They would not like the idea of replacing them with trust even if it is supported by thorough testing.
This has to be taken into account by leaders who wish to bring Data Science and, especially AI, into their businesses.
Machine learning will most likely follow and it may find resistance internally when people would not trust algorithms and would want lots of manual controls undermining the whole idea of AI automation.
Demanding explicit rules is just a matter of habit and tradition in business operations that emerges from the time when business rules were executed by humans.
Today, business operations are largely automated and the only reason they are written in the human-readable format is that they are designed by humans.
With the adoption of Machine Learning, business rules don’t have to be human readable any longer.
It will be just a matter of trust for businesses to give up demanding fully explainable AI and have more confidence in testing.
The remains of the Roman Forum, the birthplace of the modern civilisation.
These structures were built without structure analysis theory.
They were built on an architect’s practical experience and best practices of the time.
Image by jacqueline macou from PixabayThere are many other important fields, where practice comes before theory, and where people used to trust tests before they have answers to all their whys: civil engineering, the pharmaceutical industry, etc.
Civil Engineering, for example, was based on architect’s practical experience for thousands of years before theoretical models of structural analysis were widely adopted in the 20th century.
At the end of the day, neuroscience still can’t explain how our brain makes decisions.
SummaryMachine learning is very closely tied with Data Science and AI, which means its implications must be considered too.
These implications may be challenging for established business culture when AI and Machine Learning are introduced into business operations:Machine Learning is not rule-based and therefore traditional business rules will not work in solutions, based on Machine LearningMachine Learning is example-driven.
In order to “train” an algorithm to behave in a desired way, a business must provide a set of relevant data examples from their real practice.
Most of Machine Learning algorithms lack the transparency of business rulesMachine Learning will not make Business Analysts redundant, but it will make them differentThese implications will naturally cause a feeling of losing control over business operations when using Machine Learning.
The next natural reaction in business would be compensating this loss by putting multiple manual ‘old-style’ controls undermining the whole idea of AI automation.
The best way to avoid that is by developing trust in Machine Learning solutions through careful and diverse testing.
This is hard work for both Business and Data Scientists, especially in the beginning of AI journey, but this needs to be done for successful adoption of AI and Data Science in businesses.
Like a few decades ago with the introduction of computers to workplaces, successful introduction of AI automation into business operations will not pose a threat to Business Analysts.
It will make them different and more productive… eventuallyYou can find me on LinkedIn, Twitter, Facebook.