Shailendra Rai

Demand forecasting is a very crucial aspect of improving the resource allocation, product development and strategic planning aspect of an organization. Traditionally and currently these things are happening with data driven approach where experienced personnel are working with the data, market research and trend analysis to forecast the future business aspects and take certain assumptions. An AI solution can add a signification value to this work and make system more robust considering the AI ability to analyze the data and learn from the historical trends , it can also bake-in the decisions and high-low aspects on the seasonality, market shifts and Realtime issues which typically missed in traditionally forecasting. Further the forecasted value such as revenue, EBIDTA, risk and compliance attribute become the input to scenario-planning exercises. Considering the domain on I work, insurance. Forecasting the premium with the changing market landscape and competetion is a very tough exercises and very complex experience with the underwriters to arrive to an optimum score which provide reasonable price of premium to clients as well value to stakeholders. AI solution can run in waste amount of data and can assess thousands of the past underwriting contacts to obtain meaningful information and help underwriters to manage risk and provide a balanced approach to take decisions. ML algorithms can learn from our historical records and provide a better baseline with customized approach for each client, their product types, and geographies. AI Solution can also expand the horizon of research to next level and provide meaning full insights which can be overlooked by people Benefits of integrating the underwriting work with an AI solution, Balanced inputs - AI models can continuously learn and improve their predictions based on new data, leading to more reliable inputs for decision making Customized solutions - AI systems can provide scenario-based output for better alignment for org level requirements . Market Research – market research can be very easy, give the best available AI LLM and models. Cost efficient – Underwriting is an expensive skill and it can reduce lot of ad-hoc work for UW teams

At present AI innovation is in its initial stage and it works as a support agent or catalyst to provide reasoning, summarize inputs and search best bit solutions to the problems. AI can help on financial planning and wealth management where innovation can strike with basic management of products across the market and can provide best fit to the users. As an example, if we are going to buy a insurance for self, the market is flooded with many options and offerings. Here an AI solution can take inputs from users covering their financial and spending limits, capture benefits and look for the best available options, this way an AI solution can be build and drive better outcomes for the end users as well as the company so right match as per user requirements and increase revenue. Specially the investment advisor to an extant providing the return forecast and initial investment estimates with customs inputs for each segment. Also, it can summarize the offering taking out the hidden terms and conditions which user many do not want to ops for. AI solution can make a lot of innovation to segment and match the right offering to the right party. Gradually once user provide more inputs further this solution can assist users on risk assessment, future wealth planning and suggesting right instruments for investment with risk and time horizon for ease of making decisions. Here decision part still be taken by human as AI cannot take the decision on someone’s behalf , considering regulation, data anomalies and changing market dynamics. It can also provide an early heads-up for recent market shift, rate change and even policy change which one can miss. To ensure these AI-generated ideas are both creative and feasible, a detailed multilevel approach can be defined as below, Data and Analysis – Data should be considered from the source such as government and regulation bodies to avoid bias Input and Suggestion – Solution can provide user inputs to refine their decision with reference of the data Ideas and Innovation – Advance AI algorithm can assess large scale data and provide better input for solution and drive innovation User Behavior analysis – AI solution should investigate the pattern of user and based on the preferences provide best fit as per need and planning Human in Loop – The inputs can be validated by user to prevent any mistakes or data anomalies and train the model to refine future suggestion and ideas Prototype & testing – AI solution can run as pilot to check the feasibility and accuracy on the analysis and output. A comprehensive testing can refine the solution to a accepted level Regulations- This solution should callout as reference for decision making with an effective governance through regulatory councils This approach will ensure that the AI solution provided ideas and suggestion are guiding factors rather than direct advise, it will make the approach balanced and calibrated. Feasibility part will be taken care by running simulations and testing, in case of any mistakes or data anomalies occur. Also, the suggestion should be long terms inputs and avoid any Realtime or recent event to inflate the suggestions. A layer of analysis through AI on these areas can spark lot of innovation through test and experiment can provide an feasible solution.

I work in insurance domain, and I can relate very well with the early risk detection as key risk parameter considering fund inflow to organization, there can be instances of financial fraudulent transactions and money laundering. These aspects can impact organizational financial health, regulatory concerns, and brand impact. There are several early risk detection measures in place to handle such issues however they are dependent on the people and processes. A system driven approach can be more advanced with a layer of human evolution can make this more efficient and robust. Organizations have measures such as fund source, background check of the payer, payment pattern and many more parameters to scrutinize but it led to high efforts, delay in payment settlement and in case of legit transaction it breaches trust of client due to high validation measures. In the given scenario, we see a lot of manual effort and resource limitations, with the AI solutions one we can increase our sampling (maybe cover all transactions) as well as we can cover any new activity with different patterns, this can also be identified and validated further by human evaluation so it will not only reduce the effort but also it will scale our validation on preventing any falsified transactions to the organization. AI driven solution model can analyze all financial transactions, qualitative aspects as client profiling, detect pattern on submitted documents, and most importantly import the history and map each transaction with a new pattern every time. The solution can leverage machine learning algorithms and can detect irregularities and emerging patterns that human analysts might miss. Some examples such as, Unusual transaction patterns or frequencies ( Volume and timeline) Sudden changes in customer financial behavior (more risky investments, or investing in low ROI products) Connections between apparently distinct accounts and entities ( Statement anomalies, money rotation etc.) Sudden change in client organizational health and suspicious indicators (High stock rates, regulation, and penalties etc.) We must ensure that the preventive alert and flag are considered, following measures could be implemented, There should a risk score for each client or party, so the changes in scores can reviewed by risk team. High Risk cases are handled on priority and updated back with finding in system for future learning of the model Defining the alerts with context so system can be more efficient Focus on false positive cases to train system with feedback and investigations Make the system cross functional so it’s an Org wide practice and early detection is possible, may be at onboarding stage. Preventive alerts and flagging through AI system with human layer can be a better way of avoiding alarm fatigue.

