Exclusive: SAP's Manoj Swaminathan Wants to Kill the Enterprise Dashboard

When SAP launched R/3 in 1992, it set the template for how enterprise software would work for the next three decades. Employees logged in, navigated to a module, completed a transaction, and logged out. Firms like Accenture, Deloitte, and Infosys built entire practices around that model, certifying workforces screen by screen, module by module.

Manoj Swaminathan, General Manager and Chief Product Officer for SAP’S Business Suite, now says that model is ending. "From a user perspective, you will [no longer] see [the] static experiences [we] used to have, users going into a particular application, logging in, going to a dedicated screen to do their activities," Swaminathan says to AIM Media House. "How do we move from systems of record towards systems of action and the systems of intelligence?"

If SAP’s vision materializes, the next generation of enterprise workers may never need to know which application performs a task. They will simply state an objective, and a network of AI agents will execute it.

The Shift Enterprise Software Has Not Seen Since the Cloud

Robert Kugel, executive director of Business Research at Information Services Group (ISG), highlighted disruption in ERP in a January 2025 report, saying, "ERP systems, which have served as the central nervous system of enterprises for decades, are in the midst of a technology revolution." ISG predicts that almost all ERP providers will incorporate AI fully by 2027.

According to a Gartner report, 40% of enterprise applications will embed task-specific AI agents by the end of 2026, up from less than 5% in 2025, describing the shift as one that will transform enterprise applications from tools that support individual productivity into platforms that enable autonomous workflow execution.

Cut to SAP Sapphire, the company’s flagship event in Orlando last month, CEO Christian Klein announced what he described as the most consequential evolution in SAP's history: the Autonomous Enterprise, built around a Business AI Platform and an Autonomous Suite of more than 50 Joule assistants and 200 specialized agents across finance, procurement, supply chain, HR, and customer engagement. Klein told a 30,000-strong audience, both in person and virtually, that SAP was becoming "a business AI company."

Swaminathan describes what that looks like at the workflow level. "They now have this one simple conversational experience," he says, describing a model where most routine processing runs touchless and exceptions are handled by humans in the loop. "So they are purpose driven, truly becoming app-less experiences."

Context, Explainability, Trust

For AI agents to operate reliably within enterprise workflows, Swaminathan says three things must hold: the agent must understand the business context it is working in, its decisions must be explainable, and the data it touches must be governed. SAP's case for the autonomous enterprise rests on all three.

On context, he says general-purpose AI models fall short in enterprise settings because they lack the process knowledge specific to the organization deploying them. SAP's answer is a knowledge graph, built on 50 years of ERP data, business process structures, and industry configurations.

"The business context, process aspects are absolutely honored and well understood in a way the customer has tailored it for their business, and that's super important, and that is where it starts," the Chief Product Officer says.

SAP also trains frontier models on its own codebase, processes, and data rather than deploying them unchanged. "When developers create new agents or extend our agents, they get [a much] better recommended code, because it already understands SAP code context, SAP security principles," Swaminathan says.

This context takes time to replicate. "Others can catch up; it's going to take them many years before they can actually gain the data, the process, more than what SAP has today," he adds.

Meanwhile, on explainability, SAP embeds a visible indicator into every AI-driven output across its platform, allowing users to see the data factors and reasoning behind a result.

The company also monitors every agent instance through an agent hub, tracking for drift from expected execution paths. "If the agent drifted from the standard expected code path, or the defined guardrails of the code path that you had, you can detect [that] right away," Swaminathan says.

On data governance, each agent operates within defined access boundaries. SAP certifies its platform against International Organization for Standardization (ISO) and System and Organization Controls (SOC) requirements.

"Right from the construct of data access, data privacy, [and] data residency aspects of things, we want to honor those," Swaminathan says. For finance and other regulated functions, auditability matters when AI-generated outputs are reviewed by external auditors, he says.

Beyond Technical Evals

On the competitor front, Oracle announced plans to embed ROI measurement directly into its platform in March 2026, expanding its AI Agent Studio for Fusion Applications with a dedicated dashboard that tracks time saved, cost reductions, and productivity gains at the individual-agent level. Microsoft's Dynamics 365 2026 release wave 1, covering April through September, moving Copilot into the agentic layer across finance, supply chain and HR.

The accountability problem both are addressing is significant. The IBM Institute for Business Value's 2025 CEO Study, which surveyed 2,000 chief executives across 33 countries, found that only 25% of AI initiatives have delivered expected returns, and just 16% have scaled enterprise-wide.

Swaminathan draws a distinction between technical evaluations, which measure whether a model performed correctly, and business evaluations tied to financial results. "We have something called business evals," he says. "What is the outcome that we are defining for every agent? You can now realize 10 to 15% productivity with this agent in billing, so much cost reduction with this agent in your aging bucket determination."

The value outcome is expressed in monetary terms, per agent instance. "Sometimes it's productivity, sometimes it's cost, sometimes it's revenue generated, or revenue secured, or loss prevented," he says. "The loss prevention is described as a dollar value sometimes, right? So it's a monetary value in terms of [the impact on the business] that you can model as well."

For Prem Udayavarma, CIO of Aditya Birla Renewables, the direction is clear, and the timeline is not short. His company completed its SAP digital core implementation last year and is now evaluating AI for legal, contracts management, and statutory compliance work.

At SAP Now Mumbai, the product that stood out to Udayavarma was Joule Studio, SAP's environment for building custom agents without infrastructure setup. "Some of the agents which come on SAP themselves release, which will generate, so it may solve, let us say, 50-60% of the problem in that business process," he told AIM Media House. "You always need to have customized tailor-made agents for your business context."

Udayavarma estimates 12 to 18 months to reach meaningful autonomy in selected workflows. In the renewables sector, he notes that every procurement and every site configuration differs, which limits how far general-purpose agents can go.

Swaminathan describes the path through invoice processing. A company at 30% touchless processing today, with 70% of invoices requiring human review, can use agent monitoring to identify where that review adds nothing. "Eight out of 10 of these situations, humans didn't do anything, humans just reviewed it and hit approve," he says.

Those cases become candidates for full automation. The system retains learning from every remaining exception, what SAP calls company memory. "In some cases, it will be faster, while in some cases it will take time for them to be able to mature their processes," Swaminathan says.