From Chat Windows to Workflow Execution Layers
Enterprise AI strategy is pivoting from conversational assistants to agentic systems that can actually move work across tools. Instead of treating large language models as smart text boxes, vendors are embedding AI agents directly into business process automation. These agents can trigger actions, update records, launch experiments, or assemble content without human operators nudging them prompt by prompt. This shift is reshaping how platforms position value: not "write my email faster," but "run this campaign end to end and report back." The result is a new category: AI agents workflow automation that lives inside existing stacks rather than beside them. As enterprises tire of proof‑of‑concept chatbots, they are demanding production‑grade automation tied to metrics like throughput, experiment volume, and campaign output—moving AI from novelty interfaces to core execution infrastructure.
Optimizely’s Opal Shows What Agentic AI at Scale Looks Like
Optimizely’s Opal AI agent orchestration platform illustrates how quickly enterprises are embracing workflow-native AI. The company reports 42% quarter-over-quarter ARR growth for Opal, with nearly 1,700 customers building more than 4,000 custom AI agents and running over 172,000 executions across marketing workflows. Crucially, more than 97% of activity comes from customer-built agents rather than prepackaged assistants, signaling that teams are investing in bespoke enterprise software agents tailored to their own stacks. Around 32% of executions involve multi-step tasks, indicating that agents are handling end-to-end flows, not just single prompts. Downstream metrics tell the same story: more concluded experiments, higher campaign production, and increased digital asset reuse when Opal supports Optimizely’s content tools. In practice, Opal is less about content generation and more about business process automation AI that can span CMS, DAM, analytics, and experimentation systems in a governed way.
Agentic AI Platforms Become a Monetization Engine
As AI moves into workflows, vendors are reshaping commercial models around agent-based platforms. Salesforce’s leadership has explicitly framed its AI agent offerings as a “very high margin opportunity,” signaling that the next phase of SaaS monetization revolves around embedded agents rather than add-on chat features. The company is also leaning heavily on coding agents to accelerate how it builds and implements its own software, tightening the feedback loop between internal efficiency and productized automation. While some deals reportedly include capped-price access to AI agents, analyst warnings about renewals highlight the emerging tension: long-term revenue will depend on how deeply customers bake agentic AI platforms into critical processes. For buyers, this reinforces a key evaluation question: will AI agents workflow automation simply reduce manual effort, or will it fundamentally change how revenue, support, and marketing engines operate—and be priced—over time?
Legacy Platforms Rebundle AI Into Customer Experience Workflows
Incumbent platforms across CRM, marketing, and collaboration are increasingly weaving AI agents into the customer experience itself. Instead of selling standalone copilots, vendors are bundling AI into workflows that span sales outreach, service routing, marketing orchestration, and analytics. Zoom and other established providers are turning AI features into integrated steps inside meeting, support, and engagement journeys, creating new revenue streams tied to usage and outcome-based value, rather than simple seat licenses. This rebundling changes the adoption pattern: users do not have to learn a new AI tool; they interact with business process automation AI embedded in the flows they already rely on. For the platforms, it is an opportunity to increase stickiness and expand wallet share by owning not just the data and interface, but the automation logic that connects them.
From Experiments to Production-Grade Automation
The move from conversational interfaces to autonomous agents reflects a broader maturation in enterprise AI adoption. Early chatbot deployments were often experiments, judged on novelty and user satisfaction more than on hard operational metrics. Agentic AI platforms, by contrast, are evaluated on execution rate, cycle times, and capacity gains—how many campaigns shipped, how many experiments were concluded, how much reusable content was activated. Optimizely’s Opal data suggests organizations are comfortable letting AI carry multi-step tasks to completion, forcing investments in governance, QA, and standardized approval flows. Meanwhile, the debate around a possible “SaaS‑pocalypse” underscores a deeper question: if coding and workflow agents can assemble entire applications, what is the long-term role of traditional SaaS vendors? For now, the winners are those who treat agents as the orchestration fabric of their ecosystems, not as isolated AI features.