The enterprise AI landscape is moving quickly.
Microsoft recently introduced its evolving Copilot experience around Home, Code and Autopilot, with Chat and Copilot Cowork coming together within Home. OpenAI's DevDay introduced Dots, alongside further developments across ChatGPT and Codex. Anthropic continues to evolve Claude, Claude Code and Managed Agents, and is now bringing its Chat and Cowork experiences together into a unified Claude experience.
These aren't the only platforms and models enterprises are evaluating. Google Gemini is another significant part of the enterprise AI landscape, while Microsoft is expanding its own MAI model family alongside its partnerships with other model providers. Meta continues to develop its AI ecosystem, and xAI's Grok is expanding its business and enterprise presence. Platforms such as AWS Bedrock and IBM watsonx also play important roles in building, deploying and governing enterprise AI.
For this article, I'm focusing specifically on Microsoft Copilot, OpenAI and Anthropic, and how their increasingly overlapping capabilities could shape the multi-AI workplace.
Looking across these developments, one thing stands out to me:
AI is rapidly moving beyond chat.
We started by asking AI questions. Then we started using it to create content and complete tasks. Now we are moving toward AI that can take on larger pieces of work — and increasingly, continue working toward an objective.
For organizations investing across multiple AI platforms, this creates both opportunity and complexity.
Three Platforms, Increasingly Overlapping Capabilities
Microsoft Copilot, OpenAI and Anthropic have different strengths and starting points, but their capabilities are increasingly overlapping.
Here is a simplified view:
| Enterprise capability | Microsoft | OpenAI | Anthropic |
|---|---|---|---|
| AI assistant | Microsoft 365 Copilot | ChatGPT | Claude |
| Delegated knowledge work | Copilot Cowork | ChatGPT work/agent capabilities | Claude / Cowork capabilities |
| Persistent / asynchronous work | Autopilot (private preview) | Dots (rolling out) | Managed Agents (beta) |
| Developer AI | GitHub Copilot | Codex | Claude Code |
| Building/customizing agents | Copilot Studio | OpenAI agent development tools | Claude Agent SDK |
| Enterprise context & connectivity | Microsoft 365 context, Work IQ and connectors | Apps, plugins and connected enterprise data | MCP, connectors and integrations |
| Enterprise strength | Deep M365, identity and governance integration | General-purpose AI, cross-platform work and agents | Knowledge work, coding and flexible agent development |
These are not exact product equivalents, and they shouldn't be treated as such.
Microsoft's Autopilot, previously known as Scout, is being positioned as a persistent, proactive digital teammate with its own identity, memory, computer and workspace. That is particularly interesting from an enterprise perspective because persistent agents introduce new questions around identity, permissions, governance and accountability.
OpenAI's Dots similarly moves beyond a traditional chat session. A Dot can be given a goal, connected to the applications it needs and continue making progress between conversations using its own cloud computer.
Anthropic's Managed Agents, currently in beta, provides managed infrastructure for long-running and asynchronous agent workloads with persistent sessions.
There is also an interesting relationship between the two Cowork experiences.
Microsoft's Copilot Cowork and Anthropic's Claude Cowork share more than a name. Microsoft says it worked closely with Anthropic to bring the technology powering Claude Cowork into Microsoft 365 Copilot. They remain separate products, delivered and governed through their respective platforms.
Anthropic is now also bringing its Chat and Cowork experiences together into one Claude experience, beginning with Pro and Max users.
Microsoft's Code is another capability worth watching separately. It is aimed at knowledge workers rather than traditional developers, enabling people to create apps, dashboards, widgets and automations using natural language. That makes it different from GitHub Copilot and is why I haven't grouped the two together as developer AI.
Multi-AI May Become Normal
Many organizations are already invested in Microsoft 365 and Copilot while also evaluating ChatGPT Enterprise, Claude and other AI platforms for different business requirements.
That doesn't necessarily mean choosing one and removing the others.
But the growing overlap does make capability and licence optimization increasingly important.
Copilot itself can incorporate technology and models from other AI providers. Copilot Cowork's relationship with Anthropic is a good example. An organization separately licensing multiple AI platforms should therefore look beyond vendor names and understand which capabilities, models and business outcomes it is actually paying for.
That doesn't automatically make overlapping platforms unnecessary. Different products can provide very different enterprise context, user experiences, governance, security and integration even when some underlying AI technology overlaps.
The question is whether that overlap creates additional business value.
Use the Governance Structure You Already Have
This connects with something I wrote about earlier on building an Agentic CoE.
Every new AI platform doesn't require another Center of Excellence.
If your organization already has a dedicated AI or Agentic CoE managing AI workloads, align these evaluations with that team.
