Three-quarters of organizations now use agentic AI to some degree, and nearly all expect to deploy AI agents within two years, according to a Deloitte survey of more than 3,200 executives released earlier this year. The most advanced companies aren't just adding AI agents to existing workflows—they're rethinking entire business processes and models around the technology. Twenty percent of enterprises surveyed already report increased revenue from the shift.
The survey reveals a sharp acceleration in adoption timelines and integration depth. Today, 20% of organizations say they're using agentic AI moderately, 2% extensively, and 1% have fully woven it into core operations. Within two years, 46% expect moderate use, 23% extensive deployment, and 5% full integration as a key business component. Meanwhile, 30% of enterprises are already redesigning critical processes around AI, while 34% are using it to fundamentally transform their business. The technology is showing transformative impact on efficiency and productivity at more than double the rate from a year ago, according to Deloitte's findings.
AT&T achieved "five-X cash ROI" on its AI investment last year, says Andy Markus, the company's chief data and AI officer. The telecom giant has built its own model-agnostic AI platform and now has more than 1,000 agentic workflows either in production or under development. New York Life Group Benefit Solutions is taking a similar path, building a custom AI business engine using cloud-native capabilities and open source tools rather than relying on embedded AI from commercial platforms. Matt Marze, CIO at New York Life Group Benefit Solutions, says the group has deployed agentic AI for compliance audits, which traditionally required labor-intensive evidence gathering, saving significant time.
The shift from chatbots to agents represents a fundamental change in what AI can do for businesses. November 2024 marked a major transition point when Anthropic released MCP as an open standard, allowing AI models to directly interact with data sources and tools rather than just answer questions. This capability lets AI carry out business processes that were previously too complex, unpredictable, or costly to automate. But deployment comes with rising costs—Gartner reported in August that AI inference costs per agentic workflow will increase more than five times through 2028, as agents repeatedly query models. AT&T is managing this by routing requests through an AI gateway that directs simpler tasks to older, cheaper models, and building small language models that can be just as accurate for specific use cases while saving 90% of the cost.
Companies that build their own AI infrastructure early are positioning themselves ahead of competitors who wait for vendor solutions to mature. AT&T's strategy is to solve problems at the leading edge now so the technology becomes standard practice by the time it's commonplace elsewhere. New York Life Group Benefits is working on a persona-based command center experience that will replace the need for employees to navigate multiple work systems directly, orchestrating tasks and surfacing priorities based on role and objectives. The company plans to roll out the first version later this year. Organizations face a choice between adopting vendor platforms and building custom solutions—a decision that will likely determine competitive positioning as AI costs, governance demands, and the complexity of multi-agent systems continue to grow.