Enterprise AI in Production: Mumbai Call for Proposals

AI in IT Operations

{Traditional IT Service Management (ITSM) has largely focused on ticket volumes, resolution times, and SLA adherence. However, enterprises increasingly need IT operations to deliver measurable business outcomes—not just operational efficiency. A payment failure in banking, a checkout outage in retail, or downtime in a healthcare system can have consequences far beyond IT, affecting revenue, customer trust, patient care, and regulatory compliance. The challenge was to bridge the gap between IT incidents and business priorities while reducing manual effort, accelerating resolution, and improving service quality. We addressed this through Mastek’s AI-powered ITSM Agents, designed to bring contextual intelligence, intelligent orchestration, and automation into the service management lifecycle. Our approach combines AI-driven incident acknowledgement, resolution intelligence, root-cause analysis, and operational dashboards to help service teams move beyond reactive ticket handling towards proactive, insight-led operations.

A key design decision was to adopt an API-first, platform-agnostic architecture, integrating with enterprise ITSM platforms such as ServiceNow and leveraging LLMs, enterprise knowledge sources, and multi-agent orchestration. Rather than treating every incident equally, the solution focuses on contextual understanding, recommending relevant resolutions, orchestrating workflows, and enabling human oversight where required. Mastek’s Response.AI delivers 100% response SLA compliance, while Resolution.AI has demonstrated approximately 40–44% improvement in Mean Time to Resolution (MTTR) in reported deployments. The AI-powered ITSM Dashboard replaces manual weekly and monthly service reporting with data-driven operational visibility, helping leaders identify improvement opportunities and track outcomes. These capabilities establish the foundation for the next evolution: connecting incidents to vertical-specific business KPIs, including revenue protection in retail, transaction continuity and risk reduction in BFSI, clinical system availability in healthcare, and production uptime in manufacturing. The outcome is a shift from ITSM as a cost-centre function to an AI-enabled business-performance driver—improving operational efficiency today while building towards predictive, autonomous, and business-aware service management.}

{End-to-end data flow

1. Ingest: An incident or alert arrives from ServiceNow or an integrated operational system.
2. Enrich: The system retrieves relevant incident history, knowledge articles, service context and runbooks.
3. Reason: AI agents classify the issue, recommend a resolution, investigate potential causes and coordinate tasks.
4. Act: The system recommends or executes approved workflows, escalating uncertain or high-risk cases to engineers.
5. Learn and measure: Resolution feedback and operational metrics support quality improvement and business-outcome reporting.Components, models and tools
   Layer Technologies and purpose
   Integration REST APIs, ServiceNow and monitoring integrations
   AI models Large language models (LLMs), including GPT-4o and GPT-5 in the discussed solution evolution
   Knowledge Retrieval-augmented generation (RAG), enterprise knowledge sources and vector search
   Orchestration Multi-agent workflows using LangGraph / Node.js
   Hosting Microsoft Azure
   Execution ITSM workflows, API-triggered actions and controlled remediation
   Analytics AI-powered ITSM Dashboard and operational KPI reporting

}

{Production Evidence: Mastek’s AI-First Managed Services

Scale and time in production: Mastek introduced Response.AI and Resolution.AI in 2025 and progressively embedded agentic AI across the ITSM lifecycle. As of 2026, the solutions are deployed across 100+ customer accounts in production, spanning BFSI, healthcare, public sector, retail, manufacturing, and other enterprise sectors. Response.AI is live across the reported account base, while Resolution.AI is mandated on core ITSM platforms. Named deployments include AVEVA Solutions (UK), AMICO Operations Holdings (US), South Lanarkshire Council (UK), Mercer & Marsh (APAC), and Samsonite (US); additional customer identities are anonymised due to contractual confidentiality. Production duration varies by account and rollout phase.

Measured results: Mastek’s submitted operational figures indicate up to 40% faster resolution, with reported MTTR improvements of approximately 30–50% in AI-enabled deployments. Response.AI achieves 100% response SLA compliance in the reported deployment scope, while overall SLA compliance is reported at above 98%. Additional reported outcomes include 20–30% lower recurring incidents, 20–30% reduction in L1/L1.5 ticket-handling effort, and 25–40% lower operational costs in AI-enabled accounts. These figures are approximate, account-dependent results and should not be interpreted as independently audited or universally achieved.

