amit.suri

Let's take a typical example of an AI-powered Workforce management system where it is identifying employees, who are giving higher customer satisfaction during peak hours, to the shifts during the weekends or late nights, despite employees submitted their preferences to avoid those times due to person commitments. This AI behavior may result in higher attrition cases in the process and may impact productivity and quality of deliverable. Instead AI system should submit proposals based on historical data and forecast needs.

Scenario: Multi-agent collaboration in Vendor Registration in Procurement Process In Procurement process, vendor registration is a key activity that includes multiple steps requiring accuracy, compliance checks and system integration. In the proposed solution, 3 specialized AI agents will collaborate to accomplish the required task. Agent 1: Vendor Integration Agent Responsibility: Collect initial vendor data through a secure web portal or a chatbot. Tasks: • Gathers vendor profile, contact info, product categories and banking details. • Guides vendor on required documents (e.g., GST, PAN, ISO certification, etc) • Ensures data is captured in a structured format and confirms completion. Agent 2: Compliance & Document Verifier Agent Responsibility: Validates submitted documents for authenticity and compliance. Tasks: • It will use OCR and ML to extract data from uploaded document. • Verifies legal/tax compliance (e.g., GST registration, etc.) • Checks documents validity, expiry dates and flag discrepancies. Agent 3: ERP Integrator & Workflow Trigger Agent Responsibility: Will be responsible for updating the internal procurement system and initiates approval workflows. Tasks: • Creates and updates Approved Vendor List (AVL) in ERP. • Notifies Procurement, Finance and Legal teams for approvals. • Triggers automatic onboarding emails and vendor codes. Challenges in Coordination • Standardized & Consistent Data: Ensuring clean hand-offs between agents without data loss. • Status updates: Keeping all agents aware of process status to avoid rework. • Error handling: Managing incomplete/invalid documents and re-engaging the vendor smoothly. Design for Smooth, Accurate, Explainable Interaction • Workflow Design: Workflow to be designed in such a way that all 3 agents work sequentially (1-->2-->3) • Central Database: Central vendor registration databased shared across agents. • Explainability: Each Agent will maintain its logs for traceability, audit and compliance. • Fallback Mechanism: If AI agent is unable to handle the query, it should flag it for manual intervention. The above design will ensure a smooth, accurate and explainable interaction in this agentic approach.

There are a couple of ways of doing this: The AI agent should be designed to provide reasoning for a particular response, for example, Referencing and analyzing six months' historical data. The AI agent must ask the human user for the suitability or correctness of the response. Human users must have a mechanism to validate the responses periodically and confirm if the response is correct and if not human user should feed in the feedback for AI to learn. Feedback for the exceptions and the right solution must be fed back to the AI agent to continually learn. I believe 2-way communication will help AI agent to earn the trust of human users.

Scenario Let's take an example of a Quality Process of a VIP client where Quality checkers have to follow the Quality Checklist to remain within the scope and timelines and escalate, any possible non-compliance which may have a higher dollar impact, to the Quality Manager for further investigation. Trade-off Following the checklist will ensure the Quality process remains within the scope and timelines; however, if any transaction gets escalated for further investigation, it may delay the process timelines but may help in detecting a possible major non-compliance AI Implementation AI should follow a risk-based approach (as per ISO 9001) to assign risk score, based on some rules, like, Possible defect impact exceeds a pre-defined financial threshold of $5000 Analyze the past NC data for high-risk patterns. Higher score for high-risk patterns. Analyze already deployed controls and check for their effectiveness. Low scores for weak controls. The Logic AI will follow the logic below: IF the transaction's possible defect impact >$5000 AND past NC risk pattern score is >5/10 OR the deployed control effectiveness score is <5/10, the transaction will be escalated to the QA Manager for further investigation. This way AI will logically decide which transactions require further investigation.

We can take example of Document Verification Process for employee onboarding. Candidates have to submit various documents, like, resumes, certificates, IDs, marksheets, etc. Company has implement AI-enabled document verification system using Computer vision, which includes OCR and Image Processing, and Natural Language Processing (NLP). Any mismatches in the documents, formats, etc, are called out as exceptions by tagging them as Incomplete document, Missing Document, Invalid Format, etc, and sent for human intervention. Designated HR official goes through the exceptions, captures the reasons and the required solution, which is again fed in the system for further learning. AI continually learns from the responses given by the designated HR and starts handling exceptions over a period of time on its own. The above set-up is an example of Supervised Learning as the documents are labelled and the exceptions are also categorized and labelled.

The Invoice Creation Process can largely be handled by an AI agent; however, it can also escalate if the criteria are not met, as agreed in the contracts. The whole process requires An AI agent to collect the monthly input. A form can be designed to collect the input data from the operations team, such as transactions processed, number of processors, number of working days, etc. Another AI agent will be created to fetch information from the Knowledge base, primarily the contractual requirements, CRs, amendments, pricing, etc. The first AI agent will collect the input and call the second AI agent to validate the input with the information stored in the Knowledge Base and take decision if Invoice to be created or an escalation to be made. The invoice will be created if the input matches all the contractual requirements. The escalation will be made to Process Owner if the AI agent detects a mismatch of any of the criteria. For e.g., COLA mismatch or FTE mismatch. Bringing AI into the Invoice Creation process will help avoid calculation errors, leakages, and compliance issues.

One rule where I would trust the AI would be to create the audit/assessment calendar and allocate resources accordingly and the one where I would not trust the AI rule would be to interpret the legal or statutory and regulatory requirements, as mentioned in the contracts, because of various nuances associated to Stat & Reg or legal interpretations.

Marking a situation as Major or Minor Non-compliance (NC) as part of ISO 9001 Internal Audits is one thing I feel is too human to hand over to AI. The reason for stating the same is because the auditor has to review and interpret the situation holistically after reviewing evidence, matching it against the standards, talking to the auditee, looking the intent, etc., and then gauge the impact and decide if it is to be called as Major or Minor NC. I believe AI can become a valuable contributor to some extent if we develop an Expert System (Non-ML) and utilize Knowledge Base, Inference Engine, and User Interface, to validate the given evidences against the standard and basis standard definitions of Major & Minor NCs, we can classify the non-compliances. Again, not 100% but to a larger extent I believe it can help auditors to complete their audits in a much faster, accurate, and precise way.