Agentic AI for business: Microsoft, Google and Anthropic just changed the deal
TL;DR
Microsoft, Google, and Anthropic have each embedded autonomous AI agents inside the software your business already pays for. DeepSeek V3.2 has cut the cost of running these agents by 90–95%, eliminating every 'too expensive' argument. Meanwhile, 67% of Australian workers are already using unapproved AI tools, creating serious data and governance risk. If you do not have an AI agent strategy in place right now, you are already behind.
What just happened in the last 48 hours?
Three of the largest technology companies in the world made simultaneous moves that changed the nature of work. Not a product update cycle. A category change.
Microsoft rolled out Copilot Cowork, an autonomous agent embedded inside Microsoft 365. Not a sidebar. Not a chatbot. An agent with a job. It scans your emails and calendar, identifies critical meetings, pulls client history from your shared files, prepares detailed briefing documents, drafts follow-up emails, and creates marketing collateral, proactively, before you ask. If your business runs on Microsoft 365, this is already sitting inside your tools.
Google embedded Gemini so deeply into Workspace it is practically part of the furniture. Give it one high-level instruction, "Create a comprehensive business proposal for Project X targeting the APAC region", and it executes across apps: a detailed proposal in Docs, financial projections with sourced market data in Sheets, and a compelling presentation in Slides. One prompt. An entire project delivered.
Anthropic updated Claude to operate as a persistent agent, subtle, but arguably the most significant shift of the three. Claude now retains context across sessions. Assign it an ongoing task, such as monitoring a competitor's marketing activities, and it keeps working. It remembers. It refines its analysis. It delivers weekly summaries without you prompting it each time. That is not a tool. That is a team member.
What does 'agentic AI' actually mean for your business?
The paradigm has shifted from asking AI a question to giving AI a job. That is not a minor evolution, it is a complete rethinking of what a workforce looks like.
For years the dynamic was unchanged: a human operator using a digital tool. Email replaced the fax. Cloud storage replaced the filing cabinet. Every step improved efficiency, but humans were still directing every action. Agentic AI removes the human from a significant portion of that equation. The tools are starting to run themselves.
Lead generation, customer onboarding, financial reporting, supply chain management, the answer to "Can an AI agent do this?" is increasingly yes.
The real question is no longer whether you have access to AI. It is whether you have a strategy to direct it.
How cheap is this getting, and why does that change everything?
The assumption that agentic AI is the domain of large enterprises with nine-figure IT budgets is now wrong.
DeepSeek V3.2, a model out of China, delivers performance on par with GPT-4, the model powering many of these top-tier agent systems, at a 90–95% reduction in cost. The raw compute power required to run sophisticated AI agents is becoming a commodity accessible to anyone with a credit card.
This creates two simultaneous realities:
- Opportunity: An agile SME can now build a team of digital agents handling marketing, sales, operations, and finance for less than the cost of a single human employee.
- Threat: Your competitors have access to the exact same power. The gap between what is technologically possible and what most SMEs are actually doing is no longer a gap, it is a chasm, and it is widening by the hour.
The excuse that this technology is too expensive or too complex has evaporated. The real differentiator is no longer access to the technology. It is the strategy for how you implement it.
See where AI fits in your business. Free.
A 45-minute audit. We map the highest-value automations and what they're worth in time and money. No pitch, no pressure.
What happens when 67% of your team goes rogue with AI?
Here is where it gets uncomfortable. The tools are cheap and accessible, so your team is already using them, just not in a way you have approved.
Recent studies show 67% of Australian workers are using AI tools that have not been approved by their employers. This is the shadow AI problem, and it is a significant operational and legal risk.
They are feeding sensitive company data, financial reports, customer lists, strategic plans, into free public tools without a second thought. No governance. No security review. No strategy. The consequences include:
- Data breaches and intellectual property leakage
- Compliance failures and regulatory exposure
- Fragmented workflows with no integration across teams
- Wasted effort as twelve teams use twelve tools that do not connect
It creates a veneer of innovation hiding a foundation of risk. Power without control is just chaos.
What does serious AI implementation actually look like?
National Australia Bank (NAB) is the benchmark worth studying. They built what they call a 'customer brain', a massive integrated AI system comprising over 3,500 models all working together. This is not a single tool bolted onto existing processes. It is a central nervous system for the entire organisation.
The result: NAB can personalise services, anticipate customer needs, and make smarter decisions in real time across millions of customer touchpoints. They did not buy a tool and hope for the best. They invested in architecture, established governance from day one, and built a strategic capability, not a collection of unsupervised experiments.
