Here’s a scenario I’m hearing from more and more business leaders right now. They’ve rolled out Microsoft 365 Copilot, a handful of people on the team are using ChatGPT, and someone in the marketing or legal team has started working with Claude.
The tools are in the building, people are getting value, and the next question is becoming clear: how do we turn this into something that actually changes the way our business works?
That’s the shift from AI tools to an AI operating model. Microsoft has a name for the businesses making that shift. They call them Frontier Firms.
The difference isn’t the tools they use. It’s how deliberately they’ve designed the way work flows between their people and their AI agents.
What Is a Frontier Firm, and Why Does It Matter for Your Business?
Microsoft’s 2026 Work Trend Index, published in May 2026, introduces the idea of the Frontier Firm as an organisation that doesn’t just use AI, it redesigns work around it. Human involvement doesn’t disappear. It changes shape. People move away from step-by-step execution and toward direction, quality control, and judgement.
The research is clear about what’s already happening.
A privacy-preserving analysis of more than 100,000 Microsoft 365 Copilot chats found that 49% of all conversations support cognitive work: analysing information, solving problems, and thinking through decisions. A further 58% of AI users now say they’re producing work they couldn’t have created a year ago.
What’s holding most businesses back isn’t access to AI. According to Microsoft’s Work Trend Index data, organisational factors like culture, manager support, and talent practices account for more than twice the AI impact of individual behaviour.
The real question isn’t whether your people have the right tools. It’s whether your business is designed to get value from them.
That’s the Frontier Firm gap. The businesses closing it are the ones thinking about AI as an operating model challenge, not just a software rollout.
The Four Patterns of Human and AI Collaboration
Before you can design an AI operating model, it helps to have a clear picture of how humans and agents actually work together. Microsoft has identified four patterns, and they’re showing up across every function in the businesses we work with.
Author
You’re doing the work, with AI helping as needed. Drafting an email, generating a chart, checking a calculation. AI is a tool you pick up and put down.
Editor
You set the intent and AI produces the first draft. You review, shape, and approve the output. You’re still in the loop, but the starting point is AI-generated.
Director
You hand off a full task and AI executes it in the background. You define what needs to happen and review the result, but you’re not involved in each step.
Orchestrator
You design a system where multiple agents run in parallel across a workflow, flagging exceptions and escalating decisions back to you when needed.
Most businesses are running a mix of all four right now, often without realising it. The Frontier Firm difference is being deliberate about which pattern belongs where, and who’s accountable for outcomes at each stage.
Microsoft 365 as Your AI Backbone
For most Australian businesses, Microsoft 365 is already where work happens. Email, documents, data, meetings, approvals. It’s all there. That makes it the natural backbone for an AI operating model, and Microsoft has been building fast to make that real.
Microsoft 365 Copilot now handles multi-step tasks across Word, Excel, Outlook and Teams. Copilot Cowork, which is now available on iOS and Android, extends this into a cross-device experience where staff can delegate outcomes and pick them up later across devices.
Native plugins connect Copilot to Dynamics 365 and Microsoft Fabric, and federated connectors bring in real-time data from partners like HubSpot, Moody’s, Notion and LSEG.
The other piece worth understanding is Microsoft Agent 365, which launched as part of Microsoft 365 E7 in March 2026. Agent 365 is the control plane for managing AI agents across your business. It gives IT a unified view of what agents are running, what they can access, and how to govern them.
For businesses that have read our blog on the Agentic System of Work, this is the infrastructure that makes coordinated, human-led, agent-operated work possible at scale.
Copilot handles the flow of everyday work, Cowork takes on delegated multi-step tasks, and Agent 365 keeps it all governed. Together, they give Microsoft 365 a strong claim as the operating backbone for most Australian businesses, even when they’re also using OpenAI or Anthropic.
Where OpenAI and Anthropic Fit In
Microsoft 365 is where most of the work happens. But OpenAI and Anthropic are doing serious work on enterprise agent capabilities, and for certain parts of your operating model they’re worth understanding on their own terms.
OpenAI Workspace Agents
In April 2026, OpenAI launched Workspace Agents as the enterprise successor to custom GPTs. These are persistent agents that live inside ChatGPT and connect directly to tools like Slack, Microsoft 365, Salesforce, Google Drive, and Notion. They can be built by teams to handle specific workflows and shared across the organisation with admin controls, a Compliance API, and usage analytics.
Where Workspace Agents shine is in cross-tool research and coordination work. These are scenarios where information lives across multiple platforms and an agent needs to pull it together, summarise it, and hand it to a person in a usable form. For sales teams, communications teams, or operations managers working across many systems, that’s a real use case.
