Agents Execute. Control Planes Decide. Enterprises Adapt.

What happened
At Google Cloud Next 2026, Google’s most consequential announcements were not bigger models but tools to control them. InfoWorld notes that Google emphasised a “Knowledge Catalog” to ground agents in trusted data, an agent inbox in Gemini Enterprise to monitor long‑running processes, and new Workspace controls to audit and restrict agent data access. The message was clear: agents need supervision.

Why it matters
Enterprises are falling in love with AI agents yet are unprepared for rogue behaviour. Google’s move signals that the agentic era requires governance as much as innovation, without oversight, agents with budgets and credentials risk embarrassing or bankrupting organizations.

What’s next
Expect major cloud providers to compete on control planes rather than just model performance. Agent registries, monitoring dashboards and security audits will become table stakes as businesses insist on stopping, pausing or auditing agents at any time.

What happened
ConnectWise expanded its zofiQ platform for managed service providers, reporting that early adopters cut ticket escalations by 86 % and felt like they added two extra analysts without hiring. The platform now goes beyond recommendations: zofiQ automatically triages and routes tickets, performs administrative tasks and learns from real service data. A new Oversight Agent monitors tickets and agents for SLA compliance and risk.

Why it matters
This marks a shift from assistive bots to autonomous execution with built‑in supervision. Managed service providers can offload repetitive work while maintaining service standards—crucial for scaling operations without adding headcount.

What’s next
Expect similar “agent plus overseer” architectures to proliferate in IT service and operations. Vendors will pitch autonomy paired with governance as the formula for trusted agentic workflows.

What happened
Customer‑service software provider ASAPP launched a system of five purpose‑built agents within its Customer Experience Platform (CXP). The company said its goal was to deliver end‑to‑end AI‑powered customer service rather than a single chatbot. The five agents—Discovery, Developer, Simulation, Insights and Optimization—handle tasks from identifying automation opportunities to building agents, stress‑testing them, mining context graphs and continuously improving workflows.

Why it matters
Enterprises struggle to run AI agents consistently in production. By integrating multiple agents with governance, ASAPP promises more reliable deployment, faster setup and improved task completion consistency.

What’s next
Multi‑agent platforms that coordinate specialized agents will become the norm for complex CX environments. Vendors will differentiate on ease of deployment, agent interoperability and built‑in monitoring.

What happened: The Pentagon’s GenAI.mil platform, which serves 3 million eligible users, added access to Google’s Gemini 3.1 Pro model only eight weeks after commercial customers. Usage is soaring: 1.3 million users leverage the platform to automate tasks and summarize data, with examples such as building an automated database in three months instead of several years and drafting statements of work in hours instead of weeks. Users have built more than 100,000 AI agents via the platform’s Agent Designer tool, with authorizations for sensitive data and minimal coding required.

Why it matters
The Defense Department is rapidly operationalizing generative AI at scale. By onboarding Google’s frontier model and enabling agents to work with classified data, the Pentagon demonstrates how government agencies can keep pace with commercial innovation while building secure, domain‑specific workflows.

What’s next
With growing demand, expect DoD to prioritize agent governance and security. Future updates may include more model options and expanded use cases beyond document summarization and database automation.

What happened
Microsoft announced general availability of real‑time voice agents in Copilot Studio and rolled out new AI agents and features across Dynamics 365 Contact Center, Sales and Customer Insights. The Customer Assist, Quality Assurance and Service Operations agents handle high‑volume calls, evaluate interactions for quality and let leaders configure operations via conversation. In Dynamics 365 Sales, five new features—including a Sales Opportunity Agent and voice‑to‑CRM notes—automate research, data enrichment and post‑call updates.

Why it matters
By embedding real‑time voice in its platform, Microsoft moves beyond text chat to more natural, multilingual customer interactions. Automating routine tasks frees CX and sales teams to focus on empathy and strategy.

What’s next
As voice becomes the primary interface for customer service, expect more advanced voice agents with cross‑channel memory and integration with existing telephony systems. Competitors will likely respond with their own voice‑enabled agentic suites.

