We think the cost of unmanaged cognitive load over the next decade will be enormous, and that most organisations are measuring it with the wrong instruments.1
Burnout has become a global crisis. Around 83% of workers report feeling its strain, and companies are estimated to lose $322 billion in productivity in 2026.2 This paper sets out what the current tools miss, and what an agentic coach does instead.
The crisis
Organisations already know their people are tired. What they lack is a signal that arrives before the damage. Engagement surveys are retrospective, annual, and answered by people who have learned what a safe answer looks like. By the time a score drops, the person has usually already decided to leave.3
Physiology moves earlier. Heart rate variability, resting heart rate and sleep architecture shift weeks before someone reports feeling depleted. Behavioural load moves earlier still: meeting density climbs, focus blocks disappear, messages arrive later at night. Both signals exist inside most companies today. Neither is being read together.
The symptoms are visible everywhere: rising turnover in critical roles, more sick leave taken for stress related mental health issues, and managers who can see their teams are exhausted but cannot pinpoint where the pressure is coming from or how to relieve it without jeopardising delivery.
Historical context: how we got here
Shift limits, rest breaks, safety standards
Surveys, apps, optional coaching. Reactive and voluntary
Track, then display. Interpretation left to the depleted
Leaner teams, no boundaries, higher throughput
Burnout is not new, but the shape of it has changed.
From industrial fatigue to cognitive overload
In the 20th century, workplace fatigue was largely physical: long hours on factory floors, shift work, and manual labour. Occupational health frameworks evolved to measure and manage that risk with maximum shift lengths, rest breaks, and safety standards.
As work shifted into knowledge and service economies, the dominant form of strain moved from muscles to minds. The stressors became:
- •always on communication
- •constant context switching
- •blurred boundaries between work and home
- •performance metrics that reward responsiveness over recovery
The language of wellness emerged, but the systems designed to manage risk did not keep pace.
The first wave of wellness: programmes and surveys
The 2010s saw a surge in corporate wellness:
- •mindfulness apps
- •annual engagement surveys
- •optional coaching programmes
- •mental health days
These were important steps, but they shared a common limitation: they were reactive and voluntary. They relied on people who were already depleted to opt in, self report accurately, and then act on generic advice.
Surveys captured sentiment after the fact. Programmes reached a small, often already engaged subset of employees. The core work design and workload patterns remained untouched.
The second wave: wearables and dashboards
The 2020s brought wearables and AI wellness platforms into the workplace. For the first time, organisations could see physiological signals: stress scores, sleep duration, activity levels.
But the model remained the same: track, then display.
Dashboards showed managers and individuals a stream of numbers. The interpretation, and the decision to act, was left to the person least able to make it while depleted. A readiness score of 41 is a fact, not a plan.
At the same time, workload data proliferated: calendars, chat logs, ticket systems, meeting analytics. But these lived in separate silos. HR had surveys. IT had collaboration metrics. Individuals had their wearable app. No one connected physiology to workload in real time, and no system intervened inside the flow of work.
The 2026 moment: structural exhaustion
By 2026, several forces have converged:
- •Hybrid and remote work dissolved the remaining boundaries between office hours and personal time.
- •AI driven productivity expectations have increased throughput without reducing meeting load or administrative overhead.
- •Economic pressure has led to leaner teams absorbing more work, with fewer buffers.
- •Digital fragmentation means attention is constantly pulled in multiple directions.
The result is a workforce that is not just stressed, but structurally exhausted. Burnout is no longer an individual failure, it is a system level condition.
The gap
The central gap is not data. It is integration and action.
Most organisations now have:
- •some form of wellbeing survey
- •some form of wearable or health data, often voluntary
- •rich behavioural data in calendars, chat, and work tools
But these are measured separately, aggregated too late, and presented passively. The outcome is a lagging, incomplete picture that cannot guide timely interventions.
What current tools miss
Surveys lag behavioural change by one to two quarters in most published studies. By the time a team's engagement score falls, the underlying patterns have been entrenched for months.
A low recovery score after a hard training week means something different from the same score after four consecutive days of back to back meetings. Without workload, the number is noise. With it, the number becomes an instruction.
Even when risk is identified, there is rarely a built in mechanism to change the conditions that created it. The advice is generic: take a break, speak to your manager, use your benefits. The system that created the load remains unchanged.
The responsibility to interpret and act falls on the individual. But cognitive depletion impairs exactly the capacities needed to make good decisions: attention, planning, and impulse control.
Why existing solutions fail
Engagement surveys
- •Frequency. Annual or semi annual, not continuous.
- •Bias. People learn what safe answers look like.
- •Granularity. Aggregated at team or org level, too coarse to act on specific workload patterns.
- •Timing. Retrospective, they measure sentiment after the damage is done.
