Is this right or wrong? The idea that business advisers using AI to find intelligence worth finding?
AI is best at automating predictable, repetitive and simple decisions. They may be of critical importance like medical diagnosis but that follows a simple decision tree branching yes/no process. Prognosis is where it gets more complex and the health professional takes over.
Most insurers access only a small percentage of the data they collect- maybe 20% if one is optimistic. Will AI help access, join and analyse the remaining 80%? Not without the data. No data, no model.
In Fraud Analytics, Detection and Prevention AI is still woefully short of all the data it needs to detect fraud, hence the high 25% white noise. AI rarely gets the volume of data needed to effectively train the AI to a model that would work. Even the simplest of image detection processes could take 72 hours to run over a daily image workload until the model is stable enough to be reliable. This could take months if not years.
If AI today were so ubiquitous then police forces, HMRC in UK and IRS in USA would have no need for investigators. They would just apply AI and track all criminals, tax evaders.
In fact it is more nuanced than that as they use investigative tools like 360Retrieveto access, join and analyse data. That is not enough as this must be combined with the enormous capabilities of sceptical (of human nature) investigators to test assumptions, test intuition with a mix of technology and human intuition.
Maybe part of the problem is the term AI. This seems to cover a continuum
from application of simple algorithms, through machine-learning and RPA to full blown deep-learning and an increasing and unfathomable and automated decision making.
To add to the mix is the constraint of GDPR- data supplied by individuals can only be used for the purpose agreed. Google and the NHS have already got in to deep water by forgetting this in the application of deep-learning to NHS patient data (even though it was anonimized).
AI is a servant of the business and not a leader. It has inbuilt bias from the human originators of algorithms and from the rules specifying deep learning processes.
And as always data is at the heart of this argument. Without the breadth and depth of data AI is a dangerous tool. Master Data Management is as important as ever.
Nicholas Taleb's "Black Swans" should teach us that the improbable is the tipping point for many a disruption. The improbable is hard to predict and AI may be a barrier to spotting black swans before they land.
It is a beguiling idea but is it putting the AI cart before human intuition? : -
"If business advisers can further what is right, to get insurers to more easily and expeditiously do the right thing before challenges arise, the insurance industry will benefit as a whole. If AI can substantiate what a business adviser recommends, if that recommendation is brief yet bold, if that business adviser can prove his recommendation is right, an insurer can succeed."
Lewis Fein Insurance Thought Leadership March 29th 2019
More to the point, business advisers are an independent class—hence their advisory role—in which they do much more than translate data into reports. They use artificial intelligence (AI) to find intelligence worth analyzing. They use intelligence to advance wisdom, because it takes skill to convert ones and zeros into a message that is as concise as it is compelling; it takes a different class of advisers to actualize a future that is close but hard to see; it takes verbal facility and visual acuity to present the future—to make the future present—for the insurance industry.