What happened
JumpCloud unveiled Agentic IAM, a control plane that registers and manages every human, non‑human and autonomous agent as a corporate identity. The platform discovers agents across an organization, provisions them with appropriate credentials and enforces zero‑trust policies so agents can only access authorized resources. It verifies the health of hardware across operating systems and includes human‑in‑the‑loop checkpoints for high‑impact actions.

Why it matters
As autonomous agents proliferate, they become high‑velocity identities with access to sensitive systems. Traditional IAM tools cannot track these dynamic entities. JumpCloud’s approach treats agents like employees, bringing visibility and governance to the entire agent lifecycle.

What’s next
Identity management vendors will rush to support non‑human identities. Look for standards around agent discovery, credential rotation and agent‑to‑agent trust to emerge as enterprises demand secure AI workflows.

What happened
The NextPlatform reports that the AI landscape is pivoting from chasing larger models to building agentic platforms that orchestrate agents, data and hardware. A PwC survey of 300 executives found 88 % plan to boost AI budgets and 79 % have already adopted agents, with 66 % citing productivity gains. At Google Cloud Next, CEO Thomas Kurian said more than 75 % of Google Cloud customers use AI products, and the company plans to invest over $175 billion in ML infrastructure this year. Google unveiled an expanded Vertex AI platform (Gemini Enterprise Agent Platform) featuring an Agent Studio, Agent Development Kit and a Memory Bank for long‑running agents, along with centralized identity, registry and gateway tools to monitor and secure agents.

Why it matters
The rapid cadence of model releases and the proliferation of agents mean enterprises need a unified stack—chips designed for models, models grounded in enterprise data and agents secured by infrastructure. Without such integration, organizations accumulate technical debt and expose themselves to security risks.

What’s next
Expect major vendors to bundle models, orchestration frameworks and governance features into cohesive platforms. Enterprises will evaluate AI providers not just on model performance but on the robustness of their agentic stack and control mechanisms.

What happened
Chinese service‑robot leader Pudu Robotics opened its U.S. headquarters in Dallas on April 27. The new facility centralizes operations, product showrooms and warehousing, enabling more efficient logistics and regional coordination. Pudu has deployed nearly 15,000 robots across the Americas, fueling 285% year‑over‑year revenue growth. Its product lineup spans service delivery robots like BellaBot, commercial cleaners such as the CC1 and BG1 series, industrial delivery bots and general embodied‑AI robots.

Why it matters
The investment signals continued confidence in service robotics despite economic headwinds. By localizing manufacturing and support, Pudu aims to penetrate sectors ranging from hospitality and healthcare to logistics and retail.

What’s next
As competition intensifies, expect service robot vendors to differentiate through regional manufacturing, integrated AI platforms and partnerships with major enterprises. Expansion into new industries like industrial delivery and large‑facility cleaning will test the scalability of these embodied AI systems.

What happened
Japan Airlines (JAL) announced the phased rollout of humanoid robots for ground handling at Japanese airports, starting May 2026, in partnership with GMO AI & Robotics Trading. The robots will automate baggage loading, cargo movement, and cabin cleaning.

Why it matters
With Japan’s aviation sector facing a severe worker shortage, humanoid robots could bridge the gap and set a precedent for broader adoption in logistics and transportation.

What’s next
The project will expand in phases, with robots gradually taking on more complex and physically demanding tasks alongside human staff.

Agents are no longer just suggesting—they’re executing, with real access to systems, data, and decisions. The winners won’t be those with the smartest models, but those who can control, monitor, and trust what their agents actually do.

Build your own “agent + oversight” stack in under an hour. Use Google Cloud Vertex AI Agent Builder (or Gemini tools) to create a simple task agent—like summarizing emails or routing tickets. Then layer in governance: simulate identity control with JumpCloud (define what your agent can access), and add a human-in-the-loop checkpoint using Microsoft Copilot Studio or a simple approval workflow. Bonus: map how an MSP would automate this using ConnectWise-style ticket flows.