Wearables and wellness apps
- •Focus. Track stress, not workload.
- •Interpretation. Left to the user, without integration into work systems.
- •Action. Generic recommendations that do not change meeting schedules, deadlines, or role design.
- •Adoption. Often voluntary, skewing towards already health conscious employees.
Manager training and coaching programmes
- •Scale. Limited by human coach capacity, they reach a fraction of the workforce.
- •Timing. Often triggered after a crisis, such as a performance drop, sickness, or resignation risk.
- •Context. Managers may lack visibility into the real drivers of load, such as cross team dependencies, hidden work, and tool friction.
Productivity and workload analytics
- •Focus. Efficiency, not wellbeing.
- •Risk. Can optimise for output at the expense of recovery if used in isolation.
- •Privacy. When not designed with strong safeguards, these tools can feel surveillant, eroding trust.
In short, the current landscape is a patchwork of partial signals and partial solutions. None of them closes the loop from signal, to context, to intervention, to changed workload.
The agent
Sloane models each person's own baseline rather than a population average, then watches for drift away from it. When drift appears, Sloane acts inside the tools people already use: it suggests moving a session, protecting a focus block, or taking a restorative break at the hour it will actually help.4
Dashboards ask for attention from people who have none left. Agents spend their own attention instead.
How it works
For each individual, Sloane learns a personal baseline across physiological signals such as HRV, resting heart rate and sleep patterns, behavioural signals such as meeting density, after hours activity and focus time, and self reported states when provided. The model is adaptive, accounting for normal variation such as travel, illness and life events, while flagging sustained drift.
Sloane fuses biometrics with workload data from calendars and collaboration tools. This allows it to distinguish a tough but planned sprint from a chronic overload pattern with no recovery.
When risk thresholds are crossed, Sloane does not just notify. It proposes concrete changes: rescheduling or shortening meetings, carving out protected focus blocks, prompting micro breaks at moments likely to restore rather than interrupt flow, and suggesting workload rebalancing to managers when patterns persist.
Individuals retain control. Suggestions are optional and explainable. Over time, the agent learns which interventions are helpful for each person and adjusts.
Why agentic coaching matters
- •Timing. Interventions arrive when they can still change the day, not weeks later in a report.
- •Personalisation. Advice is tailored to the individual's baseline and current context.
- •Integration. Coaching happens inside existing tools such as calendar, chat and task managers, not in a separate app that requires extra effort.
- •Scale. An agent can support thousands of people simultaneously, far beyond the reach of human only coaching.
The multi-modal training architecture
You cannot pour raw heart rates into a language model and expect a useful coaching prompt. A heart rate of 95 beats per minute is panic at a desk and perfectly normal on the stairs. Raw numbers mean nothing without context, so the system runs as three separated tiers, each doing a job the others cannot.4
Tier 01, the biometric normalisation layer
Light time-series models learn each person’s baseline, what their body looks like under normal, high-stress and recovery states. Strain is scored against the individual, never a population average, so the same number means different things for different people.
Tier 02, the telemetry contextualiser
This layer reads the digital footprint: calendar density, back-to-back meetings, and messaging velocity across Slack and Teams. Metadata only, never message content, so it sees the shape of the working day without reading a single word.
Tier 03, the agentic reasoning engine
The coaching agent maps the two vector streams together and reasons. High strain plus a high-stakes meeting in 45 minutes means a high risk of panic, so it triggers a decompress-and-defer prompt. The agent only wakes when both tiers agree, and nothing is sent until the person confirms.
The agent does not see raw numbers. It sees meaning, and it reasons about what to do next.
The leader view
Leaders get a view of their teams' cognitive capacity, always aggregated, never individual. Capacity, recovery and risk sit alongside the workload decisions that shape them, so a hiring plan or a launch date can be checked against the capacity that has to absorb it.
No manager sees an individual's biometrics. Aggregation thresholds and minimum group sizes are enforced in the data layer, not in policy alone.5
What leaders can see
- •Team level capacity trends over time, with clear markers of sustained overload.
- •Workload drivers: which meeting patterns, project phases, or cross team dependencies correlate with capacity drops.
- •Risk hotspots: teams or functions where cognitive load consistently exceeds recovery.
- •Scenario testing: the ability to model the impact of proposed changes, such as deferring a launch, adding a hire, or redistributing work.
What leaders cannot see
- •Individual biometric data.
- •Individual workload breakdowns below aggregation thresholds.
- •Any data that could be used for performance evaluation or disciplinary action.
This separation is intentional. The purpose is to improve work design, not to monitor individuals.