Anthropic Claude Managed Agents
Anthropic released Claude Managed Agents in April 2026 as a suite of composable APIs for building and deploying cloud-hosted agents at scale. Each agent runs in a sandboxed environment with persistent memory, scoped permissions, and end-to-end tracing built in. You define the task, the tools, and the success criteria. Claude handles execution and state management.
Claude Managed Agents are particularly strong for technical and analytical workflows, including document processing, code review, root cause analysis, and financial and legal document handling. Microsoft Agent 365 covers governance across your Microsoft environment. Claude Managed Agents fill a different role by providing a managed, auditable runtime for specialised workflows that need tight control and traceability.
What’s worth noting right now is that these providers are not entirely separate. Claude Opus 4.7 is now available inside Microsoft 365 Copilot Chat, and GPT-5.5 Instant has rolled out to Copilot Studio. The lines between platforms are blurring, which is exactly why having a clear operating model matters more than picking a single provider.
Designing Your Cross-Vendor Operating Model
This is where it gets practical. A Frontier Firm operating model isn’t about choosing one platform and ignoring the others. It’s about being clear on which work belongs where, and putting the right human and governance controls in place across all of it.
Here are four questions I walk through with clients when they’re designing this.
1. Where does human judgement stay firmly in the loop?
Start by mapping the decisions that can’t be delegated. Legal risk, client relationships, strategic priorities. Anything where the consequences of a wrong call are significant stays with people. AI can inform those decisions, but the orchestrator pattern isn’t appropriate for all of them.
2. Which workflows belong in Microsoft 365, and which in specialist agents?
Microsoft 365 Copilot and Copilot Studio are the right home for workflows that live inside your Microsoft environment: document drafting, financial approvals, Teams meeting actions, Dynamics 365 data summarisation, Power Platform automations. OpenAI Workspace Agents are well suited to cross-platform coordination work where your people already use multiple SaaS tools. Claude Managed Agents are worth considering for high-complexity, auditable processes like compliance checks, technical analysis, or document-heavy workflows that need precise tracing.
3. How will you govern agents across platforms?
This is the step most businesses skip and then regret. In Microsoft 365, use Agent 365 plus Microsoft Purview and Defender to manage agent identities, data access, sensitivity labels and compliance. In your OpenAI environment, use the admin console to define which connectors agents can use, who can build and share agents, and how activity is logged. In Anthropic, use the Claude Console to set spend limits, monitor sessions, and connect to your compliance dashboards.
Tie agent use to outcomes that matter to your business. Think time saved per process, error rates, volume handled, and escalation frequency. Microsoft’s research is clear that the return on AI investment comes from how work is redesigned, not just from access to the tools. Build a scorecard using Agent 365 reporting, ChatGPT Enterprise analytics, and Claude’s session data so you have a joined-up view of what’s working.
What a Practical First Step Looks Like
You don’t need to have the whole model designed before you start. What I recommend is picking two or three workflows that matter, mapping where they sit against the four collaboration patterns, then deciding which platform is the right home for each one.
Common starting points we see with Australian businesses include customer communications and triage in Microsoft 365 Copilot or Copilot Studio, cross-system research and synthesis using OpenAI Workspace Agents, and document review or compliance checking using Claude Managed Agents or Copilot.
The real risk isn’t moving too slowly. It’s letting agents accumulate without a design behind them, which is how you end up with shadow agents, ungoverned data access, and staff unsure about who’s responsible for AI-generated outputs.
The businesses building operating models now, not just running pilots, are the ones that will be hardest to catch in 2027.
How CG TECH Can Help
At CG TECH, we help Australian businesses build the foundation for exactly this kind of operating model. That means getting your Microsoft 365 environment, data governance, and agent infrastructure into the right shape, and then helping you make deliberate decisions about where OpenAI and Anthropic capabilities fit alongside that.
If you’d like to talk through where your business sits right now and what a practical path to the Frontier Firm model looks like for you, reach out to our team. We’d love to have that conversation.
About the Author
Carlos Garcia is the Founder and Managing Director of CG TECH, where he leads enterprise digital transformation projects across Australia.
With deep experience in business process automation, Microsoft 365, and AI-powered workplace solutions, Carlos has helped businesses in government, healthcare, and enterprise sectors streamline workflows and improve efficiency.
He holds Microsoft certifications in Power Platform and Azure and regularly shares practical guidance on Copilot readiness, data strategy, and AI adoption.
Here’s a scenario I’m hearing from more and more business leaders right now. They’ve rolled out Microsoft 365 Copilot, a handful of people on the team are using ChatGPT, and someone in the marketing or legal team has started working with Claude.
