention, learning data to capability growth, mobility data to how fast the business can adapt.
of organizations can reliably tie people data to business performance.
A system of intelligence, then, isn't a fancier dashboard. It's a different operating logic, one that actually links people, skills, work, performance, and business outcomes. It's what turns HR from a function that reports on the workforce into one that helps shape its future capacity.
Skills Intelligence: The Backbone of Adaptive HR
If technology is the leaves, skills intelligence is the chlorophyll, the mechanism that actually converts raw workforce data into something useful.
We argued in the previous piece that the HR operating model of the future has to be dynamic and skills-driven. Static job descriptions and neatly plotted career paths made sense in a world where work barely changed year to year. That world is gone. Skills now shift faster than roles do, careers cut across old boundaries, and business priorities move faster than annual workforce planning can keep up with. A skills-based architecture gives some shape to all that volatility, but technology is what makes it visible at any real scale.
Research using LinkedIn profile data shows how far this has come. Dorn, Schoner, Seebacher, Simon, and Woessmann looked at self-reported skills from nearly 9 million U.S. college graduates, using data curated by Revelio Labs, and grouped raw skill listings into 48 clusters spanning general, occupation-specific, and managerial skills. Their finding: these multidimensional skill profiles explain labor-market patterns that traditional measures of education and experience simply miss (Dorn et al., 2025).
That's a big deal for HR. It means skills are becoming a structured, machine-readable layer of human capital - something that can be mapped, clustered, compared, refreshed, and tied directly to roles, projects, learning paths, and business priorities. Which is exactly what makes adaptive HR possible in the first place.
With decent skills intelligence, workforce planning stops being an annual headcount exercise and starts becoming something closer to a live capability system. Organizations can see where skills are concentrated, where the gaps are forming, where adjacent capabilities already exist, and where internal mobility could solve a problem faster than another external hire.
It also changes what a talent marketplace can be. Once skills are actually visible, people can be matched to projects, mentors, learning paths, and stretch assignments (not just job postings). Managers start seeing capability instead of job title. HR moves from managing positions to mobilizing potential.
This matters even more given what HR analytics is increasingly being asked to solve. Gerber et al. found that skills shortages and retention were among the top problems respondents wanted HR analytics to address (Gerber et al., 2024).
But skills intelligence comes with a warning label. This data is powerful, but it's far from neutral. Much of it is self-reported, inferred, out of date, unevenly maintained, or shaped by how confident, visible, or career-savvy someone happens to be. Dorn et al. are upfront about this, the risk that how and when people update their profiles skews the picture (Dorn et al., 2025).
For organizations, that means skills intelligence can't just be a technical build. It needs governance. Definitions have to be clear. Data has to be refreshed. People need to understand how their skills data actually gets used, and managers need to be trained not to flatten someone into a set of tags. Good validation blends self-declaration with manager input, learning records, work history, credentials, and actual demonstrated performance.
A skills system people trust becomes a mobility engine. One they don't trust becomes just another database people avoid, game, or quietly resent.
AI in HR: From Describing the Past to Deciding About the Future
AI raises the ambition of HR technology considerably. Traditional analytics helped HR describe and diagnose. AI is meant to predict, recommend, personalize, and in some cases, decide outright.
It can flag attrition risk, recommend learning paths, match people to internal opportunities, screen candidates, summarize performance signals, read workforce sentiment, forecast skill gaps, draft job descriptions, run planning scenarios, personalize onboarding, and speed up HR service delivery generally. In short, it pulls HR technology closer to actual decision-making.
That's the opportunity, and the risk, in the same breath. AI can push HR from reactive to proactive. But it inherits whatever trust deficit already exists in the system it's entering. If people don't trust how decisions get made today, AI showing up tomorrow doesn't automatically make those decisions feel more legitimate.
There's a useful, slightly counterintuitive finding here from Yu and Li's 2022 research on AI transparency and employee trust. Transparency does increase how effective people perceive AI to be, which can build trust. But it also increases discomfort, and that discomfort can erode trust just as easily. Transparency, in other words, is necessary. It just isn't automatically reassuring (Yu and Li, 2022).
