What is the difference between Claude Projects, Custom GPTs and Copilot Agents?

All three are 'scoped AI assistants' configured once and reused. Claude Projects are workspaces with curated documents and instructions inside Claude.ai Teams. Custom GPTs are configurable assistants inside ChatGPT with optional API-calling actions. Copilot Agents are configurable assistants inside Microsoft 365 Copilot that live inside Word, Excel, Outlook and Teams. The choice is mostly about which underlying platform fits your workflows best.

Which is easiest to maintain at SMB scale?

Claude Projects, by some distance. The Project model maps cleanly to one-per-business-function ownership, the curated-context approach is small enough to maintain (typically 10-20 documents per Project), and there is a clear named owner per Project. Custom GPTs proliferate easily because they are quick to spin up; teams routinely end up with 30+ GPTs they cannot keep tidy. Copilot Agents fall between the two.

Which has the most powerful action capability?

Custom GPTs, today. The Actions feature in Custom GPTs is the most mature interface for calling external APIs from inside a scoped assistant. Claude Projects support similar action capability through MCP-shaped tools and the Anthropic API, and the gap is closing fast. For most SMB workflows the difference is academic; for workflows that depend on calling many external services from inside the assistant, Custom GPTs still lead.

Are Copilot Agents only worth it if we are deep in Microsoft 365?

Roughly yes. Copilot Agents shine because they live inside Word, Excel, Outlook and Teams. If your team works in Microsoft 365 all day, the in-app experience compounds value. If you spend most of your day in a browser, Slack and a non-Microsoft stack, the Microsoft-native value mostly disappears and Claude Projects or Custom GPTs are the better fit.

Can we use more than one of these?

Yes. Most XLev clients run Claude Projects as the default for company-wide scoped assistants, with a small number of Custom GPTs or Copilot Agents for workflows that genuinely fit those platforms better (multimodal-heavy work in Custom GPTs, in-document M365 work in Copilot Agents). The governance question is which platform owns which workflow and avoiding duplicated assistants in three places.

← All insights

Comparison·8 min read

Claude.ai Projects vs Custom GPTs vs Copilot Agents: which fits an SMB team

The three scoped-AI-assistant frameworks compete for the same job: take an AI model, configure it once for a specific workflow, reuse it across the team. They sound similar in marketing and behave quite differently in practice. Here is the honest read on Projects versus Custom GPTs versus Copilot Agents.

The job to be done

All three platforms answer the same question: how do you take a general-purpose AI model and tune it to behave a specific way for a specific business function, so that anyone on the team can invoke it without explaining the context each time. The implementation details differ; the underlying job is the same.

Claude.ai Projects

A Claude Project is a scoped Claude.ai workspace with curated context: documents (SOPs, frameworks, examples), instructions (how to behave in this scope), and shared chat history. Live inside Claude.ai Teams and Enterprise.

Strengths: cleanest one-per-business-function model; small, curated context (10-20 docs) is easy to maintain; clear named ownership; strongest writing quality
Weaknesses: action capability (calling external APIs) is slightly behind Custom GPTs; smaller third-party ecosystem
Best for: the SMB default - one Project per business function

OpenAI Custom GPTs

A Custom GPT is a configurable assistant inside ChatGPT with custom instructions, optional Knowledge files, and Actions that can call external APIs. Custom GPTs can be shared inside a Team workspace or published publicly.

Strengths: mature Actions feature for calling external APIs; large third-party marketplace; multimodal native; fast to spin up
Weaknesses: easy to proliferate (teams routinely end up with 30+ unmaintained GPTs); ownership is fuzzy; context limits looser than Projects
Best for: workflows that depend on calling external APIs from inside the assistant, or where a relevant Custom GPT already exists in the marketplace

Microsoft 365 Copilot Agents

A Copilot Agent is a configurable Copilot assistant that lives inside Microsoft 365 - Word, Excel, Outlook, Teams, SharePoint. Configured via Copilot Studio with custom instructions, connectors and knowledge sources.