An AI system working as MIS/WFM Analyst or Specialist would be a complex system as these are rule based to an extent. However, adjustments and balancing are the real work on scheduling and resource forecasting. There have been forecasting and scheduling models which are there through rule-based approaches such as macros, RPA and integrated platforms such as workday etc. These solutions need human validation to overcome real-time changes and uncertainty to non-forecasted events. The AI system, which is responsible for an efficient and fair system, requires a wide range wide range of factors beyond basic parameters such as shift, volumes and attendance inputs. An efficient and should consider input parameters such as, Employee skills, and a skill matrix on which skills are updated on a periodic basis to cover each employee for more coverage in forecasting and scheduling Employee historical performance and proficiency data, which will help produce more efficient allocation to business or line of businesses At least a month or 3-month Individual preferences and availability, keeping leave, events and operations planning Based on geographic compliance requirements and employment regulations. Caping of working hours, mandatory holidays and avoiding overtime. Volume visibility through integrated system on a real-time with categorization Seasonality on the product and services and historical trends and special causes Business growth aspect as business grows, employees become proficient, this yielded proficiency should be considered in forecast Employee demographic details such as gender, location and environmental aspects should be called out, specifically in urban/metro cities where environmental factors and transport issues are the key input to shrinkage. Employee One-on-one inputs such as family events, future treatment and life events should be called out and updated as potentially planned shrinkage, which should be considered in the forecast. Organizational training and growth programs are another big aspect to keep in mind as inputs at process level. The AI solution should be capable enough to manage the adjustment and customize the inputs as scheduling and forecast are subject to change as we move in the timeline. AI solution should have defined input parameters as below so model can be fine tuned with percentage increment on the input values. Listing below some critical aspects to consider, Prioritization of the top work types as per SLA, Client requirements and product needs Adding the offline data incase of system downtime, so forecast volume is adjusted Real-time volume adjustments based on the input volume Cross training employees to adjust as per increased volume in critical queues Shift adjustment as per real-time employee turnout with a specific interval wise allocation, it will keep productivity intact. AI system should allocate employees from ideal to volume queue as per the interval wise demand Recommending shift swaps to balance workload and employee preferences In case of low and high volumes, it should send alert to leadership for immediate intervention Another aspect to maintain efficiency and fairness, the system should be capable of collecting in-puts not only from the Forecasting and scheduling team but also from the employees, it will have many benefits and a feedback loop to keep the system interactive, Allow employees to input their scheduling preferences and constraints with approval with managers. Keep monthly feedback from employees on how efficient the system is along with recommendations Using systems such as randomizer or some other method to plan people on shifts, it will drive transparency For exception adjustments, please leverage HR/Employee representative inputs in these cases Link the system with employee scorecard for direct inputs. Shift rotation and late-night shifts should be based on employee preferences OT, night allowance etc. should be calculated for better transparency in the system Through this approach and input we can build an efficient and fair system so employees and organization both can be benefited.

Shared service case management process looks a potential example to me. The process operates with customers case setup, review, validation and closure of the case in multiple business areas with fragmented knowledge across the value chain. The bigger problem within businesses is that over the period business processes and areas are more focused on the in-scope work and lack the up and downstream linkages. This creates a fragmented knowledge base, and they never talk to each other, resulting in the key objective. Further if people want to know the overall case summary the dependency is another concern with higher lead time and inefficiencies in the value chain. The dependency on people to update the Knowledge base based on the changes in product and service is another challenge as the information chain is so long from product to services, in many instances it observed we are falling behind the market updates on our own product. Another aspect on the risk side, where incomplete or inaccurate information has been shared with customers and stakeholders and leading to business losses and impacting brands market presence. A small example a miss on communication templates are not being updated on time, which leads to serious regularity and compliance risk on the contractual requirements. An AI based solution could be a effectively handle the knowledge management in the organization, considering KM is broadly based on the input and content through a right channel, stored and accessible to the concern party on time with lates revisions and flags to help determine the business process. We can break the AI enabled solution into segments which will provide a complete view of KM across organizations, KM Assistant Bot – We would need an AI assistance UI through which all KM can be accessed through search and expand level with summarization, bullet and key highlights on product and its features along with recent updates and any flags. Central Knowledge Repository – Consolidate information from various sources such as policies, procedures, FAQs, etc. Also use natural language processing to summaries the content as per user prompt. Dynamic Content Generation – Based on large and complex documents can be simplified with underline sense and inputs for user to understand and take decisions. Tailor knowledge delivery to individual employees based on their roles, experience levels, and learning preferences. Interactive workflows – AI solution can provide a end to end mapping of the process flows and created an interactive workflow to help on completing their tasks. Validation from multiple sources – Since AI can read all information, it can provide with best suitable input for user and work as validation agent. Self- Learning module – Skill based self-learning module for the complex topic and underline knowledge assessments to keep the employees up to date on critical organizational aspects such as data security, ethical practices and workplace security etc. Customized training courses to individual employees based on their skills, roles experience levels, and learning preferences. Proactive Insights and Flags – Some time there are irrelevant, and incorrect information gets updated due to typo or missing user validation, AI agent can flag such information can flag for the user review. Feedback Loop – Knowledge base are very sensitive to updates and sometime updates are very frequent that we can to keep a reference from the KB, here versions can be tagged, and older version can be updated with user feedback so, appropriate version are updated and may be tagged with feedback score. Given these aspects will keep the organization ahead on the accurate and precise information flow across organization. It will enable the functions to collaborate and work closely with each other reducing cycle time and create efficiency and keep the organization ahead on serval business aspects.