Himanshu_Lohani_WpY8

I am using a scenario from Insurance Broker F&A (Surplus Lines Tax Filing) In an Insurance Broker Finance environment, prioritization is critical in Surplus Lines Tax filing. Every month or quarter, thousands of policies must be reviewed and taxes filed across multiple US states, each having different regulatory deadlines and penalties. Current Practice Traditionally, teams prioritize work based on: Filing deadlines communicated by leadership Escalations from brokers or account teams State regulators known for strict penalties Manual experience of team leads For example, a team may prioritize California or New York filings first simply because they are large-volume states or historically sensitive. However, this prioritization is often experience-driven rather than data-driven. How AI Could Recommend a Different Priority If AI analyses historical data across filings, it may evaluate: Probability of filing delay Historical penalty amounts by state Complexity of policy structures Volume spikes near deadline Past error patterns The AI might recommend prioritizing Illinois or Texas filings first, even if their deadline is later, because: Those filings historically take longer to reconcile Missing documentation causes frequent rework Late filings in those states have produced higher penalty exposure So while leadership may instinctively push high-volume states, AI may recommend high-risk states. Resolving the Conflict: A Practical Decision Framework In such cases, the answer should not be AI vs Human judgment. Instead, the decision should follow a structured prioritization framework. Step 1: AI Provides the Risk-Based Priority AI should generate a prioritization score based on factors like: Regulatory penalty risk Expected processing time Documentation completeness Probability of delay This creates an objective risk-weighted work queue. Step 2: Human Leaders Apply Context Operations leaders should review the AI recommendation and consider factors AI may not fully capture, such as: A broker relationship requiring urgent support A regulatory audit currently underway A recent policy change affecting a specific state These factors may justify overriding AI in certain cases. Step 3: Governance Rule for Final Decision The best model is: AI recommends → Human validates → Governance rules decide A simple rule could be: If regulatory or financial risk is involved, AI risk scoring takes priority. If client relationship or strategic urgency is involved, leadership judgment may override AI. The override must be documented, ensuring transparency and continuous improvement of the AI model. Who Should Have the Final Say? In a regulated industry like insurance, the final decision should rest with the Process Owner or Operations Leader, but the decision must be evidence-based and supported by AI insights. AI should guide prioritization through data, but humans must retain accountability for compliance and client impact. Final Insight In Insurance Broker F&A operations, AI should not replace leadership decisions — it should challenge instinct-based prioritization with risk-based intelligence. The winning model is not Human vs AI, but Human judgment informed by AI-driven prioritization. Organizations that build this balance will reduce compliance risk, improve filing accuracy, and ensure that priorities are driven by impact rather than noise or hierarchy.

When AI Enters Finance Operations: Do We Still Need Separate Ops and Analytics Teams? In most Finance & Accounting (F&A) Global Capability Centres today, work is clearly divided: Operations teams process invoices, reconciliations, collections, journal entries. Analytics teams build dashboards, explain variances, and generate insights. Operations execute. Analytics interprets. It’s a clean handoff model. But once AI becomes deeply embedded in workflows, this separation starts to break. What Changes When AI Is Fully Integrated? Imagine this inside your ERP system: Invoices auto-classified Journals auto-drafted Variance commentary auto-generated Anomalies flagged instantly Cash flow predicted in real time Now ask yourself — if the system is already generating insights while executing transactions, where exactly does Operations end and Analytics begin? The answer: it doesn’t. The boundary starts collapsing. Execution and Insight Happen Together In an AI-enabled F&A setup: The person reviewing a transaction will simultaneously see: Risk probability Suggested root cause Predictive impact Recommended action The workflow itself becomes analytical. This means the Operations Manager can no longer say: “Analytics will interpret this later.” And the Analytics team can no longer say: “We just report what Operations sends.” AI connects both in real time. What Will the New Structure Look Like? Instead of separate teams, we’ll likely see AI-augmented value pods. Each pod may include: A process owner An AI/data steward A control specialist They won’t work in sequence; they’ll work together around a shared AI engine. The focus shifts from, managing volume to governing intelligent systems Quality Also Changes Today, Quality teams audit samples after the work is done. With AI: Errors are flagged before posting. Risk scoring happens instantly. Exceptions are predicted, not just detected. Quality moves from inspection to AI governance. That’s a big shift. Business vs Technology Will Blur Traditionally: Business defines requirements. IT builds the system. Operations uses it. But with AI: Prompt design affects financial output. Data structure affects risk exposure. Model tuning affects compliance. You can’t separate Finance from Technology anymore. Future leaders will need to understand both. So What’s the Real Boundary in the Future? It won’t be: Operations vs Analytics. It will be: Autonomous AI vs Human Judgment. AI will handle: Volume Pattern detection Drafting Predictive alerts Humans will focus on: Material decisions Ethical oversight Risk governance Value optimization Final Thought In AI-enabled Finance operations, the question will no longer be: “Who owns the data?” The real question will be: “Who owns the intelligent system that is producing the outcome?” The organizations that redesign around AI-integrated teams, instead of protecting old functional boundaries will move faster, operate leaner, and make better decisions. The structure won’t just evolve. It will reorganize around intelligence itself.

