Arosha Brouwer – “Predictive HR: change and action needed”

More and more organizations aim to prevent absenteeism and burnout before employees actually drop out. But how can early warning signs be recognized in time? And what role do data, leadership, and organizational culture play in this process? In this interview, Arosha Brouwer, CIO at TrueTribe explains how predictive HR can help organizations move from reactive to proactive workforce strategies.

Brouwer discusses why early signals are often visible long before absence rates increase, and why real change starts with the way work itself is designed. ‘Most HR functions measure and report outcomes rather than causes. Absence rates, turnover, and engagement scores reflect what has already happened, the result of mechanisms building for months or years. By the time those numbers move, the window to intervene has closed.’

Three signal categories

Earlier risk prediction rests on three signal categories:

• Work conditions
  Sustained high demands combined with low autonomy is the single most replicated predictor of health breakdown in occupational science. Karasek’s demand-control model has been validated across decades and dozens of countries – this is a structural risk factor, not a soft concern. TNO data shows approximately 16% of Dutch workers experiences this pattern structurally, yet most organizations don’t monitor it at all.
• Social and team dynamics
  Declining psychological safety, reduced informal connection, and withdrawal from collaboration are consistent early indicators. Edmondson’s research shows team social climate predicts both engagement and health outcomes; Bakker and Demerouti’s job demands-resources work identifies belonging and social support as the core buffers against burnout. When those resources erode, risk builds quietly, which is precisely why the damage is so easy to miss until it’s done.
• Sustainability of individual capacity
  Whether someone can sustain their current level of functioning over time. Observable signals include changes in output quality, working outside normal hours, and declining use of leave, but the underlying drivers are physical: sleep, health, and energy determine how much someone can carry and for how long. These behavioral shifts often precede formal absence, though lead times vary by person and context.’

Brouwer: ‘And not everything shows up in a dataset. Managers close to their teams often sense a shift before any system captures it, a change in someone’s presence or energy, a conversation shared in confidence. That observation is data, too. A structural early warning system needs to make room for it.’

Leadership

‘Fragmentation persists because each function was built to solve its own problem. HR tracks engagement, finance tracks costs, occupational health tracks absence duration. There is no definition of what good looks like, no shared data, and no shared accountability for outcomes. More than half of all work-related absence costs in the Netherlands are attributable to psychosocial workload – factors that sit in HR’s domain – yet the conversation typically starts in occupational health, by which point prevention is no longer possible.’

‘A shared view requires leadership that treats workforce health as a strategic performance indicator – with real accountability for the upstream conditions that drive outcomes, not just the absence rate that results from them. By the time absence figures move, the conditions that caused them have been building for months, sometimes years. Acting on earlier signals requires functions to look at the same data and be accountable for the same outcomes. That is a leadership question before it is a systems question.’

Prioritization and action

‘Most organizations have invested in collecting signals and generating insights, but far fewer succeed in consistently translating those insights into action and measurable outcomes. Data in itself changes nothing. The chain that actually matters is: signals, insights, action, and result. I recommend first to try to create awareness of this gap, because recognizing the problem is a prerequisite for improvement. From there, the focus shifts to making insights actionable: identifying where risks concentrate, understanding which underlying factors are likely driving them, and linking these to targeted interventions. Equally important is closing the feedback loop by measuring whether interventions improve outcomes over time. Without that learning cycle, dashboards remain descriptive rather than transformative.’

‘The Dutch data illustrates the gap precisely. TNO’s Arbobalans 2024 shows that while compliance with statutory obligations has improved – 64% of organizations now conduct a risk inventory, up from 46% in 2014 – the measures most commonly taken are process-level: promoting open culture, managing absence after it occurs. Fourteen percent of organizations took no measures at all. The upstream conditions that actually drive absence – workload, autonomy, role clarity – are rarely what gets addressed. The bottleneck is rarely data availability. It is the translation layer between insight and intervention. A manager who sees that workload is high, autonomy is low, and engagement is declining needs to know what to do next. Do they redesign roles? Change how they lead? Flag a capacity problem upstream? Without a practical protocol and organizational support to follow through, the insight sits in a dashboard, and the situation continues to deteriorate.’

‘Prioritization matters enormously. Research on occupational health intervention consistently shows that diffuse, unfocused responses produce little measurable impact. Act first on what most directly predicts serious outcomes. Not everything is equally urgent and treating it as such creates noise that makes it harder, not easier, to respond to what genuinely matters.’

Overload

‘Two patterns of signs of overload stand out consistently. The first is how organizations respond to workload pressure by adding resource without touching role design. When someone is consistently working beyond their capacity, the instinct is often to offer support – a coach, a training, a conversation about prioritization. These are not wrong, but they leave the structural conditions intact. The workload stays high, the autonomy stays low, and the support sits on top of a problem that hasn’t been addressed at its root.’

‘The second is the erosion of psychological safety – specifically, the point at which people stop asking for help. That moment is rarely visible in data, but it is one of the most consequential early signals available. When someone who used to raise concerns goes quiet, when a team that used to push back stops doing so, psychological safety has already broken down. People learn quickly whether raising a hand leads to support or to scrutiny. Once they conclude it leads to the latter, they manage their situation privately – and by the time it surfaces in absence or turnover, the window for easy intervention has long closed. These signals are felt before they are measurable, which is precisely why manager judgment, informed by data but not replaced by it, remains the most underutilized early warning system in most organizations’

Monitoring versus supporting

‘The difference between feeling monitored and feeling supported comes down to three things: transparency, consent, and what happens next. If measurement leads to a conversation that helps someone, it feels like support. If it leads to a file being updated somewhere that the employee never sees, it feels like surveillance. That distinction is entirely within an organization’s control.’