The tools are in the building, people are getting value, and the next question is becoming clear: how do we turn this into something that actually changes the way our business works?
That’s the shift from AI tools to an AI operating model. Microsoft has a name for the businesses making that shift. They call them Frontier Firms.
The difference isn’t the tools they use. It’s how deliberately they’ve designed the way work flows between their people and their AI agents.
What Is a Frontier Firm, and Why Does It Matter for Your Business?
Microsoft’s 2026 Work Trend Index, published in May 2026, introduces the idea of the Frontier Firm as an organisation that doesn’t just use AI, it redesigns work around it. Human involvement doesn’t disappear. It changes shape. People move away from step-by-step execution and toward direction, quality control, and judgement.
The research is clear about what’s already happening.
A privacy-preserving analysis of more than 100,000 Microsoft 365 Copilot chats found that 49% of all conversations support cognitive work: analysing information, solving problems, and thinking through decisions. A further 58% of AI users now say they’re producing work they couldn’t have created a year ago.
What’s holding most businesses back isn’t access to AI. According to Microsoft’s Work Trend Index data, organisational factors like culture, manager support, and talent practices account for more than twice the AI impact of individual behaviour.
The real question isn’t whether your people have the right tools. It’s whether your business is designed to get value from them.
That’s the Frontier Firm gap. The businesses closing it are the ones thinking about AI as an operating model challenge, not just a software rollout.
The Four Patterns of Human and AI Collaboration
Before you can design an AI operating model, it helps to have a clear picture of how humans and agents actually work together. Microsoft has identified four patterns, and they’re showing up across every function in the businesses we work with.
Author
You’re doing the work, with AI helping as needed. Drafting an email, generating a chart, checking a calculation. AI is a tool you pick up and put down.
Editor
You set the intent and AI produces the first draft. You review, shape, and approve the output. You’re still in the loop, but the starting point is AI-generated.
Director
You hand off a full task and AI executes it in the background. You define what needs to happen and review the result, but you’re not involved in each step.
Orchestrator
You design a system where multiple agents run in parallel across a workflow, flagging exceptions and escalating decisions back to you when needed.
Most businesses are running a mix of all four right now, often without realising it. The Frontier Firm difference is being deliberate about which pattern belongs where, and who’s accountable for outcomes at each stage.
Microsoft 365 as Your AI Backbone
For most Australian businesses, Microsoft 365 is already where work happens. Email, documents, data, meetings, approvals. It’s all there. That makes it the natural backbone for an AI operating model, and Microsoft has been building fast to make that real.
Microsoft 365 Copilot now handles multi-step tasks across Word, Excel, Outlook and Teams. Copilot Cowork, which is now available on iOS and Android, extends this into a cross-device experience where staff can delegate outcomes and pick them up later across devices.
Native plugins connect Copilot to Dynamics 365 and Microsoft Fabric, and federated connectors bring in real-time data from partners like HubSpot, Moody’s, Notion and LSEG.
The other piece worth understanding is Microsoft Agent 365, which launched as part of Microsoft 365 E7 in March 2026. Agent 365 is the control plane for managing AI agents across your business. It gives IT a unified view of what agents are running, what they can access, and how to govern them.
For businesses that have read our blog on the Agentic System of Work, this is the infrastructure that makes coordinated, human-led, agent-operated work possible at scale.
Copilot handles the flow of everyday work, Cowork takes on delegated multi-step tasks, and Agent 365 keeps it all governed. Together, they give Microsoft 365 a strong claim as the operating backbone for most Australian businesses, even when they’re also using OpenAI or Anthropic.
Where OpenAI and Anthropic Fit In
Microsoft 365 is where most of the work happens. But OpenAI and Anthropic are doing serious work on enterprise agent capabilities, and for certain parts of your operating model they’re worth understanding on their own terms.
OpenAI Workspace Agents
In April 2026, OpenAI launched Workspace Agents as the enterprise successor to custom GPTs. These are persistent agents that live inside ChatGPT and connect directly to tools like Slack, Microsoft 365, Salesforce, Google Drive, and Notion. They can be built by teams to handle specific workflows and shared across the organisation with admin controls, a Compliance API, and usage analytics.
Where Workspace Agents shine is in cross-tool research and coordination work. These are scenarios where information lives across multiple platforms and an agent needs to pull it together, summarise it, and hand it to a person in a usable form. For sales teams, communications teams, or operations managers working across many systems, that’s a real use case.