If AI governance already sits across Digital Workplace, Architecture, Cloud, Data, Security, Automation or Innovation teams, build on that existing model.
And if that capability doesn't exist today, establishing an appropriate AI governance and evaluation structure becomes increasingly important.
The name of the team matters less than having the right people around the table.
Bring the Capabilities to the Table First
Before taking every new capability to thousands of users, bring it to the table first.
Let the AI CoE or existing governance structure work with platform teams, security, architecture and the champions community. Give selected capabilities to representative business users and test them against real business use cases.
Understand:
- What business problem are we solving?
- Which platform is best positioned for that workload?
- Where do capabilities overlap?
- What enterprise data does it require?
- What are the identity, security and compliance implications?
- What does it cost?
- What measurable value does it produce?
- Who actually needs access?
Instead of asking:
"Which AI platform is better?"
perhaps the better question is:
"Which AI capability is right for this business use case?"
Some employees may need Copilot. Others may benefit more from ChatGPT or Claude. Developers may require a different combination, while power users and AI champions may genuinely benefit from access to multiple platforms.
Not everyone needs everything.
Governance May Also Cross Platform Boundaries
Another interesting development is that these ecosystems don't necessarily have to remain completely isolated.
OpenAI has indicated that specialist Dots are planned to integrate with Microsoft Agent 365, bringing another dimension to cross-platform agent governance.
It is still early and worth watching rather than drawing conclusions from today.
But as enterprises adopt agents from multiple vendors, the ability to discover, govern, secure and monitor them through common enterprise controls could become increasingly important.
A multi-AI workplace may eventually involve multiple AI providers while maintaining common governance around them.
ROI Needs to Follow the Use Case
Adoption alone isn't ROI.
Knowing that employees activated an AI tool tells us something about adoption, but not necessarily whether the investment is delivering business value.
I would look at something closer to:
Use Case → Platform → Cost → Adoption → Time Saved → Quality → Business Outcome
If AI materially reduces effort while maintaining or improving quality, there is a value story.
If the same employee has three AI licences but consistently uses only one, there may be an optimization opportunity.
And as persistent agents consume resources while performing asynchronous work, AI FinOps and observability become increasingly important.
The objective shouldn't be to give every AI product to every employee.
It should be to put the right capability in the hands of the people and processes where it creates measurable value.
A Note on Availability
Many of these capabilities are still evolving.
Microsoft Copilot Cowork is already generally available, while the newer Home, Code and Autopilot experiences are still progressing through their respective rollout stages. Microsoft says Home will begin reaching its Frontier program in the coming weeks, with Code following at the end of the month and wider access afterward. Autopilot is currently in private preview.
OpenAI Dots is rolling out gradually to eligible users, while Anthropic Managed Agents remains in beta. Anthropic is also progressively bringing Chat and Cowork together, starting with Pro and Max plans. For Enterprise customers, Anthropic says admins will hear from it at least 30 days before anything changes for their organization.
Features, licensing and availability will continue to change, so organizations should validate current vendor guidance before making platform or rollout decisions.
What This Means for Enterprises
Microsoft Copilot, ChatGPT and Claude are all evolving beyond conversational AI toward delegated work, agents, coding, application integration and increasingly persistent execution.
For enterprises, that doesn't mean deploying everything.
If you already have an AI CoE, use it. If you have an established governance model, build on it. Bring your champions and representative business users into the evaluation, test real use cases and understand the overlap before scaling.
The technology will keep changing, and today's differentiator may become tomorrow's standard feature.
So perhaps the question is no longer:
"Which AI should we buy?"
but:
"Which AI capability creates the most value for this work?"
Three Practical Takeaways
If I had to reduce this to three things for enterprise teams:
Map capabilities, not just vendors. Understand where the platforms overlap and where each genuinely adds value.
Use the governance you already have. Extend your existing AI CoE or governance model rather than creating another structure for every new platform.
Pilot before you scale. Test real business use cases with representative users, measure the value and understand the governance implications before expanding access.
From AI Cost to AI Capacity
Microsoft's Jared Spataro recently added another interesting dimension to the FinOps for AI discussion: as agentic AI becomes increasingly consumption-based, the question isn't only how much AI costs, but which business priorities should receive that capacity in the first place.
This shifts the conversation from simply managing AI consumption toward allocating AI capacity based on business value.
Central teams can provide governance, visibility and common controls, while business leaders closer to the work can help determine where that capacity should be invested. The important part is connecting the investment to an expected outcome, measuring whether it delivered, and then scaling what creates value and retiring what doesn't.
For me, this reinforces a simple principle:
AI FinOps shouldn't stop at measuring consumption. It needs to connect consumption to business outcomes.