‘In practice, there are concrete things organizations can do at every level:’

• At the individual level, give people access to their own data first. When an employee receives insight into their own wellbeing trends before anyone else does, the dynamic shifts they become an active participant rather than a subject being observed. Simple, low-threshold check-ins that feedback directly to the person completing them are far more effective than annual surveys that disappear into a report nobody sees. At TrueTribe, we match individuals with relevant self-help content and expert-facilitated interventions based on what those signals show.
• At the managerial level, the highest-leverage investment is capability, not more data access. A manager who receives a signal that someone may be struggling needs to know what to do with it – how to have an early, low-stakes conversation about workload and energy before things become critical. Most managers want to support their people but feel underprepared.
• At the organizational level, close the loop visibly. If signals flow upward and nothing changes, people stop being honest. The organizations that build real trust around their data are the ones that can point to concrete outcomes: autonomy scores were consistently low across a department, the data traced it back to how work was being allocated, and the response was a structured conversation with leadership about delegation – not a training course pushed out to the team. Or: capacity and durability signals were consistently elevated among women in a specific age bracket – the data pointed to menopause as a potential contributing factor, so the organization put concrete support in place: adjusted working arrangements, access to relevant expertise, and manager training on how to have the conversation The response must match the signal. This is precisely why we provide data and suggests interventions across all three levels: individual, team, and organizational. The right intervention at the right moment is what turns measurement into something people trust.’

Ownership

‘The organisations that appear to be progressing most successfully share several structural characteristics. Wellbeing is positioned at senior leadership levels, not as a standalone wellbeing initiative, but as a strategic performance indicator with clear ownership and accountability. These organisations have also moved beyond isolated interventions towards systems thinking. Rather than asking which programme to offer employees, they increasingly ask how work itself is organised, and whether that design supports or undermines sustainable employability. In addition, they invest heavily in manager capability, recognising that managers are the critical link between data and action, and between early signals and timely intervention.’
‘We are working with several large Dutch organisations including Intergamma and PostNL to explore how analytics can be used to move from generic absence management to a targeted and preventative wellbeing strategy.’

Opportunities and risks

‘The biggest opportunity is moving from reactive to evidence-based decision-making. Most organisations still rely on lagging metrics – absence rates, turnover, engagement scores – without understanding which underlying factors drive them or which interventions actually work. AI makes it possible to identify risk patterns at a scale no human analyst can match, integrate fragmented signals, and surface early indicators that would otherwise stay buried. But two risks deserve serious attention. The first is context collapse: an algorithm trained on absence records alone, without data on workload, autonomy, or social context, produces a partial picture and acts on it with false confidence. The EU AI Act specifically categorizes AI systems used in employment contexts as high risk, requiring transparency, human oversight, and demonstrable fairness – and rightly so. Decisions made about people’s working lives on the basis of incomplete models cause real harm.’

‘The second is measurement inflation. AI makes it easier than ever to track more, but continuous monitoring has in documented cases driven stress levels up, not down. More tracking is not better insight. This is why the governance layer matters as much as the technology itself. Working with an ISO-certified organisation means the data handling, privacy protections, and ethical frameworks underpinning any AI-driven insights have been independently verified – not just asserted. It’s a meaningful predictive signal that the system has been built responsibly, not just built quickly.’

Cultural change

‘Predictive HR without cultural change is sophisticated surveillance, nothing more. For early signals to surface, in data and in conversation, people need to feel safe enough to show them. And psychological safety is a cultural condition before it is a technical one.’

‘The honest answer to whether you can implement predictive HR without cultural change is no. The most sophisticated early warning system in the world depends on people being willing to raise their hand before they reach a breaking point. That only happens in organizations where doing so is genuinely safe, where a manager’s response to ‘I’m struggling’ is curiosity and support, not performance management. Where leaders talk about pressure in the present tense, not just in retrospect once they have recovered. Where asking for help is modeled from the top, not just encouraged in a policy document.’

‘Beyond culture, there is a structural requirement that is equally non-negotiable: the willingness to look at how work itself is designed. Proactive HR is not primarily about better support programs. It is about asking whether roles are sustainable, whether workload is realistic, whether people have enough autonomy to manage their own energy and recovery. Those are organizational design questions, not HR program questions. And they require leaders who are willing to hear uncomfortable answers and act on them, not just commission another engagement survey.’

Business risk

‘So, to improve workforce health tomorrow, I advise to have the conversation you have been avoiding. Most HR directors already know where the pressure points are in their organization. They know which teams are running too hot, which managers are not equipped to support their people, which roles are structurally unsustainable. The data usually confirms what people already sense.’

‘The gap is rarely knowledge. It is the willingness to take that insight to the board and reframe it, not as a wellbeing issue, but as a business risk with a measurable cost and a concrete intervention point. Absence does not appear from nowhere. It builds in conditions that are often visible, often known, and often left unaddressed because the conversation feels difficult or the solution feels expensive.’

Start tomorrow by asking one question in your next leadership meeting: what are the three conditions in our organization right now that are most likely to produce serious absence or turnover in the next twelve months? If the room can answer that question specifically and honestly, you have the foundation for a proactive strategy. If nobody knows, or the answers stay vague, that is your starting point. Close that gap first, before investing in any new tool or program.’

Arosha Brouwer is Chief Impact Officer at TrueTribe and Founder of Quan. TrueTribe is the main sponsor of HR Live 2026, the largest network event on HR in the Netherlands.