From Flagship Demos to Everyday Work: Lessons from Google Cloud Next AI
At Google Cloud Next AI, the narrative around enterprise AI agents shifted decisively from experimentation to operational impact. Google Cloud showcased Macy’s “Ask Macy’s” agent as evidence that sophisticated, customer-facing agents can move from idea to production in days rather than months when built on platforms such as Gemini Enterprise for Customer Experience. Embedded near the search bar on Macy’s website, the agent supports natural, conversational queries that span discovery, purchase, service and store fulfillment in a single interaction. It can, for example, recommend outfits based on personal context, check real-time inventory and secure in-store pickup, blending digital and physical commerce into what Google calls “fluid commerce.” Early signals are encouraging: Macy’s reports larger order sizes among shoppers who engage with the agent and is tracking long-term metrics like conversion, units per transaction and revenue to gauge durable value.
Why Scaling AI Agent Adoption Demands a Culture Shift
Despite impressive pilots, broad AI agent adoption inside enterprises is hitting human roadblocks. The Macy’s example shows that technical speed is solvable, but sustained value depends on user trust and behavior change. Customers need confidence that agents understand context, respect preferences and reliably resolve issues, not just redirect them to generic FAQ pages. Internally, employees must see agents as partners embedded into existing workflows rather than opaque black boxes or job threats. That requires deliberate change management: clear communication on goals, training so staff can craft effective prompts and interpret responses, and governance so decisions are auditable. Leaders also need to define where agents augment versus automate human work. Without this cultural foundation, organizations risk a proliferation of disconnected bots that people bypass or distrust, undermining even the most advanced underlying models and infrastructure.
The Rising Problem of Agent Sprawl in the Enterprise
As enterprise AI agents move beyond single flagship deployments, a new problem is emerging: agent sprawl. Different teams are rolling out their own specialized bots for development, security, operations and customer experience, often tied to distinct SaaS platforms and internal systems. According to Band, an organizational infrastructure startup, organizations are beginning to run dozens or even hundreds of agents, yet coordination is quickly becoming a bottleneck. Forecasts cited by the company suggest that roughly half of agent deployments could fail because of weak runtime governance and poor interoperability between systems. Many enterprises still rely on manual processes to pass context from one bot to another, creating fragile, non-scalable workflows. The result is an environment where agents cannot easily share information, policies are inconsistently enforced and leaders lack a unified view of who—or what—is acting on critical business data.
Inside Band’s ‘WhatsApp for AI Agents’ Vision
Band is positioning itself as a communication and collaboration layer designed specifically for enterprise AI agents—a kind of “WhatsApp for AI agents.” Founded by serial entrepreneur Arick Goomanovsky and multi-agent systems veteran Vlad Luzin, the company argues that the future lies in many smaller, specialized agents working together rather than single, monolithic bots. To make that architecture viable, agents must communicate with each other and with humans in secure, real-time, two-way channels. Band’s platform provides organizational infrastructure to manage and coordinate agents across domains, connecting internal enterprise systems, SaaS applications and external partners. A central control center enforces policies, defines permissions and improves transparency, making actions more traceable and auditable. With initial revenues and about ten partners already, Band aims to reduce the friction of multi-agent collaboration so enterprises can scale automation without losing oversight or control.
What CIOs Should Prioritise for AI Agent Orchestration in 2026
For CIOs, the next phase of enterprise AI agents is less about creating one impressive assistant and more about disciplined orchestration. First, treat AI agent orchestration as core architecture: define standards for communication, handoffs and shared context so agents can collaborate reliably across departments. Second, invest in a governance and communication layer—whether homegrown or via platforms like Band—that unifies policy enforcement, permissions and monitoring. Third, tie AI agent adoption to explicit business outcomes, mirroring Macy’s focus on metrics such as conversion, units per transaction and customer value. Fourth, plan for training and change management, ensuring employees understand how to work with agents and when to override them. Finally, security and auditability must be non-negotiable; as agents trigger real-world actions, leaders need clear visibility into what each agent did, on whose behalf and under which policy constraints.