AI agents are no longer waiting inside one enterprise application. HCLTech, Google Cloud, and ServiceNow are pushing them into field service, customer experience, factory operations, and IT workflows, raising the harder ERP question: who governs the action when the data, model, and workflow live in different platforms?
HCLTech has expanded its collaboration with Google Cloud and ServiceNow to put AI agents inside operational workflows, from field services to customer experience to the factory shop floor, all running on Google’s Gemini Enterprise platform. For SAP practitioners, the issue is whether that layer can coexist with SAP Business AI without splitting data ownership, governance, and ROI accountability across too many control points.
This is a three-party story, and that matters. Gemini Enterprise supplies the foundational AI, ServiceNow supplies the workflow orchestration and its Blueprint for Agentic Business, and HCLTech supplies implementation, integration, and industry context.
“The future of enterprise AI lies in orchestrating intelligent agents across the business as a connected system of action,” said Michael Park, ServiceNow’s Senior Vice President of Global Partnerships and Channels. The phrase to sit with is a connected system of action, because connection across systems is Precisely where SAP governance gets complicated.
Agents Move into Operational Workflows
Per HCLTech, the initial solutions span two high-impact domains plus a manufacturing assistant. In Field Services, Gemini Live, integrated with ServiceNow Field Service Management, provides technicians with real-time audio and visual intelligence to enable faster issue resolution. In Customer Experience, the goal is to preserve customer intent across channels. A next-generation Factory Shop Floor Assistant delivers real-time operational intelligence to manufacturing environments. There is also an ITOps ServiceNow Agent available on Google Cloud Marketplace for Gemini Enterprise for incident management and remediation. The useful detail is the placement of AI inside the process flow, not only in a separate assistant interface.
In a manufacturing scenario, that could mean an assistant flags a quality deviation, initiates a service workflow, or supports a downstream operational decision. In an ERP environment, the underlying data, material, production, maintenance, or order records may live in ERP, while the agent acting on it operates in a different layer. That pattern works when integration and controls are explicit. It also raises a governance question many ERP teams have not settled: Which platform is accountable when an AI-assisted action is wrong, and which audit trail is authoritative?
The Control Plane Starts to Fragment
ERP customers are increasingly asked to run multiple AI execution and governance layers at once: SAP Business AI inside ERP and line-of-business suites, Google’s Gemini Enterprise as the foundational model layer, and ServiceNow as the cross-system workflow orchestrator. Each brings its own data ingestion pattern, policy model, and definition of an agent. Tellingly, the announcement itself leans on ServiceNow’s AI Control Tower for “visibility and governance of AI agents” inside Gemini Enterprise, which is the same control-plane question, just answered in ServiceNow’s favor rather than SAP’s.
SAPinsider’s 2025 AI Adoption research shows AI Leaders use a wider portfolio of partners and platforms than Adopters or Beginners. That is a maturity correlation, not a deployment instruction. AI agents do not become governed because the architecture diagram has more arrows. Without clear ownership, the result is agent sprawl, committee work with better branding. The Leaders reporting roughly 13% cost savings and 25% productivity gains from AI generally achieved these gains after resolving data readiness, ownership, and repeatable deployment models, not before. Those fundamentals get harder, not easier, when execution moves outside the primary system of record.
Governance Becomes the Implementation Test
The announcement by HCLTech, Google Cloud, and ServiceNow uses the language of responsible AI and scalable deployment. The harder work is operational: clarifying how policy ownership is allocated when an agent touches both SAP Business AI and a Gemini-plus-ServiceNow stack, and where ROI is actually measured.
This is where systems integrator accountability matters. HCLTech is positioning its role around implementation, integration, and industry-specific acceleration, and its own CTO, Vijay Guntur, frames the work as moving clients “beyond pilots to sustained, enterprise-wide impact.” For ERP customers evaluating this kind of offering, the deployment questions are concrete: which platform’s governance model applies when policies conflict, how ERP-originated data is logged when an external workflow agent acts on it, and how the ROI baseline is set before the deployment is assessed.
What This Means for ERP Insiders
ERP teams must map every agent to the system it touches. Field service, shop floor, customer experience, and IT agents may read or trigger ERP data even when they execute in another platform. ERP leaders should document which business objects each agent can access, which system owns the record, and which audit trail becomes authoritative when an AI-assisted action affects operations.
Leaders must choose a control plane before agents scale. ServiceNow, Google Cloud, SAP, Microsoft, Oracle, and other platforms all want to govern parts of the agentic workflow stack. CIOs and enterprise architects should decide which platform owns policy enforcement, exception handling, approval logic, and accountability before AI agents spread across business functions.
Systems integrators must prove governance, not just connectivity. The real work is reconciling policies, logs, data ownership, and ROI measurement across ERP, workflow, and AI platforms. Buyers should push SIs to define how agent actions will be traced, measured, escalated, and defended before treating multi-platform AI as an enterprise-ready operating model.
Editor’s note: A version of this article was originally published on SAPinsider on 6/25.
