AI News Deep Dive, May 6: The Open MCP Agent Platform Race Hits Your HR Stack
ServiceNow used Knowledge 2026 to plant a flag this week. The open MCP agent platform it announced on May 5 is not another agent or copilot. Instead, it is a bid to be the place where every other vendor’s agent comes to do real work, with audit trails, OAuth, and a single governance plane attached. For founders and HR leaders, the stakes are concrete. The company that wins the agent execution layer also wins your HR tech procurement for the next decade. On the same day, Novaworks shipped a Core HR operating system on top of it, and Microsoft moved Agent 365 to general availability. The agentic AI race just turned into a platform war.
What Happened: The Open MCP Agent Platform Just Went Live
ServiceNow announced four pieces at Knowledge 2026 in Las Vegas on May 5. First, Otto pulls Now Assist, Moveworks, and the AI Experience layer into one AI surface across the platform. Second, Action Fabric ships as an open MCP server. Any third-party agent, including Claude, Copilot, Gemini, or a custom build, can use it to execute governed work on ServiceNow. Third, twenty new AI Specialists arrived for IT, CRM, HR, finance, procurement, legal, security, and risk. Finally, AI Control Tower, the company’s governance layer for every agent in the building, is now offered free for the first year, framed as a $2 million value (Bloomberg, Constellation Research).
Read together, those four moves describe a bet. ServiceNow no longer wants to be a workflow vendor. It wants to be the operating system every other vendor’s agent runs on top of. CEO Bill McDermott set a $30 billion subscription revenue target for 2030, with AI products expected to deliver more than 30% of annual contract value by then.
For HR leaders, the test is concrete. If your HRIS, payroll, ATS, or LMS cannot expose tools to an MCP server, an external agent cannot complete a job on it end to end. Therefore the question for every HR vendor in your stack today is simple. Is there an MCP integration in the next two quarters, or is there not? If the answer is vague, your shortlist for next year’s RFP just got shorter. The companies already using AI agents for HR workflows will move first.
Why It Matters: The Open MCP Agent Platform Changes HR Tech Procurement
This is bigger than ServiceNow. The MCP standard, introduced by Anthropic in late 2024, is now an industry-wide spec. Microsoft, Google, AWS, and OpenAI all support it. So when ServiceNow opens Action Fabric to “any agent, any system,” the practical effect is portable agents.
Here is the part that should make every HR director look up. For the past two years, agentic AI has been a vendor-by-vendor story. Workday built its own agent. ADP built its own. Greenhouse, BambooHR, and HiBob each shipped their own. The promise was always the same: the agent will read your HR data, take action, and save your team time.
However, the work an HR team does spans systems. Onboarding crosses HRIS, identity, payroll, devices, benefits enrollment, and learning. A vendor-locked agent cannot finish that job because it cannot reach across the fence to the next system. An open MCP agent platform can. As a result, ServiceNow’s pitch carries weight, and Novaworks chose the same day to launch its Core HR operating system on top of ServiceNow rather than as a standalone product (PR Newswire via Morningstar). Ateko, a Bell Canada company that serves financial services, telecom, and government clients, is the launch customer.
The new procurement question is not “Does your tool have an AI feature?” It is “Whose execution layer will my agents run on, and is yours one of them?” Companies running top AI tools for HR on top of one platform will move faster than companies running ten point solutions that don’t share an audit trail.
Under the Hood: How the Open MCP Agent Platform Actually Works
Action Fabric exposes ServiceNow’s full system of action: workflows, approvals, playbooks, catalogs, audit trails, and the configuration database. External agents talk to it through MCP. In addition, the Agent2Agent (A2A) protocol, also embedded in the AI Agent Fabric, lets ServiceNow and third-party agents collaborate on multi-step work without a human re-prompting at every step (Reworked).
The control plane
The MCP Server Console handles tool discovery, OAuth, session management, consumption metering, and audit logs. Built-in observability flags drift, hallucination, and policy violations before they reach a customer. AI Control Tower sits one layer above. Specifically, it discovers every agent, model, and workflow running across the enterprise, whether or not it originated on ServiceNow, and applies governance policies against them. Picture air traffic control for AI in your organization. That phrasing comes from ServiceNow itself in the investor briefing.
The Specialists and the front door
The 20 AI Specialists split into vertical functions. The L1 IT Service Desk Specialist, CRM Specialists, and the employee-service AI Specialists are available now, covering HR, workplace services, legal, finance, procurement, supplier management, and health and safety. IT Specialists hit June. Security and risk Specialists go to preview in June and general availability in September. EmployeeWorks sits on top as the conversational front door for employees. Workers ask in natural language across web, mobile, and chat tools. The Specialist underneath does the work and reports back through the same surface. Therefore HR leaders will feel this layer first.
The build is not a green-field rebuild. ServiceNow is layering this on top of the Now Platform, the Configuration Management Database, and Workflow Data Fabric. As a result, existing customers are not asked to migrate. The day-one promise is “your existing flows now have agents inside them.” For the largest enterprises that already standardized on ServiceNow for IT and HR cases, that is the easiest sales motion in enterprise software.
What HR Leaders Do Monday
Three concrete actions, in order.
First, audit your HR stack for MCP support. Pull the integration roadmap from each of your top three vendors, including HRIS, payroll, and ATS. If any of them cannot answer “When will you ship MCP?” inside two quarters, escalate. The penalty for picking the wrong execution layer in 2026 is the same as picking the wrong cloud in 2014. Switching costs in three years will be brutal.
Second, decide where your agents will run. There are now four credible contenders for the open MCP agent platform crown: ServiceNow (HR + IT + finance), Microsoft Agent 365 (productivity-led, with Copilot Cowork shipping a Claude-powered agent), AWS (Bedrock plus Connect Talent for AI hiring interviews), and Salesforce (Einstein, Agentforce). Pick one as the default and let the others work as plug-ins. Running your own without one of these underneath is feasible at the 200-employee mark. Above 1,000 employees, it isn’t.
Third, write an AI-agent policy this month. Who can spin up a Specialist? Who approves the tools each Specialist can call? What is the audit trail when an agent makes a benefits decision? AI Control Tower-style governance is becoming table stakes, and the AI skills gap in HR is widest exactly here. The teams that already wrote AI payroll automation playbooks have a template. The rest will write theirs in the next 90 days, ready or not.
If you are building a global team and want one payroll and HRMS layer that already speaks AI agents, Asanify’s HRMS is built API-first for exactly that integration model. The market just told you why that matters.
FAQs about the Open MCP Agent Platform Shift
What is an open MCP agent platform?
An open MCP agent platform exposes a software vendor’s actions, workflows, data, and audit trails to any AI agent, built by any vendor, through the Model Context Protocol standard. The agent can complete real work on the platform with full governance, rather than only answering questions about it. ServiceNow’s Action Fabric is the most prominent example shipping in May 2026.
Will agentic AI replace HR teams?
Not the team. Specific tasks will change. For example, onboarding triage, benefits queries, leave approvals, and case routing are already moving to AI Specialists in early-deploying companies. As a result, the HR team’s role shifts toward policy design, employee experience, and exception handling.
How should small companies prepare for an open MCP agent platform stack?
Start with an HRIS that supports API-first integrations. In addition, avoid vendors with closed data models. Keep your agent vendors open enough that your HR data can move between them. The 50-person company that picks an open stack today saves a re-platform project at 500 employees.
Not to be considered as tax, legal, financial or HR advice. Regulations change over time so please consult a lawyer, accountant or Labour Law expert for specific guidance.