Strengths: in-app experience inside the Microsoft 365 suite; tight SharePoint and Teams integration; enterprise governance via Microsoft 365 admin tools
Weaknesses: value evaporates outside the Microsoft 365 ecosystem; configuration tooling (Copilot Studio) is more involved than Projects or Custom GPTs
Best for: businesses deep in Microsoft 365 where the value sits inside Word/Excel/Outlook workflows

How we pick with clients

The default we install for most Australian SMBs is Claude Projects, one per business function. Five to seven Projects covers most companies: Operations, Sales, Marketing, Finance, Support, plus one or two function-specific ones (Product if you make one, Legal if you have an internal counsel, etc.).

We reach for Custom GPTs in three situations: when a workflow needs heavy external-API calling (Custom GPT Actions are still the most mature), when a relevant Custom GPT already exists in the OpenAI marketplace, or when a specific multimodal workflow (image-heavy or voice-heavy) genuinely fits ChatGPT’s strengths.

We reach for Copilot Agents when the business is deeply on Microsoft 365 and the most-valuable workflows live inside Word, Excel and Outlook. For these businesses Copilot Agents are the cleanest path; Claude Projects compete poorly against in-app value.

Governance scales differently across the three

Three observations from running rollouts:

Claude Projects proliferate the least because they require an explicit owner and the curated-context model penalises sprawl
Custom GPTs proliferate the most because they are fast to spin up and ownership defaults to whoever created them
Copilot Agents sit in the middle - Copilot Studio is involved enough that casual sprawl is rare, but real ownership still requires governance discipline

The pragmatic stack

For most Australian SMBs in 2026:

Claude.ai Teams with Projects as the company default
A small set of Custom GPTs for specific external-API workflows
Copilot Agents only if you are already deep in Microsoft 365 and the in-app value matters

The decision is not which platform is best in the abstract; it is which platform fits which workflow. Most teams need exactly one default plus a few exceptions. Pick the default deliberately and govern the exceptions tightly.

Frequently asked questions

What is the difference between Claude Projects, Custom GPTs and Copilot Agents?: All three are 'scoped AI assistants' configured once and reused. Claude Projects are workspaces with curated documents and instructions inside Claude.ai Teams. Custom GPTs are configurable assistants inside ChatGPT with optional API-calling actions. Copilot Agents are configurable assistants inside Microsoft 365 Copilot that live inside Word, Excel, Outlook and Teams. The choice is mostly about which underlying platform fits your workflows best.
Which is easiest to maintain at SMB scale?: Claude Projects, by some distance. The Project model maps cleanly to one-per-business-function ownership, the curated-context approach is small enough to maintain (typically 10-20 documents per Project), and there is a clear named owner per Project. Custom GPTs proliferate easily because they are quick to spin up; teams routinely end up with 30+ GPTs they cannot keep tidy. Copilot Agents fall between the two.
Which has the most powerful action capability?: Custom GPTs, today. The Actions feature in Custom GPTs is the most mature interface for calling external APIs from inside a scoped assistant. Claude Projects support similar action capability through MCP-shaped tools and the Anthropic API, and the gap is closing fast. For most SMB workflows the difference is academic; for workflows that depend on calling many external services from inside the assistant, Custom GPTs still lead.
Are Copilot Agents only worth it if we are deep in Microsoft 365?: Roughly yes. Copilot Agents shine because they live inside Word, Excel, Outlook and Teams. If your team works in Microsoft 365 all day, the in-app experience compounds value. If you spend most of your day in a browser, Slack and a non-Microsoft stack, the Microsoft-native value mostly disappears and Claude Projects or Custom GPTs are the better fit.
Can we use more than one of these?: Yes. Most XLev clients run Claude Projects as the default for company-wide scoped assistants, with a small number of Custom GPTs or Copilot Agents for workflows that genuinely fit those platforms better (multimodal-heavy work in Custom GPTs, in-document M365 work in Copilot Agents). The governance question is which platform owns which workflow and avoiding duplicated assistants in three places.

Where this fits

Claude Implementation

Install Claude properly across your team - Claude Code, Claude.ai projects and skills, custom Anthropic SDK builds.

Explore Claude Implementation →Book a free 30-min discovery call