The Agent Era Is Here, and the Price of AI Keeps Falling

The biggest story in AI right now is not a new model. It is that AI agents stopped being demos and started doing real work. OpenAI, Anthropic, Google and Microsoft all ship agentic products that businesses can hand actual tasks to, and the cost of running them keeps dropping.

Agents moved from chatbot to employee

Claude Code from Anthropic and Codex from OpenAI are now everyday tools for developers. Google's Gemini agents work inside Workspace, drafting email chains and updating spreadsheets. Microsoft pushed Copilot agents deep into its business software. Each product does multi-step work: gather information, make a decision, complete a task, report back. That is a different category from the chatbot that answers one question at a time.

The pattern is the same across all of them. The model is given a goal, a set of tools, and permission to work until the job is done. Early versions were unreliable. The current generation handles long tasks with far fewer mistakes, and companies are starting to trust them with routine work.

Intelligence got cheap, and that changed everything

DeepSeek's open-weight R1 model, released in early 2025, proved that reasoning models could be built without billion-dollar budgets. The response from OpenAI, Anthropic and Google was a rapid series of price cuts and faster, cheaper models. By 2026 the cost of a capable model call is measured in fractions of a cent.

Cheap reasoning changes the business case for automation. Tasks that were too small to justify software development are now worth automating. A cafe can have an AI draft its weekly menu descriptions. A plumber can have one write clear quotes from a few notes. The economics work for businesses of every size, not just enterprises.

Regulation is arriving in phases

The EU AI Act's obligations for general-purpose AI models began applying in August 2025, with stricter rules phasing in through 2026. Australia has taken a lighter approach: a voluntary AI safety standard published in 2024, alongside sector guidance. Australian businesses are not facing hard compliance deadlines yet, but the direction is clear. Anyone using AI for customer-facing work should expect transparency and human oversight to become expected practice.

The practical takeaway

For Australian small businesses, the useful question is no longer whether to use AI. It is which tasks can be handed over safely. Start with drafting, summarising and routine administration. Keep a human in the loop for anything that affects customers or money. The tools are cheap enough to experiment with, and the cost of waiting is mostly opportunity cost.

This article was written by Triweb AI's editorial team based on analysis of today's leading AI news sources.