Beyond Apps: Multi-Platform AI Agents Are Here

AI agents took center stage at Workday’s annual Rising conference earlier this month, signaling how quickly enterprises may shift from siloed applications to agent-driven workflows. 

Workday showcased a cross-platform AI agent where an employee could interact with a self-service agent built in Microsoft Copilot Studio. That Copilot agent then exchanged information with a Workday agent to complete the user’s career-goal tasks, without the employee ever logging into Workday directly. During processing, another agent might notify the user in Microsoft Teams if the goals require refinement or are approved.

The key enabler? Workday announced that AI agents developed on the Microsoft platform can now be registered in the Workday Agent System of Record (ASOR). This effectively treats each agent like an employee, assigning security roles based on least privilege and managing its lifecycle.

This shift raises a set of practical and strategic questions for IT executives, enterprise architects, project managers, and developers:

1.     Scope of Agent Deployments. 

How will IT organizations supporting multiple platforms (Workday, Microsoft, Salesforce, etc.) handle cross-platform agents? Users naturally develop preferences for the tools they find most productive. If some teams use Microsoft Copilot to integrate with Workday, others may demand similar capabilities from Slack or another channel. Project scoping, operational support, and business cases should proactively account for likely expansion of agent use cases.

2.     Bidirectional Flow of Information.

The opportunity isn’t just about a user submitting a request. Cross-platform agents – and downstream agents – can analyze data, detect anomalies, and raise proactive recommendations back to users. However, people will want follow-ups in different places (Copilot, email, Workday tasks, Teams, etc.). How can organizations flex their business processes yet maintain control over the bidirectional flow of information? What new safeguards are needed as data crosses multiple platforms?

3.     Multiple Agent Registry Options.

Workday’s ASOR is one model, but it’s not the only one.  Organizations already use tools like ServiceNow AI Control Tower, Salesforce Agentforce, and SAP’s LeanIX AI Agent Hub. As vendors converge human and AI agent management, the question becomes: which “system of record” should you trust  – or will different registries coexist across the enterprise?  At the same time, professional services firms are rolling out their own frameworks (PwC’s Agent OS, KPMG’s AI Agent Hub).  Rather than a single “home” for all agents, many organizations may need to coordinate across multiple registries.

4.     AI Agent Management of Human Resources.

Several demos showed agents performing tasks traditionally done by supervisors. In one case, an agent handled front-line scheduling when an employee called in sick, checking availability, skills, recommending a substitute, and coordinating communications automatically. Another example: an agent aggregated information from multiple systems and drafted an employee performance review. Because agents can run on schedules or continuously monitor conditions, organizations may soon shift from employees augmented by AI to employees managed by AI. Which functions will transform first?



5.     Agent Performance Management and Lifecycle.

As agents take on increasingly critical roles – supporting employees, managing workflows, and even making supervisory recommendations – their performance will need to be continuously monitored.  Different agents will be built on different models, tuned for specific use cases, but not all will interact smoothly.  Data handoffs across multiple agents risk producing the AI version of the human “telephone game,” where messages get distorted downstream.  To manage this complexity, organizations will require robust metrics and monitoring tools to track utilization, effectiveness, and redundancy. Underperforming agents may need to be retrained, consolidated, or decommissioned – much like employee offboarding – through a formal lifecycle process.

Why It Matters

The proliferation of AI agents, the abundance of deployment options, and the diversity of user interaction preferences promise to reshape how work gets done. For IT leaders and enterprise architects, agility – in both technology and governance – will determine who benefits and who struggles. Flexible, secure, and compliant architectures are no longer optional; they’re the foundation for realizing the promise of cross-platform AI agents.