Future performance
- •Earlier detection of risk
- •Fewer acute overload episodes
- •Higher reported support
- •Lower turnover in high risk roles
- •Stable performance across quarters
- •Capacity aware planning
- •Capacity as strategic risk
- •Recovery designed into work
- •Talent advantage
Figures in this paper describe the market context Sloane is built for, not measured product outcomes. However, we can outline the performance we expect as adoption deepens and the system matures.
Short term, 0 to 12 months
- •Earlier detection of burnout risk: shifts in physiology and behaviour identified weeks before traditional surveys would flag a problem.
- •Reduction in acute overload episodes: fewer periods of sustained high meeting density with no recovery.
- •Improved subjective wellbeing: employees report feeling more supported and more able to manage workload.
Medium term, 12 to 36 months
- •Lower turnover in high risk roles, as chronic overload patterns are addressed earlier.
- •More stable performance across quarters, with fewer heroic sprints followed by crashes.
- •Better hiring and planning decisions, as leaders can see capacity constraints before committing to new work.
Long term, 3 to 10 years
We expect the cost of unmanaged cognitive load to become a central strategic risk, comparable to financial or operational risk. Organisations that integrate biometrics, workload and agentic coaching will likely see:
- •Sustained performance without relying on burnout as a hidden fuel.
- •Cultural shift, where recovery and focus become normal parts of work design, not optional extras.
- •Competitive advantage, the ability to attract and retain talent in a world where burnout is widespread.
Our hypothesis is that the delta between organisations that continue to rely on surveys and those that adopt integrated, agentic systems will widen over the decade, not narrow.
Adoption and change
Technology alone will not solve burnout. Adoption depends on trust, clarity of purpose, and alignment with existing ways of working.
Who must adopt
- •Individuals, who must feel safe to engage with the system and act on its suggestions.
- •Managers, who must use aggregated insights to adjust workload and protect capacity.
- •HR and People teams, who must integrate Sloane into broader wellbeing and work design strategies.
- •Executive leadership, who must endorse the principle that capacity is a strategic metric.
Barriers to adoption
- •Privacy concerns. Fear that biometric and behavioural data will be used against employees.
- •Scepticism. Belief that this is another wellness initiative that will not change real work.
- •Change fatigue. Teams already overloaded may see new tools as additional burden.
How to address them
Enforce aggregation and minimum group sizes in the data layer. Make data use and access controls transparent and auditable. Separate wellbeing data from performance management systems.
Run time bound pilots with volunteer teams. Measure outcomes such as turnover, sick leave and engagement, and share learnings openly. Let early success stories drive organic adoption.
Use insights to change meeting norms, deadline structures, and role design. Frame Sloane as a tool for better work, not just better feelings.
Provide training on interpreting aggregated capacity data. Give managers concrete playbooks for responding to overload signals, such as redistributing work or deferring non critical projects.
Be explicit about what data is collected, how it is used, and who can see it. Reiterate that the goal is to improve systems, not to monitor individuals.
Safety and governance
Safety is not an add on, it is the foundation.
Data protection model
- •Minimal necessary data. Only data required to model capacity and risk is collected.
- •Encryption. Data is encrypted in transit and at rest.
- •Access controls. Strict role based access, with individual and aggregated team views technically separated.
- •Retention limits. Data is retained only as long as needed for its stated purpose, then deleted or anonymised.
See the safety page for the full data protection model, retention and access controls.
Ethical guardrails
- •No use for performance evaluation. Biometric and behavioural data must not be used in performance reviews, promotion decisions, or disciplinary processes.
- •Voluntary participation. Individuals opt in and can opt out at any time without penalty.
- •Transparency. Clear documentation of the purposes and limitations of algorithms, with no black box scoring without explanation.
Governance
- •Oversight board. A cross functional group covering HR, legal, security and employee representatives reviews policies and major changes.
- •Audit trails. All access to data is logged and auditable.
- •Incident response. Clear procedures for data breaches or misuse, with prompt notification and remediation.
Regulatory alignment
Sloane is designed to align with major data protection and employment regulations, including:
- •GDPR and UK data protection law
- •emerging AI and algorithmic accountability frameworks
- •sector specific guidance on employee monitoring and wellbeing data
The result
The result is healthier and more engaged employees, lower turnover, and performance that holds across a quarter rather than being rescued one person at a time. Coaching arrives at the moment it matters, inside the flow of work, at a scale no human coaching programme can reach.
Human data into human potential.
- 1Figures in this paper describe the market context Sloane is built for, not measured product outcomes.
- 2Aggregate estimates of burnout prevalence and lost productivity across knowledge work, 2026.
- 3Annual engagement instruments lag behavioural change by one to two quarters in most published studies.
- 4Sloane connects to the leading wearables, and to workload through calendar and collaboration tools.
- 5See the safety page for the full data protection model, retention and access controls.