Contrast that with 67% of a workforce experimenting unsupervised with unapproved tools and hoping nothing leaks.
Simply giving everyone access to an AI agent without a strategy is not innovation. It is risk with a productivity label on it.
What this means for your business right now
If your business runs Microsoft 365 or Google Workspace, you already have an autonomous AI agent living inside your systems. It is there right now. The question is not whether you have access to it. The question is whether you have any idea how to use it properly, whether your team does, and whether you have a governance framework to ensure it is deployed safely.
Using AI to summarise emails and calling it innovation is not a strategy. It is a hobby.
The businesses that will win in this environment are the ones that do three things deliberately:
- Audit their processes, every recurring task, every manual report, every workflow that has not changed in years. Ask: can an AI agent handle this?
- Build governance first, approved tools, data-handling rules, access controls, and training before widespread deployment.
- Redesign, not bolt on, use these agents to fundamentally rethink how the business operates, not just to accelerate old processes.
You can either be the one directing the agents, or you can be the one the agents have made redundant. That is the accurate framing of the choice sitting in front of every business owner right now.
What to do this week
- Audit your current AI exposure. Ask your team what AI tools they are already using. Do not assume. The 67% statistic is not someone else's problem.
- Open the AI features in your existing stack. If you pay for Microsoft 365 or Google Workspace, explore what Copilot Cowork or Gemini can actually do in practice, not what the marketing material claims.
- Identify three high-frequency, low-risk processes where an AI agent could take over a repeatable task. Start there before scaling.
- Draft a one-page shadow AI policy. Define which tools are approved, what data cannot be entered into external AI systems, and who to consult before experimenting with new tools.
- Commit to a strategy, not a subscription. Buying access to an AI agent is the easy part. Building the architecture, governance, and workflows to use it effectively is the work, and it is the only part that creates a durable competitive advantage.
Where to from here
Book a free 60-minute AI audit, we'll explore exactly what workflows are worth augmenting with AI.
Live with passion & AI,
Brett
Need an AI operator inside your team?
Place a Chief AI Officer, an AI Officer, or embed an Anaboo Forward Deployed Engineer for 3–6 months.
Frequently asked questions
What is Microsoft Copilot Cowork and how is it different from earlier Copilot versions?
+
Copilot Cowork is an autonomous agent embedded inside Microsoft 365 that works proactively rather than waiting for prompts. It scans your emails, calendar, and files, prepares briefing documents, drafts follow-up emails, and creates collateral, all before you ask.
How has Google Gemini changed what Google Workspace can do?
+
Gemini in Google Workspace can now take a single high-level instruction and execute a full multi-step project across apps, generating a detailed proposal in Docs, financial projections with sourced market data in Sheets, and a presentation in Slides, from one prompt.
What does it mean for Anthropic's Claude to be a persistent agent?
+
A persistent agent retains context across sessions. You assign Claude an ongoing task, monitoring a competitor's marketing activities, for example, and it keeps working, refining its analysis, and delivering weekly summaries without needing a fresh prompt each time.
What is DeepSeek V3.2 and why does it matter for small businesses?
+
DeepSeek V3.2 is a model delivering performance comparable to GPT-4 at a 90–95% reduction in cost. This collapse in price means SMEs can now run sophisticated AI agents for less than the cost of a single human employee, removing the 'too expensive' excuse entirely.
What is shadow AI and what risk does it create for Australian businesses?
+
Shadow AI refers to employees using AI tools that have not been approved by their employer. Studies show 67% of Australian workers are already doing this, feeding sensitive data, customer lists, financial reports, strategic plans, into public tools with no governance or security controls in place.
How did NAB implement AI at scale, and what can other businesses learn from it?
+
NAB built a 'customer brain' comprising over 3,500 integrated AI models working together as a central nervous system for the organisation. They invested in architecture and governance from day one rather than deploying isolated tools, enabling real-time personalisation across millions of customer interactions.
Where should a business start when building an AI agent strategy?
+
Start by auditing every recurring process and asking whether an AI agent could handle it. Then establish governance, approved tools, data-handling rules, and training, before widespread deployment. The competitive advantage is no longer access to the technology; it is the strategy for how you implement it.
Brett is a four-time founder (Darra Tyres, Gladfish, EzyTrac, Anaboo) and the operator behind AIOS, Anaboo's AI Operating System. He writes from inside the build, installing AI in his own businesses first and reporting back what actually moves the numbers. Based between Singapore, the UK and Australia.
