Beyond SaaS: How Zoho’s ‘SaaSpocalypse’ and Tech Mahindra Orion Redefine AI Agents

In 2016, the global enterprise environment has shifted from the defensive stance of digital transformation to the offensive, autonomous stance of agentic intelligence. This transformation is characterised by the emergence of systems that not only support human activity but have agency, intelligence and the capacity to perform complex workflows across siloed environments. The conceptual gap between human intention and machine action has been strengthened by the creation of “Hybrid Pods” – integrated teams of human planners and AI agents that blend the computational rigour of machines with the contextual understanding of humans. At the heart of this evolution is a recasting of software economics, led by the likes of Zoho and Tech Mahindra, and supported by NASSCOM’s policy settings, which has declared 2016 as the “Year of Agentic AI”.

NASSCOM and the National Mandate: 2016 as the Year of Agentic AI

NASSCOM’s naming of 2016 as the “Year of Agentic AI” is part of a shift in the Indian IT industry from conversational to agentic autonomy. This shift is guided by the IndiaAI Mission, a national program backed by a ₹10,000 crore investment and 40,000 GPUs to establish sovereign AI capabilities in India. This initiative seeks to minimise reliance on foundational models from abroad by nurturing the creation of small language models (SLMs) tailored to Indian languages and data environments controlled locally.

The transition from “hype to pragmatism” in 2016 is reflected in the dramatic capital shift towards agentic systems. It is estimated that the movement towards sovereign AI will add as much as $1 trillion to the Indian economy by 2035. This is enabled by the capacity of agentic systems to tackle complex, high-friction issues in industries such as manufacturing, agriculture and logistics, where decision-making is common and highly economically significant.

The Convergence of Policy and Technology

2026’s regulatory landscape has evolved. Regulators in China (the Cyberspace Administration of China – CAC) and Europe have established regulations requiring “non-human” ID for AI agents and anti-addiction controls for companion apps, offering a model for how autonomous agents interact with humans globally. NASSCOM’s approach is in line with these global developments in its focus on “AI Delivered Right,” which prioritises transparency, accountability and cost-efficiency in the use of AI. This approach ensures that as agents are being deployed into core business processes, they do so within the governance framework and not the “move fast and break things” framework, which can lead to disruption of critical enterprise processes.

The NASSCOM vision for 2026 is a shift from career ‘ladders to lattices’ and from ‘pyramids to diamonds’ in organisational structure. This shift is made possible by the emptying out of low-level roles, which are increasingly fulfilled by self-sustaining ‘Minimum Viable Agents’ (MVAs), freeing humans to concentrate on strategic and exception management.

Zoho and the ‘SaaSpocalypse‘ Philosophy: Software Economics in the New Era

Zoho co-founder Sridhar Vembu has dubbed the 2026 market correction a ‘SaaSpocalypse’, referring to the $285 billion wiped out from the market capitalization of enterprise software companies. This market correction was brought on by the recognition that the ‘inflated balloon’ of software economics based on seat-based licensing and scarcity has been burst by AI agents. The world before 2026 considered enterprise complexity as a maturity indicator; but the agentic world favours simplicity.

The End of Seat-Based Licensing

Vembu’s approach to Zoho is to cannibalise the business model. The shift is from selling ‘access’ to a software platform to selling ‘guaranteed results’ or outcome-based pricing. This is a necessary shift because AI agents can now perform the data integration, governance and regulatory compliance work previously done by hundreds of specialist SaaS vendors.

The Zoho philosophy is that businesses will cut their SaaS vendor count by 30-50% by replacing point solutions with multi-agent workflows that can reason across an entire business process. This enables point vendors to focus on knowledge and governance rather than staff headcount.

The Mechanics of Hybrid Delivery

The ‘Hybrid Pod’ concept translates this economic reality. For Zoho and EY, these pods are smaller, more responsive teams with AI agents managing the day-to-day campaign execution, data integration and reporting. The human strategists in the pod handle the creative strategy, branding and nuanced customer insights that need human empathy and understanding.