I am considering my own service as an example here, Role: Senior Policy Review Analyst – Surplus Lines Tax (Insurance Operations)In surplus lines insurance operations, a Senior Policy Review Analyst traditionally reviewed 70–100 page policy documents, verified insured details, interpreted endorsements, determined taxable states, calculated surplus lines tax, and prepared regulatory filings. The role rewarded reading speed, memorization of state rules, and manual accuracy. Today, AI tools extract insured data, identify endorsements, flag multi-state exposure, calculate preliminary taxes, and even draft filing summaries. AI handles pattern recognition and first-level analysis. The human role shifts from processor to decision owner. These changes hiring criteria fundamentally. 1. Judgment Over Processing SpeedEarlier, speed and accuracy in reading dense policies were critical. Now AI extracts key clauses instantly. What matters more is: Ability to validate AI output Recognizing when AI is technically correct but regulatorily risky Making override decisions in ambiguous scenarios Revised Hiring Test: Provide an AI-generated tax computation with subtle regulatory errors. Assess whether the candidate questions it and explains the risk impact. 2. Framing and Problem StructuringAI responds to how a problem is framed. Poor framing produces incomplete analysis. The modern analyst must: Break down exposure logically Provide precise instructions to AI systems Define assumptions clearly Revised Hiring Focus: Evaluate structured thinking through scenario breakdown exercises rather than testing memory of tax tables. 3. Exception Handling CapabilityAI performs well in standard cases. Edge cases create compliance risk. Hiring should prioritize: Comfort with ambiguity Ability to investigate anomalies Escalation judgment Instead of asking, “How many years have you processed filings?” ask, “What would you do if AI flags Texas exposure but stamping fee logic conflicts with policy wording?” 4. Risk Ownership MindsetIn regulated environments, mistakes have financial and legal consequences. When AI contributes to analysis, accountability becomes shared. The new hire must: Think like a compliance custodian Understand downstream financial impact Take ownership rather than rely blindly on automation Behavioural interviews should assess ethical judgment and responsibility under uncertainty. 5. AI Collaboration Literacy (Not Coding)The role does not require programming skills. However, it requires: Understanding AI limitations Awareness of hallucination risks Ability to critically review automated drafts AI fluency becomes more important than manual endurance. What Becomes Less CriticalMemorization of state tax percentages High manual processing volume Mechanical reconciliation repetition Pure tenure-based seniority Experience still matters, but depth of reasoning outweighs years logged. Capability Shift SummaryBefore AI After AI Data Processing Oversight & Validation Memorization Interpretation Speed Judgment Task Execution Risk Governance ConclusionWhen AI handles part of the thinking, we no longer hire for who can process the fastest. We hire for who can challenge the machine intelligently, frame problems clearly, and own the risk of final decisions. In surplus lines tax operations, the future-ready analyst is not the best data reader — but the best decision reviewer.

We have an AI with human in loop automation where AI performs activity of extracting data from documents as per a pre defined rule. If the AI is unable to extract data because of unavailability that particular case moves to exception queue to be managed by Human. We have developed 2 sets of metrics to evaluate the performance of individuals post the AI implementation, If the AI is able to extract all data points and marks the case as completed, we perform a 30% sample to ensure that the AI is performing accurately. The expectation from the processor is to identify if there was any assumptions or in correct processing by the AI. If the AI marks a case as fail, the expectation is for the processor to identify if it was correctly marked as failed or the data was available but AI was unable to extract the same. In this case the user will need to assess if the error was due to rule building or an actual failure. In summary, the measure should be: How well were AI recommendations evaluated and applied? And Include: Override rationale quality Exception handling accuracy Risk mitigation effectiveness This is purely basis the experience I have on the AI implemented within our process but I will be keen to hear other perspective to expand my horizon.

In my experience and a few AI project I have worked, AI tools are very good at providing summary of a content and most of the time is agreeable. If a process follows checklist used by humans, then AI can do the same. Most of the time these processes are rule based, follows repeatable patterns and has structured input method. In my personal experience, we have been able to implement an AI solution integrated with API's where the AI is extracting data from document and feeding the data into a workflow tool for further processing. Before deploying this solution, we had to perform an extensive testing by uploading various format of the document and building the prompt to be able to capture accurate data. Hence, it is imperative for us to provide AI with large clean data and set accurate rules and test the output for multiple scenarios to develop trust on the output from AI.