The Great Agent War Is Here
Salesforce reported $900 million in AI and Data Cloud revenue within six months of launching Agentforce. ServiceNow announced a three-year OpenAI partnership for enterprise-grade autonomous agents. Microsoft Copilot is embedded across the Office suite. Google Vertex AI Agent Engine continues to drop prices.
AI agents are no longer a developer experiment. They are a C-suite budget item.
This is great news for the industry. It validates that enterprises will pay for AI agents that deliver business outcomes. But it also raises a question that no enterprise incumbent has answered:
Who governs the other 12 departments?
The Module Problem
Salesforce Agentforce serves three departments: Sales, Service, and Marketing. These are the departments where Salesforce has data, workflows, and customer relationships. Agentforce is excellent at what it does — using CRM data to help sales reps close deals and support agents resolve tickets.
ServiceNow serves two departments: IT and HR. Their autonomous workforce agents troubleshoot IT issues and handle employee requests. Again, excellent within their domain.
Microsoft Copilot covers Office-centric workflows: email drafting, document summarization, meeting notes. Useful, but not department-specific automation.
That leaves at least 10 departments with no agent coverage from their existing vendors:
- Legal — contract review, compliance monitoring, policy drafting
- Finance — expense analysis, forecasting, audit preparation
- Operations — process optimization, supply chain, resource allocation
- Engineering — code review, documentation, incident response
- Product — user research synthesis, roadmap planning, feature specs
- Data & Analytics — report generation, data quality, insight extraction
- Executive — board prep, strategic summaries, cross-org dashboards
- R&D — literature review, experiment tracking, patent drafting
- Customer Success — health scores, renewal prep, onboarding workflows
- Procurement — vendor evaluation, contract negotiation, spend analysis
These departments have real, urgent needs for AI automation. They are not going to wait for Salesforce to build a legal contract review agent or for ServiceNow to build a finance forecasting agent. Those products are not on any incumbent’s roadmap.
Four Structural Limitations of Platform Modules
The problem with bolting agents onto existing platforms goes deeper than department coverage:
1. Vendor-Locked Data
Agentforce agents can only access Salesforce data. ServiceNow agents can only access ServiceNow data. If your legal team’s contracts are in DocuSign and your financial data is in NetSuite, platform agents cannot help.
A standalone governed agent platform works with any data source — knowledge bases, APIs, documents, spreadsheets, databases — regardless of which vendor hosts them.
2. Inherited Governance Limitations
Salesforce governance means Salesforce permissions, profiles, and sharing rules. These were designed for CRM access control, not AI agent oversight. They do not include approval gates for agent actions, budget limits on token spending, prompt injection detection, or graduated autonomy levels.
Purpose-built governance for AI agents needs layers that CRM and ITSM platforms were never designed to provide: human-in-the-loop approval, model provider fallback, cost estimation before execution, and audit logging of every LLM interaction.
3. Siloed Agents That Cannot Coordinate
Agentforce agents cannot coordinate with ServiceNow agents. There is no shared memory between a sales agent that learned customer preferences and an IT agent troubleshooting that customer’s technical issue. Each platform’s agents operate in isolation.
Cross-department institutional memory — where a legal agent can reference the context that a finance agent built during contract negotiation — requires a unified memory layer that sits above any single platform.
4. Platform Prerequisite
No Salesforce subscription? No Agentforce. No ServiceNow subscription? No AI agents. This means enterprises must maintain expensive platform subscriptions just to access their AI agent capabilities, even for departments that do not use the core platform.
A standalone agent platform has no prerequisite. It runs on cloud, hybrid VPC, or fully self-hosted.
Who Should Use What
Let us be honest about when platform agents make sense:
Use Agentforce if your primary need is AI-powered sales and support automation and you already have a Salesforce deployment. Agentforce has deep CRM data integration that a standalone platform cannot match.
Use ServiceNow AI agents if your primary need is IT service management and employee support automation and you already run ServiceNow. Their ITSM integration is best-in-class.
Use a standalone governed platform if you need AI agents across Legal, Finance, Operations, Engineering, Product, and the other departments that CRM and ITSM vendors do not cover — or if you need a unified governance layer across all 20 departments with provider-agnostic model choice.
The two approaches are not mutually exclusive. Many enterprises will run Agentforce for their sales team AND a governed agent platform for the other 12 departments. The key is not letting the CRM tail wag the enterprise AI dog.
The Governance Gap
Beyond department coverage, there is a fundamental governance gap. Enterprise AI agents need oversight that goes beyond platform permissions:
- Approval gates that pause agent execution for human review
- Budget controls with per-department, per-workflow token spending limits
- Model freedom to switch between providers without rewriting agents
- Prompt injection detection that runs during execution, not as an afterthought
- Graduated autonomy where agents earn more independence as trust builds
- Institutional memory where agents learn and share context across departments
These capabilities require a platform purpose-built for AI agent governance — not CRM or ITSM governance repurposed for agents.
What Comes Next
The enterprise agent market is entering its next phase. The first phase was “can we build AI agents?” The second phase, happening now, is “who governs them across the entire organization?”
Salesforce and ServiceNow have answered the first question for their domains. The second question — governed AI for all 20 departments — is what standalone agent platforms are built to solve.