Anthropic Claude Managed Agents
Anthropic released Claude Managed Agents in April 2026 as a suite of composable APIs for building and deploying cloud-hosted agents at scale. Each agent runs in a sandboxed environment with persistent memory, scoped permissions, and end-to-end tracing built in. You define the task, the tools, and the success criteria. Claude handles execution and state management.
Claude Managed Agents are particularly strong for technical and analytical workflows, including document processing, code review, root cause analysis, and financial and legal document handling. Microsoft Agent 365 covers governance across your Microsoft environment. Claude Managed Agents fill a different role by providing a managed, auditable runtime for specialised workflows that need tight control and traceability.
What’s worth noting right now is that these providers are not entirely separate. Claude Opus 4.7 is now available inside Microsoft 365 Copilot Chat, and GPT-5.5 Instant has rolled out to Copilot Studio. The lines between platforms are blurring, which is exactly why having a clear operating model matters more than picking a single provider.
Designing Your Cross-Vendor Operating Model
This is where it gets practical. A Frontier Firm operating model isn’t about choosing one platform and ignoring the others. It’s about being clear on which work belongs where, and putting the right human and governance controls in place across all of it.
Here are four questions I walk through with clients when they’re designing this.
1. Where does human judgement stay firmly in the loop?
Start by mapping the decisions that can’t be delegated. Legal risk, client relationships, strategic priorities. Anything where the consequences of a wrong call are significant stays with people. AI can inform those decisions, but the orchestrator pattern isn’t appropriate for all of them.
2. Which workflows belong in Microsoft 365, and which in specialist agents?
Microsoft 365 Copilot and Copilot Studio are the right home for workflows that live inside your Microsoft environment: document drafting, financial approvals, Teams meeting actions, Dynamics 365 data summarisation, Power Platform automations. OpenAI Workspace Agents are well suited to cross-platform coordination work where your people already use multiple SaaS tools. Claude Managed Agents are worth considering for high-complexity, auditable processes like compliance checks, technical analysis, or document-heavy workflows that need precise tracing.
3. How will you govern agents across platforms?
This is the step most businesses skip and then regret. In Microsoft 365, use Agent 365 plus Microsoft Purview and Defender to manage agent identities, data access, sensitivity labels and compliance. In your OpenAI environment, use the admin console to define which connectors agents can use, who can build and share agents, and how activity is logged. In Anthropic, use the Claude Console to set spend limits, monitor sessions, and connect to your compliance dashboards.
If you’d like to go deeper on the governance side, our blog on securing AI agents with Entra Agent ID and Purview covers the Microsoft piece in detail.
4. How will you measure value and risk?
Tie agent use to outcomes that matter to your business. Think time saved per process, error rates, volume handled, and escalation frequency. Microsoft’s research is clear that the return on AI investment comes from how work is redesigned, not just from access to the tools. Build a scorecard using Agent 365 reporting, ChatGPT Enterprise analytics, and Claude’s session data so you have a joined-up view of what’s working.
What a Practical First Step Looks Like
You don’t need to have the whole model designed before you start. What I recommend is picking two or three workflows that matter, mapping where they sit against the four collaboration patterns, then deciding which platform is the right home for each one.
Common starting points we see with Australian businesses include customer communications and triage in Microsoft 365 Copilot or Copilot Studio, cross-system research and synthesis using OpenAI Workspace Agents, and document review or compliance checking using Claude Managed Agents or Copilot.
The real risk isn’t moving too slowly. It’s letting agents accumulate without a design behind them, which is how you end up with shadow agents, ungoverned data access, and staff unsure about who’s responsible for AI-generated outputs.
The businesses building operating models now, not just running pilots, are the ones that will be hardest to catch in 2027.
How CG TECH Can Help
At CG TECH, we help Australian businesses build the foundation for exactly this kind of operating model. That means getting your Microsoft 365 environment, data governance, and agent infrastructure into the right shape, and then helping you make deliberate decisions about where OpenAI and Anthropic capabilities fit alongside that.
If you’d like to talk through where your business sits right now and what a practical path to the Frontier Firm model looks like for you, reach out to our team. We’d love to have that conversation.
About the Author
Carlos Garcia is the Founder and Managing Director of CG TECH, where he leads enterprise digital transformation projects across Australia.
With deep experience in business process automation, Microsoft 365, and AI-powered workplace solutions, Carlos has helped businesses in government, healthcare, and enterprise sectors streamline workflows and improve efficiency.
He holds Microsoft certifications in Power Platform and Azure and regularly shares practical guidance on Copilot readiness, data strategy, and AI adoption.
Sources
Recent Posts
Popular Categories
Archives