Measurement methodology: Results are assessed using ITSM ticket records, incident creation and resolution timestamps, SLA compliance reports, incident recurrence and problem-management trends, and service-effort and cost data. Response SLA performance is measured against contractual acknowledgement targets; MTTR is compared across defined pre- and post-deployment periods; recurring incidents are tracked through problem and incident classifications; and effort and cost improvements are evaluated using operational workload and account-level delivery metrics. The AI-powered ITSM Dashboard consolidates operational KPIs to support ongoing monitoring and management review. Comparisons should use consistent measurement windows, account scope, ticket severity, and workload baselines to ensure fair attribution of improvements to AI.}

{One key challenge during the implementation of Mastek’s AI-powered ITSM Agents was balancing resolution speed with the accuracy and reliability of AI-generated recommendations. Early iterations of Resolution.AI highlighted that enterprise knowledge repositories are not always consistent, complete, or up to date. Similar incidents could have different resolutions across accounts, while incomplete ticket descriptions and unstructured work notes sometimes resulted in irrelevant recommendations or lower confidence in AI-generated responses. A further trade-off was between autonomy and control: allowing agents to execute remediation actions independently could accelerate resolution, but introduced risks when dealing with complex incidents, production environments, and account-specific procedures. These challenges reinforced that deploying an LLM alone is insufficient for reliable enterprise-grade ITSM automation.

We addressed this by treating AI adoption as a continuous service-improvement journey rather than a one-time technology rollout. We strengthened contextual retrieval and grounding in approved enterprise knowledge, incorporated feedback from service-management SMEs, refined agent instructions and orchestration workflows, and introduced validation and human oversight for higher-risk scenarios. We also emphasised ongoing UAT, resolution-quality reviews, and KPI monitoring before expanding adoption. The key learning was that accuracy, trust, and measurable business outcomes must take precedence over maximising autonomous actions. In hindsight, we would establish account-level knowledge-readiness assessments, standardised evaluation datasets, confidence thresholds, and explicit success criteria earlier in the rollout. We would also measure not only whether an AI recommendation was accepted, but whether it led to a correct, sustainable resolution. This experience shaped our approach towards controlled autonomy, continuous learning, and evidence-based scaling—essential foundations for moving from AI-assisted ITSM towards dependable, business-aware autonomous operations.}

{Connect AI to business KPIs, not just IT metrics. Identify how incidents affect revenue, customer experience, patient safety, production uptime, or regulatory compliance. Prioritise AI use cases based on measurable business outcomes rather than ticket automation alone.

Build trust through contextual intelligence and controlled autonomy. Ground AI recommendations in approved enterprise knowledge, validate outputs with subject-matter experts, and introduce human approval for high-risk actions. Start with well-defined use cases, establish confidence thresholds, and expand autonomy as accuracy improves.

Measure outcomes continuously and scale based on evidence. Establish clear pre-deployment baselines for MTTR, SLA compliance, recurring incidents, analyst effort, and operational costs. Track post-deployment results consistently, incorporate user feedback, and refine AI agents through structured testing and continuous improvement before scaling across teams or business units.}

{This session is designed for IT and technology leaders, CIOs, CTOs, ITSM and IT operations professionals, service delivery and managed services leaders, enterprise architects, AI practitioners, and digital transformation decision-makers looking to move beyond traditional ITSM automation towards AI-driven, business-outcome-oriented operations. It will be particularly valuable for organizations across healthcare, BFSI, retail, manufacturing, and other industries where IT service reliability directly influences revenue, customer experience, operational continuity, and regulatory compliance.

Attendees should have a basic understanding of IT Service Management concepts, such as incident management, service-level agreements (SLAs), knowledge management, and service operations. Familiarity with enterprise ITSM platforms such as ServiceNow, monitoring tools, or managed services environments would be beneficial but is not mandatory. No hands-on AI or machine learning expertise is required. The session will focus on practical implementation lessons, real-world applications of AI agents, measurable operational outcomes, and approaches to responsible adoption. Attendees will leave with a framework for identifying high-impact AI use cases, aligning them with industry-specific business KPIs, measuring value, and scaling AI-enabled ITSM responsibly within their own organizations.}

{a 30-minute talk including questions}

{Anuradha Mhatre

AI Centre of Excellence Leader | Managed Services | Mastek

Anuradha Mhatre is a senior technology and business leader at Mastek, leading initiatives across the AI Centre of Excellence and the Managed Services charter. With over 25 years of experience in IT service management, service delivery, program management, and enterprise technology transformation, she specialises in leveraging AI, Generative AI, and intelligent automation to drive operational excellence and measurable business outcomes.

Anuradha has been instrumental in advancing AI-led Managed Services through the development and adoption of agentic AI solutions for IT Service Management (ITSM), including Response.AI, Resolution.AI, and AI-powered operational intelligence. Her work focuses on transforming traditional service operations into predictive, intelligent, and increasingly autonomous models that improve resolution times, strengthen SLA performance, optimise costs, and enhance service quality. She is passionate about connecting AI innovation with tangible business value and helping enterprises move beyond operational automation towards outcome-driven transformation.

Contact details

Organisation: Mastek

Email: anuradha.mhatre@mastek.com

LinkedIn: https://www.linkedin.com/in/anuradha-mhatre-861b8b2a/}

{ Confirm your availability to speak in Mumbai on 31 October 2026, between 11 AM and 4 PM IST, and participate in editorial preparation. All talks will be recorded, and live streaming will be available for Fifthel annual members}

{Optional: add a link to draft slides in PDF or PPT format, with commenting access enabled where supported}