For most of 2025, the enterprise AI conversation was largely about potential. Pilots were running. Proof-of-concepts were generating impressive demos. Strategy decks were filled with roadmaps that pushed meaningful deployment to some unspecified future quarter.
That phase is over.
In the space of a single week in late April 2026, OpenAI, Anthropic, and Google each made significant enterprise-grade agent moves. GPT-5.5 shipped on April 23, built specifically for autonomous planning and execution across tools. Claude Managed Agents moved into public beta with session-level pricing designed for persistent, long-running deployments. Google's Gemini Enterprise Agent Platform, rebranded from Vertex AI, now supports business-user-built agents across the full Workspace suite without requiring engineering involvement.
The competitive window for pilots is closing. What comes next is platform selection, and that decision will be harder to undo than most enterprise technology leaders currently appreciate.
What Each Major Player Is Actually Competing On
The vendor narratives are diverging in ways that matter for how you evaluate them.
OpenAI is competing on capability and reach. GPT-5.5 is designed to plan, execute, verify, and iterate across tools without constant human oversight. The launch targeted enterprise buyers with a clear message: this is no longer about interactive conversations, it is about deploying intelligence that works autonomously. The dropdown integration inside ChatGPT for Plus, Pro, Team, and Enterprise users means the adoption path is remarkably frictionless. Millions of employees already have the interface; the agent mode is now a selection, not a procurement process.
Anthropic is competing on reliability and developer control. Claude Opus 4.7 reclaimed benchmark leadership in coding tasks, and Claude Managed Agents provides a hosted infrastructure layer for running agents without organisations needing to manage their own sandboxing, state management, or session persistence. The pricing at $0.08 per session-hour plus tokens makes the cost of a persistent, long-running agent deployable and predictable. For enterprise technology teams that want control without infrastructure overhead, this is a meaningful offer.
Google is competing on integration depth and distribution. Google Cloud's renamed Gemini Enterprise Agent Platform combines Workspace Studio, the agent-building tool that works across Gmail, Docs, Sheets, Drive, Meet, and Chat, with the broader Vertex infrastructure. Google's own AI Agent Trends report found that 89% of business teams are already using AI agents and that the average organisation runs 12 simultaneously. If that data is even directionally accurate, the question is no longer whether to deploy agents, but how to govern the ones already running.
The Consolidation Question
Three major platforms competing simultaneously in enterprise agents looks like healthy competition. It is also a vendor lock-in risk of the first order.
Enterprise agent deployments are not plug-and-play. Agents are connected to internal systems, trained on proprietary data, integrated into workflows, and embedded into team habits. Moving an agent from one platform to another is not equivalent to switching a SaaS subscription. The cost is in the rewiring, the re-training, and the institutional disruption of pulling out a system that has become operational.
Technology leaders making platform decisions now need to think of this not as picking a vendor but as picking an ecosystem. The A2A protocol, Google's Agent-to-Agent interoperability standard, is an attempt to address this concern by enabling agents built on different platforms to collaborate. Whether it delivers genuine interoperability or becomes a proprietary moat dressed as openness will determine how much of the market follows Google into its architecture.
What This Means for Enterprise Technology Leaders
What should enterprise technology leaders do now about AI agent platform selection?
The honest answer is that there is no clean, risk-free choice. But there are three things that reduce the strategic risk of this decision. First, define your evaluation criteria before you evaluate: capability benchmarks matter, but integration depth, governance controls, data residency, and vendor stability matter more for most enterprise environments. Second, avoid full consolidation onto a single platform until the interoperability question is more settled. Running agents across two platforms in parallel is operationally complex, but it preserves optionality in a market that is moving quickly. Third, treat the agent governance question as seriously as the agent deployment question. Google's own data shows that the average enterprise is running 12 agents simultaneously; the companies that know what their agents are doing, what data they are accessing, and how they are escalating failures will be in a significantly better position than those that deployed first and governed later.
The Signal Beneath the Headline
The pace of this week's releases is itself the signal. OpenAI, Anthropic, and Google do not coordinate their product launches. Three major agent announcements in the same week means each of them is responding to the same competitive pressure: the sense that the enterprise platform battle is entering a decisive period and that the organisations that capture developer and IT decision-maker commitment now will be difficult to displace.
For enterprise technology leaders, that urgency is real but should not translate into rushed decisions. The organisations that took a structured approach to cloud platform selection in 2013 and 2014 ended up in a better position than those that defaulted to whoever was in the room. The same principle applies here.
For investors in listed technology companies, this week represents a meaningful data point on enterprise AI adoption velocity. The supply side, the platforms, are clearly ready for scale deployment. The constraint on growth is now enterprise readiness and governance capability, not model quality or platform availability. Companies that can demonstrate genuine enterprise agent deployments with measurable productivity or cost outcomes will be the ones with a differentiated investor narrative in 2026 and beyond.
At ARC, we work with technology companies and ASX-listed operators to translate complex technology shifts into clear, credible investor and market communications. The AI agent transition is one of the most consequential platform shifts in a decade, and how companies communicate their position in it will shape how investors value them through the cycle.
The Practical Implication for ASX-Listed Technology Companies
Listed technology companies with AI agent exposure need a clear investor narrative now. Institutional investors are increasingly capable of distinguishing between companies with genuine agent deployments and those with AI aspirations dressed as strategy. The questions are specific: what agents are running, what work are they doing, what is the measurable output, and what is the path to revenue recognition.
If your investor communications on AI are still at the strategy and roadmap level, the gap between your narrative and what your peers are demonstrating is growing. The time to sharpen the story is before the next earnings cycle, not during it.