When Automation and AI Are Not the Right Solution

Many organisations are under pressure to introduce automation and AI. New tools appear constantly, and leaders are often encouraged to apply them widely across a business.

In practice, some work benefits greatly from automation. Routine and rule based processes can often be completed faster and with fewer errors.

Other activities are very different. They depend on judgment, context, relationships, or creativity. When automation is applied to these areas, it can create new risks rather than improvements.

Understanding where automation is unsuitable helps organisations make better decisions about where technology should and should not be used.

Processes requiring human judgment

Some office processes depend heavily on interpretation and context. The available information may be incomplete, and the outcome may depend on balancing several competing considerations.

Automation performs best when tasks follow clear rules and produce predictable outcomes. When these conditions are absent, automated systems struggle.

Research by Mark Torres into automation decision frameworks shows that tasks with low standardisation and difficult to verify outcomes are poor candidates for automation.

Examples include performance reviews and hiring decisions, legal interpretation and compliance decisions, and strategic planning or resource allocation.

Automation tools may assist with data analysis or information gathering. Final decisions still require people who can interpret context and weigh competing priorities.

High variability and exception heavy processes

Automation systems rely on stable processes. They work best when the steps rarely change and exceptions are limited.

Some organisational activities look structured on the surface but involve frequent variation. These processes often contain unusual cases, unexpected inputs, or changing requirements.

Attempting to automate this type of work can lead to brittle systems that break whenever the situation changes.

Automation guidance highlights that processes which change frequently or occur infrequently are rarely worth automating. Designing the automation can take longer than performing the task manually (Medium).

Examples include:

• crisis management and operational incident response,
• one off projects or irregular executive requests,
• unusual customer service issues or complex escalations.

Automation may handle the routine part of a process. The remaining situations still require people who can interpret the problem and decide what to do next.

Work centred on relationships and trust

Many organisational activities involve human relationships. These interactions depend on empathy, trust, and emotional understanding.

While software can automate routine communication, it cannot replace genuine human connection.

Analysts note that automation is effective for repetitive tasks but cannot replicate emotional intelligence or interpersonal understanding (TechTarget).

Examples include:

• resolving customer complaints,
• mentoring employees or supporting career development,
• negotiating with clients or partners.

These activities depend on subtle signals such as tone, hesitation, or body language. They also rely on building long term trust.

Automation may assist with information or preparation. The interaction itself usually needs to remain human.

Creative and ideation driven work

Creative work requires generating new ideas, exploring alternatives, and making subjective judgments.

Automation tools can analyse data or produce drafts. They are far less effective when originality and interpretation are required.

Creative activities often involve combining insights from different areas, questioning assumptions, and experimenting with new directions.

Examples include product and service design, marketing concepts and brand development, and research and development planning.

Industry analysis notes that automation can support creative processes but cannot replace the human thinking involved in generating original ideas (ProcessMaker).

Organisations often benefit from using technology to assist creative work. However, the creative decisions themselves remain human.

Ethical or high impact decisions

Some organisational decisions directly affect people’s rights, livelihoods, or opportunities. These decisions carry ethical and legal responsibilities.

Automation can introduce risks if the reasoning behind a decision is unclear or if bias appears in the data used by the system.

European data protection authorities warn that automated systems used for decisions such as recruitment, credit approval, or social benefit eligibility can introduce opacity and discrimination risks (European Data Protection Supervisor).

Examples of sensitive decisions include:

• hiring, promotion, and disciplinary actions,
• financial approvals or credit decisions,
• access to services or benefits.

Automation may assist by providing information or identifying patterns. Human oversight remains necessary to ensure fairness and accountability.

When errors are costly

Some tasks appear suitable for automation but carry a high cost of error. In these situations, the risk introduced by automation can outweigh the potential efficiency gains.

This often occurs when:

• the outcome is difficult to verify automatically,
• errors could create legal or financial consequences,
• the task happens too rarely to justify automation.

Automation frameworks describe these tasks as having low verifiability. When mistakes are hard to detect or expensive to correct, full automation is risky.

Examples include:

• regulatory filings and legal documentation,
• handling sensitive financial or compliance data,
• rare audits or special investigations.

Automation may still support parts of the process. Human review often remains necessary before final submission or approval.

Where organisational knowledge matters

Some work relies on tacit knowledge built through experience.

Employees develop an understanding of how the organisation operates, how decisions are made, and how stakeholders behave. Much of this knowledge is informal.

It may include:

• awareness of internal relationships,
• understanding of organisational culture,
• experience with previous decisions or projects.

These insights are rarely documented in a structured way. They develop through observation and experience.

Automation systems cannot easily replicate this type of knowledge. Decisions that depend on organisational awareness require experienced people who understand the context.

Choosing automation carefully

Automation and AI are powerful tools when applied to the right problems. They work well when tasks are repetitive, rule based, and clearly defined.

Many organisational activities do not meet these conditions. Work involving judgment, relationships, creativity, ethics, or unpredictable situations often remains better handled by people.

Introducing automation into these areas can increase risk, create confusion, or reduce trust.

Organisations benefit from identifying which processes genuinely suit automation and which should remain human led.

Clarity about how work actually happens makes these decisions easier. Once the process is understood, it becomes clearer where automation may improve outcomes and where it may not.