The fact that AI is gaining steam and being leveraged by diverse organizations to create new business models cannot be disputed. 2017 saw unprecedented progress with AI in real-world applications such as healthcare diagnoses, predictive maintenance, customer service, digital oil fields, automated data centers and smart homes. In a short span of time, the effectiveness of Machine Learning has tremendously improved owing to its ability to quickly learn across domains. For example, our email spam filters now classify emails quite accurately and are continually learning based on the inputs we provide.
There’s no doubt that organizations can gain significant competitive advantage through the use of AI – however, I want to emphasize the need to have a structured approach to using AI. I suggest identifying the opportunities that will deliver the maximum returns for your enterprise without disrupting your existing operations.
Harder to get the elephant to dance
Traditional, large enterprises find it harder to transform due to their legacy foundation, unlike new businesses that are born digital and hence much nimbler. A measured approach to identifying automation areas requires one to consider the business value while being practical about the change management needed during the transition. It is natural for enterprises on the AI adoption track to aim for differentiation. However, it is important to balance it through a calculated selection of areas that do not disrupt the operations during the transition. Another option is to start with the low-hanging fruit rather than go all-in.
Impactful AI needs to be business process contextualized
While components of AI are widely available, a use-case backward approach ensures integration with overall business process and ease of adoption. I cannot over-emphasize the importance of data and domain knowledge to get the desired accuracy levels. Data powers AI, as the software learns from examples. Focusing on specific applications of AI (e.g., extracting critical clauses from enterprise contracts, payments anomaly detection or automating KYC process) can hence help the organization get more accurate results as data volumes increase.
For instance, anti-money laundering processes (AML) in a bank require the Know-your-customer (KYC) analyst to review documentation received from customers, evaluate high-risk cases through due diligence and independent verification of customer’s information. The analysts are trained and know how, and where to look for information, how to interpret it and arrive at results. This requires knowledge of country-specific standards (such as Banking Secrecy Act, FinCEN requirements, etc.) in addition to risk assessment policies and procedures. Through the application of cognitive automation, this effort by the maker has been significantly reduced – in addition to supplying evidence and bookmarked source documents to the checker for verification. In the initial stages, high accuracy is guaranteed through a human in the loop.
AI aids knowledge workers; doesn’t replace them
It is important to note that new technologies will augment the humans, making them more productive and efficient. In a business process, only a subset of all the tasks will get automated, and the remaining will continue to involve human value-add. In the KYC process example mentioned above, information search, aggregation, and recommendations are automated. The human makes the decision based on this information and can over-ride the system’s advice if needed. The objective is to reduce the human effort while improving the cycle time for end customers.
These are exciting times, and as the capabilities of AI get stronger, it is certain that organizations and executives that constantly scan and thoughtfully apply this technology to business operations will have an advantage over those that don’t.