That's directly relevant to HR. Employees might appreciate understanding how an algorithm recommends a career path or flags risk. But they can also get uneasy once they realize how much data is feeding that recommendation, or how reductive the underlying logic actually feels. Yu and Li tie this to what they call algorithmic reductionism: the sense that AI boils down complex human situations into flattened, decontextualized variables, missing nuance, intent, or personal circumstance that a person would have caught (Yu and Li, 2022).
That's really the core tension of AI in HR. People don't experience HR decisions as technical outputs, they experience them as judgments about their worth, their potential, their future.
A model sees a skills gap; an employee sees a blocked career path.
A model flags attrition risk; a manager sees someone going through a hard stretch.
A matching algorithm returns low fit; a candidate experiences exclusion with no way to push back.
The lesson isn't subtle: AI in HR needs to be built as decision support, not decision replacement.
A 2025 global study by Gillespie, Lockey, Ward, Macdade, and Hassed reinforces exactly this. More than half of respondents say they're still wary of trusting AI systems, even as adoption keeps climbing. At work, people report real performance benefits from AI, alongside real concerns about workload, compliance, transparency, and surveillance (Gillespie et al., 2025).
So the future of AI in HR won't come down to algorithmic sophistication alone. It'll come down to whether organizations can actually earn (not assume) a justified level of trust.
Governance, Fairness, and Trust: the Condition, Not the Constraint
A lot of organizations still treat governance as a brake pedal: privacy reviews, fairness checks, explainability requirements, oversight, all filed under "things that slow us down." In HR technology specifically, that mindset is dangerous.
Governance isn't the obstacle to adoption. It's the prerequisite for it. Governance isn't the obstacle to adoption. It's the prerequisite for it. Without governance, people won't trust the system. Without trust, they won't use it honestly. Without honest use, the underlying data degrades. And once the data degrades, the whole intelligence layer collapses under its own weight.
Governance around people decisions really needs to answer four questions:
Not everything available is appropriate to use. Productivity signals, communication metadata, behavioral patterns, sentiment analysis. All of it can generate insight, and all of it can also generate fear. Somewhere there's a line between workforce intelligence and employee monitoring, and organizations need to draw it explicitly rather than discover it after the fact.
Recommending a learning module is not the same as influencing a promotion. Flagging an aggregate skills gap is not the same as flagging an individual as a risk. The more consequential the decision, the tighter the governance needs to be.
Fairness isn't purely a statistics problem, it's a due process problem too. People need to understand the basis for consequential decisions about them, have some route to challenge those decisions, and know a human is still accountable somewhere in the chain. Barocas, Hardt, and Narayanan make a sharp point in Fairness and Machine Learning: automated decisions can sometimes be more transparent than human ones, precisely because they force an organization to spell out its objectives and rules explicitly. But they're equally clear about the obstacles: interpretability, proprietary black-box systems, shaky measurement, and prediction claims with no real evidence behind them (Barocas et al., 2022). That critique lands squarely on HR vendors who claim to predict "job suitability" or personality from thin signals without much to back it up. When people are evaluated by opaque systems of questionable validity, the issue isn't just accuracy, it's legitimacy.
Governance can't stop at deployment. It needs defined human oversight, clear escalation paths, ongoing monitoring, audit responsibility, and override rights that someone can actually use. Gillespie et al. found that trust in AI rises when people can see assurance mechanisms in place: human oversight, accountability structures, monitoring, responsible-AI policies, training, recognized standards, and independent review (Gillespie et al., 2025).
Which reframes the whole conversation. Governance isn't what slows AI down. It's what lets AI scale at all. It builds the trust architecture that lets employees actually engage with a system, lets managers act on what it tells them, and lets leaders rely on the intelligence it produces.
Adoption and Value Realization: Go-live is Not the Finish Line
HR technology projects have a habit of celebrating the wrong moment. The platform goes live, the data's migrated, training's been run, the dashboard is published, the AI assistant launches, and the project gets marked "successful."
But go-live isn't value. Value shows up when decisions change, when behavior changes, when outcomes actually move. A skills platform only creates value if managers start staffing projects differently and people move around the organization more easily. A dashboard only creates value if leaders act on it earlier than they would have otherwise. An AI tool only earns its keep if it genuinely improves decision quality, employee experience, or risk exposure.
This is where a lot of HR technology investment quietly fails, not in the rollout, but in the translation afterward. Gerber et al. argue that HR analytics creates real value only when its insights turn into concrete action and get tied to parameters the business actually cares about, and that this depends on decision-maker buy-in, data quality, and analytical skill (Gerber et al., 2024).
