Where AI excels (based on personal experience)

 

• Detecting patterns across enormous datasets: AI is exceptional at finding weak signals across millions of records. We’re not. Why would we NOT use AI?
  
  
• Consistency: We get tired, have limited cognitive bandwidth/attention span and are distracted. AI applies the same logic every time and can stay focused 24/7.
  
  
• Analysing information at speed: Tasks that consume days of human effort can be completed in minutes.
  
  
• Connecting information across silos: My clients are using AI to simultaneously review deviations, CAPA's, complaints, audit findings, batch history and annual product quality reviews looking for patterns and trends. Quality oversight has never been better.
  
  
• Remembering everything: Humans forget. AI does not. Every previous investigation, procedure, inspection observation and historical trend remains available for analysis.

 

 

Five applications where I’ve seen AI outperform people:

 

1. Deviation investigations

Instead of investigating one deviation at a time, AI can identify recurring systemic causes across hundreds or thousands of investigations.

 

Result: Less symptom treatment and more prevention. How can you justify NOT doing this?

 

2. CAPA effectiveness
AI can detect whether CAPA's are genuinely reducing recurrence or simply creating paperwork that appears compliant.

 

3. Batch record review
I’ve used AI to identify inconsistencies, missing information and unusual relationships across extensive documentation significantly faster than manual review. Human oversight will always remain essential, but AI can do a lot of the grunt work

 

4. SOP simplification

AI can rapidly identify duplication, conflicting requirements and unnecessary complexity across hundreds of procedures while preserving regulatory intent.

 

Simpler systems =  fewer human errors. Why would you NOT want to reduce errors?

 

5. Predictive quality risk management

Rather than waiting for failures, AI can identify combinations of weak signals that frequently precede deviations, complaints or inspection findings. 

 

This moves quality from reactive investigation toward proactive prevention. Not using AI is indefensible.

 

 

So, in conclusion: The real risk may notbe AI

 

The pharmaceutical industry has spent years asking:

 

"What could go wrong if we use AI?"

 

Although an important question, should we not be asking

 

"What could go wrong if we don't?"

 

If two companies manufacture the same medicine, one uses AI to detect emerging contamination risks weeks earlier. The other relies solely on retrospective manual review. Which organisation is applying the best available science? Which is better protecting patients? Which will regulators eventually expect to become the norm?

 

 

Could failure to adopt AI become a compliance issue?

 

Not any time soon but regulatory expectations evolve. Paper batch records became electronic. Manual trending became statistical process control. Quality Risk Management became expected rather than optional. Continuous process verification became increasingly accepted.

In my 44 years experience every tech transition follows a similar pattern

 

• Early adopters were viewed cautiously.
• Good practice became accepted.
• Eventually, not adopting the improved approach required justification.

 

Will AI follow the same trajectory?

 

 

 Questions for your leadership team

 

• If AI demonstrably reduces risk in a validated, properly governed application, what is our scientific justification for not implementing it?
• Are we managing regulatory risk or are we allowing fear of regulatory interpretation to become a barrier to scientific progress?
• Five years from now, which decision is more likely to be criticised: adopting appropriately governed AI or failing to adopt technology that could reasonably have prevented quality failures?

 

Look, I know the title of this article is hypothetical but the principle behind it is not. The future may not belong to the organisations with the most AI. It’s more likely to belong to those who can demonstrate they have applied the best available science, managed the associated risks, and improved protection for patients.
Perhaps the most important question is no longer "can we trust AI" but "can we continue justifying decisions that ignore it?"

 

I just wanted to get everyone thinking from a different ‘frame of reference’ (a uniquely human quality!)
Let me know what you think!