Artificial Intelligence and the germs of major project risk

In 1721, the average age of death in UK cities was about 32 years. Doctors still did not know that the underlying cause of many sicknesses was germs, and they had no understanding of, and little answer to infection.  

In the 18^th century, there were many theories about infection, and some herbs and natural remedies were effective some of the time, and basic practices (like cleanliness) reduced the risk of infection.  

So too do current risk management methods contribute to a reduction of poor project outcomes.  They are good practices, but not very effective. The performance of projects continues to be poor, and current risk management methods do not overcome the underlying “germs” which drive poor outcomes.  (Indeed, some practices like Monte Carlo analysis of forecast costs and/or time are about as useful as the 18^th century practice of “touching the king” to cure skin disease.)

Want proof? Look at the statistics of major project performance.  Nine out of ten major projects will exceed their timeframes, and budgets.

Louis and Marie Pasteur laid the foundations for antibiotics in the 19^th century, but it was not until 1928 that antibiotics were developed.  That changed whole game when it came to treating infection.

This is where we are in the evolution of risk management.  Artificial Intelligence is the antibiotic to the germs of risk, it is a game changer.  Its impact will be significant and rapid, and projects will be healthier because of it.

We are on the threshold of the most disruptive change ever to impact risk management practice.  This is a common topic of discussion with many commentators claiming insights. But with a very few exceptions, the insights are not based on evidence from functioning AI innovation, because there is very little functioning AI for project management.  

Our work using an artificial intelligence system (Octant AI) demonstrates that machine learning is better at cost risk management than professional risk mangers or any current non-AI based system is.  What is more, the more it does it, the better it gets, and it does in seconds what takes people days or weeks to do.  

Typical performance metrics:

• 200 million data points,
• 1000 variables
• 10 seconds
• 200% performance improvement

Readers of this article may be sceptical about the achievement, and I encourage that.  As with any new technology, there will be many who claim AI capability whose systems are either not AI or ‘aspirational’.  There will be no shortage of “snake oil salesmen”.  I urge those tasked with risk management consider three things:

1. Learn about AI.  Go to courses, listen to webinars, understand its limitations and open your minds to its potential in major projects.  Focus on AI that is applicable to risk management (as opposed to that which is applicable to construction efficiency for example).
2. Be sceptical of promises.  AI in projects is very new, so there are not be many mature systems.  It is not easy to get right, and there is a danger of not understanding how the machine reached its conclusions.  You can expect there to be innovation challenges and you will need to invest your company’s time and money if you want to be ahead of the disruption. But if possible, it is better to embrace systems that are already working and have solid world class technology capability and deep domain knowledge.   
3. Data is critical.  Huge amounts of data has been created by risk management processes, and to a large extent this data is a stranded asset.  Current methodologies simply cannot leverage very well it in large complex and dynamic project environments.  A fundamental part of AI performance is the way it uses data, and it does not do this in the same way as humans do. This means that the data asset must be considered in a different light to our current paradigm.

Risk management practice will need to adapt to a new “socio-technological” paradigm which may be challenging, and the way things will play out is certainly uncertain.  Adapting may not be easy and our work to date provides and insight into how disruption may emerge. However, now is the time to become involved.   Those who do are more likely to adapt and benefit.