// Prickled.ai · Four platforms, one table

Four AI platforms, one table

Microsoft 365 Copilot, ChatGPT Enterprise, Gemini Enterprise and Claude, compared across eight dimensions in one table. All four now ship a credible version of the same stack. Here is exactly where they still differ, and where they quietly do not.

// Quick read · current as of August 2026

The short version

  • All four platforms now cover the same four layers: substrate, retrieval, method, action. The real differences are in defaults and governance, not features.
  • Microsoft 365 Copilot and Gemini Enterprise ground on your company data by default. Claude and ChatGPT Enterprise wait until you connect a source.
  • None of this decides whether the platform pays off. That is fluency, the fifth layer, and no vendor sells it on the pricing page.
01 The matrix

Side by side, in one table

Eight dimensions, four platforms, one screen. Read the row before you read the vendor's pitch deck.

Four platforms, eight dimensions. Current as of August 2026.
DimensionM365 CopilotChatGPT EnterpriseGemini EnterpriseClaude
Substrate you supplyNone. Microsoft-owned tenantAll of it, connected liveGoogle Workspace plus your connectorsAll of it, connected live
Default groundingInherits permissions. On by default. Cannot be disabledLive fetch. Off by default on Enterprise and EduInherits permissions. On by default, per-source toggleNothing until a Project or connector is attached
Reusable method objectPrompt Library. Unversioned, single promptWorkspace Agents. Job plus tools plus triggerGems. Shareable, knowledge-source-limitedSkills. Versioned, portable, org-provisionable
Action-layer protocolConnectors (ingest) plus MCP, GA 2026Connectors plus MCP developer mode, betaVertex/Data Cloud plus MCP and A2A, GA Aug 2026MCP-based Connectors only, live
Oversharing and governancePurview DLP, sensitivity labels, Restricted Content DiscoveryRBAC, admin console, audit logModel Armor guardrails, Drive trust rules, Vault eDiscoverySCIM, audit logs, Compliance API, IP allowlisting
Entry price, per seat$30/mo, annual commitment requiredBusiness: $25/mo, $20/mo annualNot published. Reseller estimates ~$21 to $30+Team: $25/mo, $20/mo annual
Enterprise tierSame $30 seat, requires qualifying M365 licenseContact sales, annual billing onlyContact sales, annual billing onlyFrom $20/seat plus usage, annual billing
2026 compliance highlightPurview plus Restricted Content Discovery bundledSOC 2 Type 2, HIPAA BAA (API); no ISO listedFedRAMP, HIPAA, CMEK, VPC-SCSOC 2, ISO 27001:2022, ISO 42001:2023, FedRAMP High (Jul 2026)
Prices and tiers change monthly in this market. Anything not marked "published" above is a reseller estimate. Reverify before it appears in a client contract.
02 What the table doesn't show

The fifth layer isn't on this table

Every row above is a layer-one-through-four decision: substrate, retrieval, method, action. None of it decides whether the platform pays off.

Layer 05
FluencyWhether the person knows what it is for. Not on this table, and not for sale.
Layer 04
ActionHow it reaches systems outside itself.
Layer 03
MethodHow the work gets done the same way twice.
Layer 02
RetrievalHow content becomes findable to the model.
Layer 01
SubstrateWhere the content physically sits.

That fifth layer is fluency, in Prickled's Fluency Layer framework: whether the person using the tool actually knows what it is for. It is not on any vendor's pricing page, and it is the one variable every row on the table above depends on to pay off.

FAQ Questions people ask

The platforms, answered

Straight answers, in case you skimmed, and the ones search engines tend to quote.

Which AI platform is best for enterprise: Copilot, ChatGPT, Gemini or Claude?

There is no single winner. All four now cover the same four layers, substrate, retrieval, method and action, so the real differences are in defaults and governance, not features. Microsoft 365 Copilot fits teams already standardized on Microsoft 365. Gemini Enterprise fits Google Workspace shops. ChatGPT Enterprise and Claude fit teams that want a portable assistant not tied to one productivity suite, with Claude ahead on compliance certifications as of August 2026.

Do Microsoft 365 Copilot and Gemini Enterprise see my company data automatically?

Yes, by default. Both inherit your existing file and folder permissions and start grounding answers in your content immediately, and Copilot cannot be fully disabled from doing so. Claude and ChatGPT Enterprise take the opposite default: neither sees your company data until you connect a source or attach it to a project.

How much do Copilot, ChatGPT Enterprise, Gemini Enterprise and Claude cost per seat?

As published in August 2026: Microsoft 365 Copilot is $30 per user per month with an annual commitment. ChatGPT Business is $25 per month, or $20 per month billed annually. Claude Team starts at $25 per month, or $20 per month billed annually. Gemini Enterprise does not publish a list price; reseller estimates put it around $21 to $30 or more per seat. Enterprise tiers on all four require contacting sales, and prices in this market change monthly, so reverify before it goes in a client contract.

What is Prickled's Fluency Layer framework?

A five-layer way to evaluate any AI platform. Layers one through four, substrate, retrieval, method and action, are what a vendor's comparison table measures: where your content sits, how the model finds it, how work repeats the same way twice, and how it reaches systems outside itself. Layer five is fluency: whether the person using the tool actually knows what it is for. Fluency is not sold on any vendor's pricing page, and it is the layer that decides whether the first four pay off.

What is the biggest governance difference between Copilot, ChatGPT, Gemini and Claude?

Where the controls live and what they are built for. Microsoft 365 Copilot leans on Purview for data-loss prevention, sensitivity labels and Restricted Content Discovery. ChatGPT Enterprise offers role-based access, an admin console and an audit log. Gemini Enterprise adds Model Armor guardrails, Drive trust rules and Vault eDiscovery. Claude offers SCIM provisioning, audit logs, a Compliance API and IP allowlisting, and as of July 2026 is the only one of the four with FedRAMP High authorization.

+ Sources

Where these figures come from

Primary vendor documentation only. Reseller and third-party estimates are labeled in the table above.

// This is what Prickled.ai does

Don't just pick a platform. Build the fluency underneath it.

Prickled helps you discover why you want AI, then places each tool where it actually fits, so the platform is never the strategy. You get a system built with you and owned by you, so the fifth layer, fluency, gets built alongside the tool instead of skipped.

Built with you · owned by you · Prickled, LLC.