From Chat Interfaces to Autonomous AI Execution
AI in the enterprise is undergoing a major shift: from chat-style prompts to agents that execute work inside real business systems. Instead of asking a chatbot to draft a single asset, teams are now deploying workflow automation software that can plan, create, test, and report with minimal human intervention. This shift is redefining what buyers expect from AI agents enterprise offerings. Rather than isolated copilots embedded in one application, organizations want AI agent platforms that can operate across their existing stack and deliver measurable throughput gains. The conversation layer still matters for control and oversight, but the real value now lies in autonomous AI execution of repeatable tasks. As vendors race to keep up, the competitive question becomes less about whose model writes better copy, and more about whose agents can reliably move work from brief to outcome across applications.
Inside Opal: 42% ARR Growth and 4,000+ Customer-Built AI Agents
Optimizely’s Opal platform illustrates how quickly this new model is taking hold. The company reports a 42% quarter-over-quarter increase in annual recurring revenue for its AI agent orchestration offering, driven by deeper usage in marketing workflows. Nearly 1,700 customers using Opal have built more than 4,000 custom AI agents and run over 172,000 executions. Crucially, more than 97% of activity comes from customer-built agents rather than prepackaged assistants, showing that teams are investing in reusable automations tuned to their own processes. About 32% of those executions involve multi-step tasks, a strong signal that agents are being trusted with end-to-end work instead of one-off prompts. Together, these metrics suggest Opal is moving beyond pilot experiments and into scaled deployment, a key inflection point for any AI agents enterprise program.
From Content Generation to Workflow Automation Software
The Opal data reveals a broader directional change: enterprises want AI that executes processes, not just generates content. Optimizely points to downstream gains across its digital experience platform: more concluded experiments, a 26.4% experiment win rate, higher campaign production when its content marketing platform is paired with Opal, and increased digital asset reuse. These are classic workflow metrics rather than vanity AI usage numbers. They indicate that agent-based workflow automation software is helping teams increase throughput—more tests, more campaigns, more reuse—without matching headcount growth. In this model, AI agents don’t sit at the edge of work; they are embedded in the core lifecycle from insight to execution. That shift pushes budgets toward platforms that can prove faster cycle times, greater consistency, and expanded capacity, instead of tools that only improve drafting speed.
Agent Orchestration Becomes the New Enterprise Battleground
As AI agents enterprise adoption accelerates, orchestration is emerging as the competitive battleground. Marketing teams already rely on CMS, DAM, analytics, experimentation, and collaboration tools; the challenge is coordinating AI-driven work across all of them. Optimizely’s Opal emphasizes agents that can act across common systems like Salesforce, Google Analytics, Figma, and Atlassian. This elevates AI from a feature inside a single product to a process layer that spans the entire stack. In a crowded enterprise landscape that includes major digital experience vendors, buyers are asking whether a platform can support cross-tool orchestration and safely scale repeatable automation patterns such as brief-to-asset, insight-to-test, and test-to-personalization. The vendors that can prove higher execution rates and throughput, rather than just smarter prompts, are more likely to capture durable AI budgets.
What Enterprise Teams Need to Operationalize AI Agent Platforms
For organizations eyeing similar gains, adopting AI agent platforms is less a tooling decision and more an operating model change. Teams need to define approved outcomes for each workflow—what agents may publish directly versus what requires human review. Measurement must extend beyond model performance to track cycle time, rework, and adoption across each stage of the process. As customer-built agents proliferate, they effectively become internal products that require clear ownership, versioning, and documentation to avoid sprawl. Cross-tool orchestration also raises permissions and audit challenges as agents act inside CRM, analytics, design, and project management systems. Optimizely’s growth, coupled with rising weekly Opal usage and participation in its enablement programs, suggests enterprises are willing to tackle these challenges to unlock automation beyond chatbots—and to move decisively toward autonomous AI execution of business workflows.