GlobalData reckons autonomous AI agents are exposing the limits of traditional graphical user interfaces across software development, cloud operations, automation and digital infrastructure.
Many AI agents still depend on browser automation, visual navigation, UI parsing and cursor-based workflows. That is a bit daft when those systems were built for humans squinting at screens, not autonomous software trying to execute tasks at scale.
The result is a push toward AI-native command-line interface environments and agent-native architectures. These systems are optimised for structured, machine-readable execution rather than making agents pretend to be interns with a mouse.
GlobalData disruptive tech practice head Kiran Raj said the old software model was built around humans seeing, clicking and interpreting.
“Enterprise software was built for people to see, click, and interpret. Autonomous AI agents need structured, direct, and reliable execution paths. GUI-led workflows force agents to spend effort on perception before action, making automation brittle at scale. AI-native command-line interface (CLI) environments give agents a cleaner command layer to connect with APIs, orchestration systems, and enterprise infrastructure.”
GlobalData disruptive tech project manager Saurabh Daga said the interface layer is becoming critical as autonomous AI moves out of the lab.
“As autonomous AI moves from experimentation to enterprise execution, the interface layer becomes critical. AI-native CLI environments provide structured, composable, and machine-readable paths for agents to execute work directly, improving orchestration reliability, enterprise interoperability, and automation across cloud, DevOps, software engineering, and business operations.”
The shift reflects a wider move toward control-plane architectures. In that model, software interaction happens through APIs, orchestration systems, AI agents and intent-driven execution layers rather than manually operated dashboards.
Humans define operational intent, AI agents decide the execution logic and orchestration layers push workflows across enterprise infrastructure.
GlobalData’s Innovation Explorer database points to more enterprise interest in agents that orchestrate workflows, manage infrastructure, process transactions and execute operations through structured systems.
The outfit cited Datris’ agent-operated data platform, AppZen’s agent-native accounts payable platform and Skai’s agent-native marketing operating system. It pointed to FPT AI Factory’s agent-native commerce infrastructure platform, launched with InFlow and Visa Intelligent Commerce.
Momentum is building around AI-native CLI tools and agent-focused operational environments.
Google’s Antigravity CLI, GitHub’s Copilot CLI and OpenAI’s Codex agent-loop work all point toward command-line, terminal-native and iterative execution models. These are being aimed at autonomous coding, orchestration, testing and software operations.
Daga said agent-native architectures and AI-native CLI environments could push enterprises toward autonomous operational ecosystems.
“Agent-native architectures and AI-native CLI environments are enabling enterprises to move toward autonomous operational ecosystems where AI agents increasingly manage execution across software and infrastructure. However, broader adoption will depend on reducing orchestration fragility, improving execution reliability, and standardising machine-readable interfaces across enterprise systems.”
Enterprise software has decades of bolted-on interfaces, weird permissions and brittle integrations hiding behind the glossy dashboards. Agents may not care about pretty buttons, but they will still trip over rubbish plumbing unless vendors stop selling vibes and start exposing proper machine-readable controls.