This approach allows firms to offer enterprise-scale solutions with much smaller teams, redefining the old IT services paradigm of “people = productivity”. Here the “system is the job” and people are hired for their ability to supply the goals, constraints and escalation policies for the agents.

Tech Mahindra Orion: The Platform of the Agentic Enterprise

Tech Mahindra has announced TechM Orion, an enterprise-class agentic AI platform in partnership with NVIDIA, to address the agentic shift. Orion is not simply a tool but an orchestration platform that enables enterprises to move from experimentation with AI to production, governance and scalability of autonomy.

Technical Stack and Implementation Details

Orion leverages the NVIDIA AI Enterprise stack, with tools such as NeMo for model creation, RAPIDS for data preparation, and NIM for model serving. Orion can be readily deployed across all major cloud providers (Amazon Web Services, Microsoft Azure, Google Cloud), private cloud or air-gapped (disconnected from the internet) on-premises systems for regulated industries such as healthcare and financial services.

The “Chat-to-Agent” (C2A) interface is a key technical feature of Orion. This conversational interface enables business analysts and subject matter experts to build and deploy intelligent agents in under a week, using a visual no-code/low-code platform. Agents can be further trained using enterprise data and linked to enterprise systems such as SAP, ServiceNow, and Salesforce.

The Core Agent Ecosystem

The platform offers a marketplace of more than 200 production-ready agents for specific enterprise use cases. They can be integrated across platforms and work in a multi-agent system (MAS) to achieve a goal.

IT Operations Agent: This agent resolves incidents by autonomously performing root-cause analysis (RCA). It detects technical problems and provides concrete solutions to reduce downtimes in critical systems.

Pharmacovigilance Agent: In the pharmaceutical industry, this agent optimises drug safety. it automatically categorises and prioritises drug reaction reports, intake of case reports, and complies with global drug safety standards.

Data Engineering Agent: To remove manual development bottlenecks, this agent creates end-to-end data pipelines. It produces YAML, SQL and PySpark scripts for production, bridging the gap between data and design.

Procurement Agent: This SQL agent streamlines worldwide procurement insights and vendor management. It provides monthly reporting and financial reports in visual form, removing the need for manual data mining.

Intelligence Fabric and Graph-Based Reasoning

Tech Mahindra’s “Intelligence Fabric” is a continuously learning fabric that sits on top of the existing IT infrastructure to support the Orion platform. The Intelligence Fabric is different from existing data fabrics, which focus on capturing historical data, by understanding the behaviour of business processes while execution is underway.

The fabric uses a dynamic semantic business graph, in which orders, inventory and customer commitments are represented as nodes in a flow instead of isolated data. The system can then reason over the graph to propagate the effect of a delay or constraint across the dependencies in real time. For instance, if a shipment is delayed at a manufacturing site, the fabric can instantly determine the financial risk and recommend alternative options for execution, before the delay affects the end customer.

This system supports what Tech Mahindra calls “context-aware autonomy” in agents, meaning they can share a real-time view of the whole organisation. Governance is baked in via the Model Context Protocol (MCP), which provides a control layer that makes all actions by agents explicit, auditable and compliant with policy.

Protocols: The Interoperability of Agentic Systems

By 2026, agentic systems must be able to interoperate with different ecosystems. Tech Mahindra’s solution leverages open protocols like the Model Context Protocol (MCP) and Agent-to-Agent (A2A) protocols to enable this.

Model Context Protocol (MCP) and A2A

MCP is a key enabler for modernization that enables AI models to be fed contextually relevant real-time information from legacy systems without having to rewrite the system. In a sophisticated multi-agent system (MAS), MCP allows a ‘Voice Agent’ to call an ‘Insurance Agent’ to confirm insurance benefits, and a ‘Triage Agent’ to assess symptoms.

Tech Mahindra’s ‘Wrapper Service’ provides agent-to-agent (A2A) message translation to REST/HTTPS-based services. This guarantees protocol compatibility and stateful communication between agents that could be designed for different protocols. To minimise the response time in these exchanges, Tech Mahindra uses Secure Low-Latency Interactive Messaging (SLIM), facilitating near real-time orchestration required in critical domains such as healthcare and autonomous manufacturing.

