AI That Clocks In: Sree Hari Subhash on Oracle Cloud Assistants and Early Alzheimer's Detection
Sree Hari Subhash shows how Oracle Digital Assistant tied to HCM/Payroll delivers instant HR answers, fewer tickets, and cleaner data. Logs and BI drive weekly fixes.
AI Assistants That Actually Work for HR: Lessons from Sree Hari Subhash
AI is no longer a side project. With moves like the Oracle-Google Cloud partnership bringing Gemini models closer to enterprise stacks, AI has become part of core operations. HR leaders need systems that cut wait times, clear ticket queues, and improve employee trust. Data engineer Sree Hari Subhash shows what that looks like in practice.
From press release to payroll: building an HR assistant that behaves like a coworker
While at Yum! Brands Inc., Sree Hari Subhash led the integration of Oracle Digital Assistant into HR and payroll. The goal was simple: give every employee instant, accurate answers without waiting on a human specialist. That meant building an AI "specialist" that fits into existing workflows, not a novelty chatbot.
The assistant was connected to Oracle Human Capital Management (HCM) and Payroll with two-way data exchange. Employees could check PTO balances, confirm pay schedules, update banking info, track bonuses, and book vacations. With the right permissions, the assistant could also make changes on request.
What changed for payroll and HR service delivery
Salary data started flowing directly to employees through the assistant. Issues and inconsistencies were caught, reviewed, and fixed, then pushed back into the system. Over time, reports were automated, and employees shifted from email tickets to fast self-service.
Subhash and the team trained the assistant on company data so responses were contextual, not generic. Each session was logged and analyzed through Oracle Business Intelligence. IBM DataStage pipelines automated reporting so performance could be monitored in real time and scripts could be adjusted quickly.
His summary of the work: "We automated processes from 401(k) calculations to payroll integrations and delivered accurate analytics on time. Employees received support faster without waiting for HR."
The practical playbook HR teams can use
- Start inside your stack: connect the assistant to HCM and Payroll first.
- Enable two-way actions with proper role-based access controls.
- Train on your policies, calendars, pay rules, and FAQs to avoid generic answers.
- Log every session; review trends weekly to fix root causes and update scripts.
- Automate recurring reports (payroll status, ticket deflection, SLA) and make them self-serve.
- Iterate on real errors: track mismatches, correct them, and push updates back to users.
Metrics that matter to HR
- Average time-to-answer and time-to-resolution
- Ticket deflection rate (email/phone to assistant)
- Payroll discrepancy rate and time-to-correct
- Self-service completion without human escalation
- Employee sentiment on accuracy and usefulness
Health assessment beyond payroll: AI research with social value
Subhash also explored how machine learning can support healthcare. One research track focused on early indicators of Alzheimer's disease using handwriting analysis. Using the DARWIN dataset from the UCI Machine Learning Repository (174 individuals), he tested SVM, decision trees, and logistic regression in RapidMiner.
The best-performing approach used an SVM with an RBF kernel. The model identified Alzheimer's cases with 86% accuracy and healthy cases with 77%. He presented "Prediction of Alzheimer's Using DARWIN Dataset" at the National Conference on Information and Communication Technology in India and is preparing an international journal submission.
Why this matters for HR leaders
The takeaway is clear: AI assistants can function like trained coworkers when they're integrated into systems, fed with company data, and measured against business outcomes. That is the difference between a demo and daily value. Subhash's work shows the path for HR: start with HCM and payroll, lock down permissions, train on policy, and refine based on logs and analytics.
About Sree Hari Subhash
Sree Hari Subhash builds AI-based assistants that act as "professional employees" across accounting, HR, and operations. He has implemented Oracle Cloud solutions at scale, automated HR analytics with IBM DataStage and Oracle BI, and pursued applied research in AI for healthcare. He is a fellow member of Hackathon Raptors and now serves as a Senior Data Engineer at KPMG, where he designs data architectures that deliver timely, reliable analytics. He holds Microsoft Azure AI certifications.
Resources
Next steps for HR teams
- Map your top 20 HR questions and actions (payroll, benefits, PTO, verification). Build flows for those first.
- Connect your assistant to HCM/Payroll with read/write, then add guardrails.
- Set weekly reviews of assistant logs and BI dashboards to remove friction.
- Upskill your team on AI fundamentals and prompt design for HR workflows.
If you're planning capability-building for your HR team, explore curated programs by role: AI courses by job and popular AI certifications.