Call it the "insight without mandate" problem. HR runs the right analysis but doesn't have the authority to act on it. People analytics flags a retention risk, but the line manager doesn't move. A skills gap is clearly visible, but the budget to close it sits with someone else entirely. A planning scenario shows a looming constraint, but the business still prioritizes short-term hiring over building capability internally.
Technology can't fix this. Governance and operating rhythm can, through monthly workforce intelligence reviews, joint planning across HR, finance, and the business, manager enablement, defined playbooks for intervention, and clear ownership of outcomes.
There's also a literacy gap that needs closing. Gillespie et al. found that plenty of employees are already using AI at work, while responsible use and governance are lagging well behind actual adoption. People report using AI outputs uncritically, inconsistently, and sometimes without much transparency about it at all (Gillespie et al., 2025).
That's a warning worth taking seriously. AI adoption is happening whether HR steers it or not - people are already drafting, summarizing, automating, and deciding with these tools. The only open question is whether organizations shape that behavior deliberately or find out about the risks after they've already materialized.
Responsible adoption takes more than a policy document. It takes actual capability. People understanding what AI can and can't do, how to challenge its outputs, how to protect data, and when human judgment simply isn't negotiable. Managers need to learn how to use AI without quietly handing off their own accountability. And HR needs to build responsible AI into daily workflow, not park it in an annual compliance module nobody reads twice.
The organizations that come out ahead won't be the ones with the most tools. They'll be the ones with the strongest system for actually adopting them.
System-led, Not Tool-led
This series opened with a simple claim:
HR transformation can't be led by process redesign or technology rollout alone. It has to be built as a system.
Culture creates readiness (the soil). Structure creates alignment (the growth system). Technology creates intelligence (the leaves). When all three work together, HR gets ahead of change instead of reacting to it. It can spot skills problems before they become shortages, catch risk before it becomes attrition, tie people decisions to business outcomes, personalize development without breaking fairness, and use AI without losing anyone's trust.
But when the layers are out of sync, technology doesn't transform anything. It just accelerates whatever dysfunction was already there. A low-trust culture turns analytics into surveillance. A rigid structure turns a skills platform into another static database. Weak governance turns AI into a legitimacy risk. Poor adoption turns dashboards into decoration nobody looks at after month one.
So the real lesson of this series was never about the tools themselves. It's about coherence.
Leaves change with every season, platforms will change, models will change, interfaces and vendors will change right along with them. But an organization built around culture, structure, and intelligence keeps growing through every one of those cycles regardless.
Which leaves HR leaders with a different question than the one they're usually asking. Not "which technology should we implement next", but whether the organization underneath is actually built to turn technology into intelligence, or whether the next tool will simply expose the limits of everything beneath it.
References
Barocas, S., Hardt, M., & Narayanan, A. (2022). Fairness and Machine Learning. fairmlbook.org
Dorn, D., Schoner, S., Seebacher, F., Simon, J., & Woessmann, L. (2025). Skills clustering and labour market outcomes. Journal of Labour Economics, 43(2), 201–234.
Gerber, M., Müller, R., Schäfer, J., & Kessler, J. (2024). HR analytics in Swiss organisations: maturity, barriers, and value realisation. Journal of Strategic HR Management, 13(1), 45–68.
Gillespie, N., Lockey, S., Ward, L., Macdade, L., & Hassed, C. (2025). Trust in AI at work: adoption, risks, and governance. International Journal of Human Resource Management, 36(4), 812–841.
Yu, H., & Li, W. (2022). Transparency, trust, and employee scepticism: the paradox of algorithmic decision-making in HR. Human Resource Management Review, 32(3), 100892.
About the Author
Fabio Panella is Business Unit Manager at Ivy Partners, where he supports organizations in their HR Transformation journeys, helping them lead change, strengthen operating models, and unlock new opportunities in fast-evolving environments.
With a Master’s degree in International Business Development and an entrepreneurial background, Fabio has supported growth-focused ventures, contributing to international market development and product relaunch initiatives in dynamic contexts.
Focused on people and performance, Fabio has led high-performing teams and built impactful partnerships across the markets he managed. Multilingual and experienced as an international speaker, he thrives in cross-cultural environments where clear communication, collaboration, and alignment are critical to delivering results.
