Automation is often presented as a simple exchange. We give a system a repetitive task, and it gives us time back.
That is sometimes true. However, convenience also changes how we pay attention, remember information and make decisions. The hidden cost of automation is not always financial or technical. It can be psychological.
In Part 7 of our AI vs Humanity series, we are looking at three connected effects:
- The tendency to trust automated outputs too readily
- The weakening of memory when we constantly offload information to devices
- The growing wellbeing challenge created by always-on digital systems
The aim is not to reject automation. It is to understand its influence clearly enough to use it responsibly.
Automation bias and the temptation to accept the answer
Automation bias is the human tendency to favour an automated recommendation over our own judgement, other evidence or professional experience.
It can appear in a cockpit, a hospital, a public-sector assessment system or an ordinary office workflow. The technology does not need to be described as intelligent for this to happen. A dashboard, alert, scoring tool or automated approval process can all acquire an undeserved sense of authority.
Two types of error are particularly important.
Errors of commission
An error of commission happens when a person follows incorrect automated advice.
For example, an automated system may recommend the wrong action while other information suggests that the recommendation should be questioned. If the person follows the system anyway, the mistake has been actively carried out.
A well-known aviation simulation demonstrated this risk. Pilots were given incorrect automated instructions during an emergency scenario, and many followed the advice despite contradictory information in the cockpit. The research is discussed in this aviation study published in PLOS ONE.
Similar concerns have emerged in healthcare. Clinical decision-support tools can help professionals identify patterns, but research has also examined how easily users may accept recommendations that conflict with clinical signs or their own knowledge. The clinical research indexed by PubMed highlights why automated advice must support, rather than replace, professional judgement.
Errors of omission
An error of omission happens when a person fails to act because the system did not flag a problem.
This can be less visible. Nobody has necessarily followed an obviously wrong instruction. Instead, something important has been missed because the system appeared to be monitoring the situation.
In a business, that might mean:
- A security alert is not raised, so nobody investigates unusual activity
- A backup is assumed to be working, so nobody tests whether data can actually be restored
- A customer issue is not escalated because an automated triage tool marked it as low priority
- A performance report looks normal, so underlying staff concerns remain unexplored
Research across aviation, medicine and public-sector decision-making shows that automation bias is a real behavioural risk. It becomes more likely when people are under time pressure, cognitive load or information overload. In those circumstances, accepting the system’s output can feel like the sensible way to reduce mental effort.
That is the important point. Automation bias is not necessarily caused by carelessness. It is often a predictable response to a pressured environment.
Why time pressure changes behaviour
Human beings use mental shortcuts to manage complexity. When there is limited time to compare multiple information sources, an automated answer can become a substitute for active investigation.
This creates a dangerous feedback loop:
- The system produces a recommendation
- The person assumes it is probably correct
- The person stops looking for contradictory evidence
- The organisation records the outcome as an accepted decision
- Future users trust the system even more
The more authoritative a system appears, the more important it becomes to design clear moments for human challenge.
Cognitive offloading and digital amnesia
Cognitive offloading means transferring a mental task to something outside the brain. We write a note, set a reminder, save a document, take a screenshot or rely on a search engine to retrieve information later.
This is not inherently negative. External tools are valuable because human memory is limited. A reliable calendar prevents us from having to remember every appointment. A written procedure allows a team to follow a consistent process. A password manager reduces the burden of retaining multiple complex passwords.
The trade-off is that information we expect a device to remember may be encoded less strongly in our own memory.
A 2025 review in Nature Reviews Psychology examined the benefits and potential costs of cognitive offloading. The review on cognitive offloading for retrospective information found that external aids can improve performance during a task while also weakening later recall if the external record is unavailable.
This is sometimes described as “digital amnesia”. It is not a medical diagnosis. It is a useful phrase for a familiar experience:
- We remember where a document is stored, but not what it says
- We recognise a person’s name only after checking our contacts
- We can find an answer online, but struggle to explain the underlying idea
- We save screenshots and bookmarks that we never properly process
- We follow digital instructions without understanding the reasoning behind them
What the brain may be doing
Recent fMRI research has provided an interesting perspective. In an experimental study, participants received instructions to remember information, offload it to an external reminder or forget it.
The researchers found that brain activity associated with offloading resembled activity associated with deliberate forgetting more closely than activity associated with active remembering. The neural representation of the information also faded in the offloading and forgetting conditions.
The findings are still emerging and should be interpreted carefully. The relevant work is available as a research preprint discussed through Sciety, rather than treated as the final word on digital memory.
Even so, the practical message is clear. When we know that information is safely stored elsewhere, the brain may reduce the effort it gives to maintaining that information internally.
