From emergency fixes to proactive performance capacity

The problem with traditional follow-up: “It's working fine” – until it isn't

A CEO asks the management team: “Is there anything I need to know?” Everyone answers: “No, it's working fine.” Two months later, three key people resign. The exit interviews all carry the same message: “We've been trying to say this for a long time.”

The problem wasn't a lack of information. The problem was that it never became visible to management.

Traditionally, organisational health and performance are measured through outcome measures such as sick leave, staff turnover, late deliveries, changes in productivity and recurring employee surveys. These KPIs matter, but they often describe the consequences of conditions that have already arisen. In other words, they are late signals.

By the time sick leave rises, collaboration deteriorates or productivity drops, the impact has already become both human and financial. Leadership therefore risks becoming reactive. Three limitations reinforce the problem:

1. Fragmented understanding: Data on HR, physiology, operational results and employee feedback often sit in different systems and are rarely connected.
2. Lack of continuity: Annual surveys or quarterly pulse surveys risk missing dynamic patterns of change and gradually increasing workload in everyday work.
3. Limited actionability: Data without context rarely gives a clear answer to what management should actually do next.

This is where Human Performance Intelligence offers a different path. By integrating perspectives from physiology, psychology, organisational research, data analysis and ethics, raw data can be turned into more actionable intelligence. The goal is to understand the human conditions that make long-term, sustainable performance possible.

How do we measure impact? Causality over correlation

A central question in modern impact analysis is how biometric data, behavioural data and organisational performance measures can be integrated to better understand the effect of long-term capacity work. Finding correlations is relatively easy. Showing that a particular intervention, routine or chain of events actually contributed to an improvement in the organisation's KPIs requires a more rigorous methodology.

One way is to combine continuously collected data in a shared analytical framework:

• Biometric data – physiological recovery: Heart rate variability (HRV) can provide an objective, non-invasive window into the regulation of the autonomic nervous system. Changes in HRV, together with context and the individual's baseline, can add information about workload and recovery.
• Behavioural data and qualitative dialogue: Coaching conversations and self-reported reflections capture the individual's experience and give context to the quantitative signals.
• Business data: Production data, lead times, quality measures or other relevant outcomes at group level show what is actually happening in the organisation.

Through longitudinal analysis, changes at group level can be followed over time and related to changes in the organisation's results. To separate random associations from more likely effects, statistical causal inference can be combined with preliminary causal models that map and control for confounding factors.

The critical management question
Which interventions work, for which groups and under which organisational conditions?

LLMs as an analytical tool: structuring the unstructured

The next challenge is all the unstructured information generated in an organisation – for example notes from coaching dialogues, reflections and written feedback. Historically, this type of data source has been hard to scale and analyse statistically without extensive manual work.

Here, Large Language Models (LLMs) can be used as more than a summarisation tool. Through semantic coding, unstructured text can be turned into structured, analysable variables that can then be combined with other data sources.

In practice, LLMs can contribute to impact analysis in three ways:

1. Semantic coding of dialogues: Linguistic patterns in coaching dialogues can be analysed to identify themes such as focus, motivation, collaboration and perceived social support.
2. Identifying behavioural change: Gradual changes in energy, concentration or self-regulation can become visible in text before they show up in traditional outcome measures.
3. Generating hypotheses about causal mechanisms: Larger bodies of text can be used to identify patterns and suggest hypotheses about which factors may be connected. These hypotheses can then be tested in more structured models.

In this way, the human, qualitative dimension doesn't have to disappear into the statistics. On the contrary, employees' reflections can gain a clearer structure and become part of a broader basis for decisions.

From signals to knowledge: AI-driven support for management decisions

The decisive question is how complex data flows can be turned into something a senior leader can actually base decisions on. Within Human Performance Intelligence, this can be described as a chain with four levels:

1. Observation: Collecting physiological signals, for example HRV from wearables, together with reflection questions and business data.
2. Analysis: Raw data is normalised and processed to identify patterns, trends, deviations and relationships over time. In complex organisations, patterns of change and variation are often more important than a single measurement.
3. Interpretation: Context is put at the centre. Workload is not negative in itself if it is short-lived, understandable and followed by sufficient recovery. Data therefore needs to be related to role, pace of work and the actual work situation.
4. Decision support: The interpretation is translated into transparent, action-oriented insights at group level for management and HR.

The result is a more coherent picture of the organisation's long-term ability to deliver. Management can follow trends in workload, recovery capacity, focus and collaboration – and thereby create better conditions for acting before the development becomes a visible problem.

If the data shows, for example, that a team has displayed a combination of increased workload and reduced focus for several weeks, the leader can look into the need to adjust resources, priorities or ways of working before the situation leads to bigger consequences.

Explainability is central. AI-based decision support must show why a recommendation is given and which patterns or relationships lie behind it. Management should be able to use the technology to sharpen its judgement – not replace it.

Ethics, privacy and human judgement at the centre

Physiological measurement and AI analysis of employee data place high demands on ethics and privacy. If the technology is perceived as surveillance, it risks damaging the trust that is essential to a well-functioning organisation. A sustainable approach therefore requires clear principles:

• Human judgement first: AI-generated insights are not absolute truths. Human performance is complex and cannot be reduced to individual data points. The technology should enrich dialogue and decisions, not replace them.
• A clear and supportive purpose: The technology should support the individual's self-leadership and help leaders understand their team. It must not be used for hidden control or scrutiny of individuals.
• Data minimisation and privacy: Only data relevant to the purpose should be collected, and sensitive individual data needs to be protected through aggregation and clear access principles.

The strategic shift for the leadership of the future

For a senior leader, the long-term benefit of multimodal AI and Human Performance Intelligence lies in a shift from merely reporting what has already happened to better understanding what is in the process of changing.

• What is happening with the people?
• What is happening in the team?
• What is in the process of changing?
• And what does the leader need to do now?

This is where the difference between reporting and intelligence becomes clear.

Through data-driven, scientifically grounded impact analysis, the gap between the individual's long-term capacity and the organisation's endurance in production can be reduced. That creates better conditions for building high-performing, sustainable organisations in a complex time.

Jenny Nilsson

Founder, Zmartrest AB
jenny.nilsson@zmartrest.com
+46 725-277 276