23 February Blog

AI in Cybersecurity Is Timely — But People Still Decide Its Success

​There is no doubt that now is an excellent time to consider upgrading your cybersecurity capabilities to include AI. Threat actors are already using AI to scale attacks, automate reconnaissance, and personalise social engineering at pace. Organisations that fail to evolve risk being outmatched.
However, while AI can significantly enhance detection, response, and efficiency, technology alone does not deliver resilience. The real question leaders should be asking is not “What can this AI solution do?” but “How will this solution work with our people?”
Before investing in any new AI-enabled cybersecurity solution, there are human-centric metrics that are just as critical as technical capability.


1. Usability Under Pressure
Metric: Can staff use it correctly when it matters most?
Cyber incidents rarely occur in calm conditions. If a solution is complex, noisy, or confusing, people will bypass it or misuse it—often unintentionally increasing risk.
Ask:

• Can non-technical staff understand alerts and actions?
• Does it reduce decision fatigue, or add to it?
• Is the interface intuitive during high-stress incidents?

A solution that works in theory but fails in practice offers little real protection.


2. Behavioural Impact
Metric: Does it positively influence staff behaviour?
The best security tools quietly reinforce good habits rather than relying on constant policing.
Consider:

• Does it encourage safer decision-making?
• Does it reduce risky workarounds?
• Does it support, rather than undermine, productivity?

If people see the tool as an obstacle, they will find ways around it—AI or not.


3. Trust and Transparency
Metric: Do people trust the AI’s decisions?
Black-box AI creates scepticism. If staff don’t understand why something is flagged, they are less likely to act on it.
Look for:

• Clear explanations of AI-driven decisions
• Confidence scoring or rationale, not just alerts
• The ability for humans to challenge or override decisions when appropriate

Trust is essential if you expect people to act decisively.


4. Cognitive Load Reduction
Metric: Does it make work easier, not harder?
AI should remove noise, not create more of it.
Evaluate:

• Does it reduce alert fatigue?
• Does it prioritise what truly matters?
• Does it help staff focus on judgement, not data sorting?

Human attention is a finite resource—protecting it is a security control in itself.


5. Cultural Alignment
Metric: Does it align with your organisation’s culture?
Security tools send signals. They communicate whether the organisation values:

• Control or empowerment
• Blame or learning
• Surveillance or support

AI solutions that feel intrusive or punitive can damage trust and culture, which ultimately weakens security posture.


6. Training and Enablement Requirements
Metric: How much effort is required to make people effective?
An AI solution that requires extensive retraining, constant tuning, or specialist knowledge may struggle to scale.
Key considerations:

• Time required for onboarding
• Ongoing learning demands
• Dependency on a small number of experts

Sustainable security works with the realities of time-poor teams.


7. Psychological Safety During Incidents
Metric: Does it support people when things go wrong?
When incidents occur, people need clarity and confidence—not fear.
Ask:

• Does the solution guide response, or just raise alarms?
• Does it help teams learn after incidents?
• Does it reinforce accountability without blame?

Resilience is built through learning, not punishment.


AI Is a Force Multiplier — For Better or Worse
AI can dramatically strengthen cybersecurity, but it amplifies whatever environment it is deployed into. In organisations where people feel supported, informed, and trusted, AI becomes a powerful ally. In environments where tools ignore human realities, AI can simply accelerate failure.
The most effective cybersecurity investments are those that recognise a simple truth:
Cybersecurity is ultimately a human system, supported by technology—not the other way around.
As you assess AI-enabled solutions, measure success not just in detection rates and dashboards, but in how well your people can engage with, trust, and sustain them over time.