The benefit is real, but so is the cost
Offloading can free mental capacity. That spare capacity may help us solve another problem, focus on a customer or handle a more demanding decision.
The problem arises when we offload everything by default.
If employees never learn the core process because a system always guides them, they may struggle when:
- The system is unavailable
- The data is incomplete
- The recommendation is wrong
- A new situation falls outside the programmed rules
- They need to explain a decision to a customer, colleague or regulator
A resilient organisation needs both dependable external systems and people who understand what those systems are doing.
This is one reason documentation, training and regular review remain important. A procedure should not merely tell people which button to press. It should also explain what the process is designed to achieve and what warning signs require escalation.
The wellbeing dimension and the always-on workplace
Automation can remove repetitive work, but it can also increase the speed and volume of work around it.
When alerts, notifications, dashboards and automated requests are always active, people can feel that they are never fully off duty. The issue is not just screen time. It is the mental switching between tasks and the expectation that every new message, alert or workflow deserves an immediate response.
This contributes to:
- Information overload
- Digital fatigue
- Fragmented attention
- Reduced recovery time
- Anxiety about missing something important
- Presenteeism, where people continue working while unwell
The CIPD Health and Wellbeing at Work 2025 report recorded average UK absence at 9.4 days per employee, the highest level in more than 15 years. Mental ill health was identified as the leading cause of long-term absence, while presenteeism also remained a serious concern.
These figures do not prove that automation alone is causing poor mental health. Workload, management, job security, personal circumstances and organisational culture all matter. However, automation can intensify existing pressures when it is used to increase pace without improving clarity or support.
Simplyhealth’s 2026 analysis of mental ill-health absence reported that the UK passed five million estimated working days lost to mental ill health on 24 March 2026. Its published age breakdown groups younger workers as 18–34, with this group reporting particularly high rates of time off for anxiety, depression, stress and burnout.
The distinction matters. We should use the available evidence accurately rather than turn a broad age category into a more precise claim than the research supports.
As Paul Schreier, Chief Executive of Simplyhealth, put it:
“Creating conditions for people to thrive at work is good for everyone.”
That principle applies to technology design as much as to workplace policy. A system that saves five minutes but creates constant uncertainty may not be improving working life overall.
Practical takeaways for businesses
The answer is not to remove automation from the workplace. It is to create sensible boundaries around its use.
High priority actions
-
Keep humans accountable for final decisions
Make it clear who reviews an automated recommendation and who has authority to challenge it. -
Build in independent checks for high-impact decisions
Where a decision affects safety, access to services, finances, employment or personal data, require a second source of evidence. -
Train people to recognise automation bias
Use practical examples of errors of commission and omission. Staff should know that questioning a system is part of responsible work, not a failure to follow procedure.
Medium priority actions
-
Show the evidence behind automated outputs
A result should be explainable enough for a trained person to review. “The system said so” is not a sufficient audit trail. -
Decide what knowledge must remain internal
Use external reminders for routine details, but ensure that people understand critical processes, risks and recovery steps. -
Test failure scenarios
Ask what happens if the system is unavailable, fed incomplete data or produces an implausible recommendation. -
Monitor wellbeing alongside productivity
Track workload, after-hours activity, absence, presenteeism and staff feedback. A rise in completed tasks does not necessarily indicate a healthier workplace.
Lower priority, but still worthwhile
-
Create notification boundaries
Define when alerts require immediate action and when they can wait. -
Review automation regularly
A process that was helpful last year may now create unnecessary complexity or cognitive load. -
Encourage short periods of focused, device-free work
People need opportunities to think, remember and solve problems without continuous prompts.
For organisations reviewing their wider IT environment, ABC Service’s managed IT services provide an example of a support model built around monitoring, documentation, security checks and regular business reviews. The underlying principle is more important than the service itself. Systems should help people observe, review and question what is happening.
Summary and takeaway
Automation changes more than the tasks we complete. It changes the way we distribute attention, responsibility and memory.
Automation bias can lead to errors of commission and omission, particularly when people are rushed or overloaded. Cognitive offloading can free valuable mental capacity, but it may weaken recall and deep understanding when we rely on external systems for everything. Always-on automation can also add to digital fatigue and workplace stress if greater efficiency simply becomes an expectation to work faster and remain permanently available.
The responsible use of automation depends on human oversight, clear accountability and healthy boundaries.
Convenience is valuable. It should not come at the cost of judgement, understanding or wellbeing.
A Reflective Question for You
Which parts of your working life have automation genuinely made lighter, and which parts may have quietly become more demanding because of it?