‘Agentic AI race’—why it matters as Oracle and Salesforce reshape work. What CIOs should ask now.

Sam Donaldston
oracle salesforce agentic ai race

As markets climbed this week, investor attention turned to enterprise software leaders betting on “agentic” artificial intelligence, a model that promises software agents that perform tasks with minimal human input. Constellation Research CEO R “Ray” Wang told Fox Business viewers that Oracle, Salesforce and others are setting the pace, a push that could change how companies sell, support and secure their operations.

Constellation Research CEO R “Ray” Wang joins Varney & Co. to explain how Oracle, Salesforce and others are leading the emerging “agentic” AI race, reshaping business operations and fueling investor optimism as markets climb.

The discussion centered on who is out front, what these tools actually do and why the shift is drawing money back into large-cap software. It also raised questions about risk, talent and timelines for real returns.

From copilots to agents: a shift in enterprise AI

Agentic systems move beyond prompts and chat. They plan steps, call tools and complete work flows. In sales, that might look like drafting proposals, scheduling follow-ups and updating records without human clicks. In finance, an agent can reconcile payments and flag exceptions before month-end.

Salesforce has pushed into this space through its Einstein platform and newer agent features that tie into CRM data and automation. Oracle has focused on embedding generative models and agents across Fusion applications and its cloud infrastructure. Both aim to cut routine work and speed decisions inside business systems that companies already use.

Wang’s framing suggests a near-term race to productize these agents in customer service, sales and operations, where time saved is easy to measure.

Why investors care right now

Public software valuations often follow proof that new tech can drive growth or margin gains. Agentic AI promises both. If agents reduce support handle times, close more deals or shrink back-office labor, revenue per employee can improve.

That math, even at small percentages, supports higher earnings forecasts. It also builds a services and consumption flywheel for cloud providers that host and fine-tune these systems.

Markets have rewarded firms that show concrete use cases, ecosystem depth and security controls. Oracle’s cloud capacity and application footprint, and Salesforce’s domain depth in customer data, fit that playbook.

Benefits and early roadblocks

The pitch is simple: automate repetitive steps, reduce errors and free staff for higher-value work. Contact centers use agents to summarize calls and propose resolutions. Field teams get automatic next-best actions. Finance teams see draft journal entries and exception queues prepared in advance.

But buyers are testing limits. Hallucinations, data leakage and compliance issues remain top concerns. Cost control is another, as agent loops can run many model calls and rack up bills without careful guardrails.

  • Security and governance: Who approves what an agent can do?
  • Quality: How are outputs reviewed and corrected?
  • Cost: How are tokens, APIs and retries monitored?

Enterprises are answering these with human-in-the-loop reviews, role-based permissions, audit trails and fixed-price tiers from vendors.

How Oracle and Salesforce differ

Oracle is threading agents across ERP, HCM and supply chain, backed by its cloud’s data and GPU capacity. The pitch stresses performance, data residency and lower total cost by running on its stack.

Salesforce is anchoring agents in CRM, where customer context is richest. Its approach leans on metadata, connectors and prompts tuned to sales and service. Partners build domain agents on top, giving customers industry-specific options.

Both are courting developers with tools to define actions, test workflows and set policies. Success may depend less on model size and more on how cleanly agents plug into real business processes.

What to watch next

Wang’s appearance signals that leaders expect 2025 to be a year of deployment, not demos. That will require clearer metrics and safer defaults. Buyers want proof of fewer tickets, faster cycle times and lower costs, not just better chat.

Standards will matter, too. Expect stricter audit logging, incident reporting and evaluation benchmarks for agent behavior. Vendors that offer simple dashboards showing savings, risks and overruns will gain trust faster.

For now, pilots are expanding where data is structured and outcomes are easy to score. Sales operations, IT service management and invoice processing are early winners.

CIOs and COOs weighing the shift can start small, pick measurable tasks and tie budgets to defined KPIs. That disciplined approach will separate hype from value as the “agentic” wave hits core systems.

The bottom line: the race is real, the tools are improving and the next few quarters will show which platforms turn interest into durable gains. Watch for adoption in cost centers, clearer ROI proofs and stricter controls as companies put agents to work.

Sam Donaldston emerged as a trailblazer in the realm of technology, born on January 12, 1988. After earning a degree in computer science, Sam co-founded a startup that redefined augmented reality, establishing them as a leading innovator in immersive technology. Their commitment to social impact led to the founding of a non-profit, utilizing advanced tech to address global issues such as clean water and healthcare.