Mohamed Aamir

Taking into consideration a hypothetical situation similar to what was discussed during class. Let us take into consideration a Loan approval process where AI processes applications and follows a rule-based approach, all applications above a specified threshold based on say payback ability / solvency or collateral fulfillment of over 40% would be approved any applications which fell below that threshold would be rejected. In this case, it is tricky for the AI to balance high customer satisfaction (typical human psychology is when something gets rejected, they are not satisfied) and following process and protocol and thus must strike the balance between contradictory priorities. Possible solutions for the AI to handle this while increasing or improving CSAT would be: 1. AI is allocated high compute bandwidth and also opens additional dialogue with customers who fall short withing 5% of the threshold (35-40%) for instance, where customers can present additional documents to support their case and add other collateral assets which could result in a quick approval despite not being within threshold. 2. If customers are too far from the threshold, the AI could respond quickly informing them of the discrepancy and give them options of how they could meet the threshold or improve their score to be approved for the loan, this would ease the burden of rejection from the customer/client resulting in improvement of the CSAT while striking a balance with protocol/policy. Overall, as a result of either or both of the options training the AI would these sort of options would then help balance Process vs Customer satisfaction as even in the case of rejection the customer gets potential guidance on what can be done to obtain a future approval thereby taking the role of a humanoid customer relationship agent who is trying to help them rather than present a flat out rejection when criteria is unmet.