The ReAct Reasoning Loop

At the level of the individual agent, the “Reason-Act-Reflect” (ReAct) loop is the de facto architecture for autonomous agents. The loop requires the agent to:

• Reason: Perceive the situation, break the goal into sub-goals, and choose an action.
• Act: Carry out the action, using tools, application programming interfaces (APIs) or environment modules.
• Reflect: Assess the result of the action, update its memory, and update the plan accordingly.

This approach goes beyond turn-based interactions to goal-driven behavior with long-term memory recall, usually implemented with vector databases such as Pinecone or Milvus using Retrieval-Augmented Generation (RAG).

Customized Blueprints: CFO Blueprint and Healthcare MAS

Agentic AI deployment in 2026 is highly domain-specific, with special frameworks for critical functional areas.

CFO’s Blueprint for 2026: Static to Agentic Automation

Tech Mahindra’s “CFO Blueprint for 2026” underscores the move from static automation (e.g., Robotic Process Automation – RPA) to agentic decision making. Static automation like RPA was efficient but struggled with data structure or process change. Agentic AI, on the other hand, is a dedicated execution layer that can:

Read between the lines: Rather than merely describe a tax rule, an agent interprets the rule and recognizes that a facility has created a new nexus that requires filing, then queues the filing.

Automate Reconciliation: SQL agents reconcile data across systems, detect inconsistencies and audit activity, automating up to 70% of reconciliation.

Real-Time Compliance: Agentic systems continuously process transaction data to apply rule engines and pattern recognition, to validate results across ERP and tax engines in real-time.

Healthcare and Medical Device Development

In the medical sector, agentic systems are used to orchestrate testing processes and ensure drug safety. A multi-agent system (MAS) can create a modular framework for patient interactions, with a Voice Agent managing the flow of the patient conversation and handing off tasks to the Triage and Insurance agents.

Medical device manufacturers benefit from continuous monitoring via agentic AI in quality management systems. Agents track deviations, prioritise quality issues by severity and model the impact of change control to ensure traceability and risk management. This is essential to meet the Quality Management System Regulation (QMSR) that replaced previous standards in February 2026.

The use of agentic AI is recasting the design of work. The traditional rule of ‘more people, more productivity’ has given way to ‘Hybrid Pods’ of humans and AI agents. These pods work together to merge accuracy and efficiency with no additional labour costs, enabling enterprises to execute complex workflows.

The Evolution of the Org Chart

In 2016, the pyramid is giving way to the diamond. This shift is driven by AI eliminating low-level jobs, which were previously centred on execution-intensive, data-driven work. There are lattices, rather than ladders, in career development, as employees move from one squad of agentic workers to another and focus on high-level problem solving.

The Human in the Hybrid Pod

In the hybrid pod, the role of the human is elevated to that of strategist, supervisor and exception handler. Humans set the goals, constraints and escalation policies for agents. This means that while agents are involved in the operational execution, humans are still needed to make decisions that involve regulatory, financial or legal considerations.

New Roles for 2026

The advent of agentic systems has led to the emergence of new roles that integrate AI engineering with domain knowledge:

• Agentic AI Architects: Experts who create multi-agent orchestration layers and semantic graphs.
• Regulation-Smart Prompt Engineers: Engineers who ensure agent prompts adhere to regulatory and safety constraints.
• Safety and Oversight Stewards: Experts who monitor ‘Chain of Thought’ logs and step in when agents become uncertain.

Transformation Squads and Workflows

For successful migration, enterprises are creating “Transformation Squads” that engage business and IT leaders from sales, operations, supply chain and finance in the discovery phase. Squads apply ‘Transformation Playbooks’ to find bottlenecks with high friction and re-engineer workflows to scale from the beginning to enterprise volumes. They are not only judged by the accuracy of the model but by revenue growth, cost reduction and cycle-time improvement.

Ethics of Delegation and Risk

The shift to self-organising agents presents operational and governance challenges. Risks of hallucinations, biases, and adversarial attacks require safeguards. Tech Mahindra’s “AI Delivered Right” program highlights the importance of ethical, transparent and economical adoption of AI.

The Chain of Thought (CoT) Audit Trail

In Q4 of 2026, national regulators are likely to mandate “Executive Liability for Negligent Delegation” for autonomous AI. Under this regime, companies must provide a cryptographically signed ‘Chain of Thought’ (CoT) audit trail of any agent decisions with serious consequences. The logs must be kept for at least seven years, and should be searchable and exportable for auditing.

The 2026 threat model rewrites the insider threat model. Malicious external agents can exploit multi-turn prompt engineering to extract budget caps or cost structures in time frames humans can’t review. To address this, sensitive agent tasks are performed in Hardware-Rooted Trusted Execution Environments (TEEs). Additionally, ‘Ephemeral Credential Minting’ provides agents with only time-limited, verifiable authorisation (Proof of Intent) for each task (credential expires in 60 seconds).

Industry Leader Quotes on the Future of Work

The transition to agentic systems is best captured through the insights of those driving the change.

Tech Mahindra CEO Mohit Joshi highlights the scale:

“We are enabling hybrid workforces across industries with our AI portfolio, built on more than 300 smart agents. As the IT industry enters the phase of autonomous and data-driven operations, our approach on multiple fronts – from AI platforms to efficiency-driven execution – seems to be working.”

Zoho Co-founder Sridhar Vembu takes a more radical view:

“AI is the pin popping this inflated balloon of software economics. This wasn’t an apocalypse – it was forced evolution: from headcount to hybrid pods, from labor arbitrage to AI orchestration, and from selling time to delivering outcomes.”

Nikhil Malhotra, Chief Innovation Officer of Tech Mahindra, looks at the pragmatic shift for enterprises:

“Today enterprises are seeking return on investment, security and roadmap for business process transformation. Paired with NVIDIA’s world-leading AI infrastructure, TechM Orion helps enterprises to transform from experimentation to transformation.”

Shankar, Zensar’s CTO, points to the evolving C-suite:

“Agentic AI, like digital transformation a decade ago, is changing the face of technology leaders. CIOs are collaborating more with COOs, CFOs and CMOs as agentic AI operates across business functions. This synergy allows CxOs to concentrate on strategy and change management, while agentic AI takes over complex processes.”

Looking ahead: The era of co-intelligence and post-scarcity

Beyond 2026, the future of agentic AI promises a ‘Post-Scarcity Era’ with autonomous discovery driving advancements in science and economics. Agentic systems are being used to democratise discovery in areas such as room-temperature superconductors and carbon capture technologies, achievements that have proven impossible for humans to achieve.

Biotech and AI

Tech Mahindra sees a future where human creativity and machine accuracy seamlessly intersect. Agentic AI can help neural interfaces to enhance brain functions without erasing identity, help overcome neurodegenerative diseases and cancer. The “future of AI is in biology” where AI can learn, evolve and build on each iteration, akin to a learning flywheel.

The Real-Time Enterprise

We now have the foundations for the ‘Real-Time Enterprise’ with smart data systems, integrated MAS blueprints and LLMOps/MLOps systems that automate the machine learning lifecycle. These platforms allow businesses to make smart decisions at scale and in real-time, delivering personalised experiences to customers and agility in the face of market upheavals.

Conclusion: Lessons for the 2026 Enterprise

The evolution from digital transformation to agentic systems is a natural progression. To flourish in the 2026 world, enterprises must progress from “one-offs” to an “AI-first operating model” that is built with governance and economics in mind. The lessons from this study are:

Redesign for Outcomes: Don’t value AI by accuracy or hours used; value it by revenue gain and cycle time reduction.

Prioritize Governance: Quality, integration and security of data are essential for autonomous execution.

Be the Hybrid Pod: Organise the org chart to support diamond-shaped pods and the “lattice” career, and emphasise human skills in high-value strategy and human relationships.

Embrace Interoperability: Use open standards such as MCP and A2A to enable agents to communicate throughout the enterprise ecosystem and existing systems.

Agentic AI is not the successor to humanity, but its ally. By embracing empathy, ethics and excellence, enterprises can take humanity to a future in which human creativity and machine efficiency combine to address the problems of our times.