Your staff may already be using AI, but on personal accounts and without rules for company data. We help you choose between ChatGPT Business, Claude Team and Microsoft 365 Copilot based on where your team works and what data may go into AI. Then we help set up company accounts and permissions (with your IT, or for you) and train people on their own tasks.
Company AI rollout at Webforte is help with choosing, setting up and training a whole team on an AI assistant. Typically this means ChatGPT Business or ChatGPT Enterprise from OpenAI, Claude Team or Claude Enterprise from Anthropic, or Microsoft 365 Copilot, which Microsoft now calls Microsoft Copilot. We choose the tool based on where the company works day to day (Microsoft 365, Google Workspace or elsewhere), what data may go into AI and what administration IT needs. According to all three vendors' terms, their business plans do not use company data to train models by default; they differ mainly in how they connect to company systems and how they are administered. We write down rules for using AI, help set up company accounts and permissions, test the tool with a pilot group and prepare shared templates and training in Czech. The service is provided by Webforte Technologies s.r.o. (company ID 23364343), Prague. Pricing is set by a non-binding quote after a consultation.
All three can write, summarise and translate. They differ in where they work, which company data they can reach, where they process data and what administration they give IT. The matrix shows which tool comes out best in which situation. We give the final recommendation only once we have seen your tasks.
| Your situation | OpenAIChatGPTBusiness and Enterprise | AnthropicClaudeTeam and Enterprise | MicrosoftMicrosoft Copilotformerly Microsoft 365 Copilot, an add-on to Microsoft 365 |
|---|---|---|---|
| The company runs on Microsoft 365Email in Outlook, documents in SharePoint and OneDrive, meetings in Teams. | WorksOffice add-ins and appsAdd-ins for Word, Excel and PowerPoint, @ChatGPT in Teams and apps for SharePoint and Outlook. Some actions need a Microsoft 365 administrator's approval. | WorksOffice add-ins and a connectorAdd-ins for Word, Excel and PowerPoint. The Microsoft 365 connector searches SharePoint, OneDrive, Outlook and Teams. | StrengthBuilt into the Microsoft 365 appsIt sits in Word, Excel, PowerPoint, Outlook and Teams. With a Microsoft 365 Copilot (Premium) licence it finds the email, files and meetings the user has access to by itself. |
| The company runs on Google WorkspaceGmail, Google Drive and documents in Google Docs and Sheets. | StrengthGoogle Drive, Gmail and SheetsAn app for Google Drive including Docs, Sheets and Slides, a Gmail connection and an add-on for Google Sheets. | StrengthGmail, Calendar and Google DriveBy default it asks for your approval before sending an email or moving a file. | LimitedMakes no sense without Microsoft 365It is an add-on to Microsoft 365. It can bring in data from Google services through a connector. |
| Long documents and analysisContracts, reports, data spreadsheets, research. | StrengthData analysis and deep researchRuns Python calculations on spreadsheets. Deep research is included in the Business plan and searches the web and connected apps. | StrengthHundreds of pages in one chatProjects with source material can be shared with colleagues in the company. | WorksResearcher and Analyst agentsThey come with the paid licence, as does the full experience in Word and Excel. Researcher draws on the web and on company files, email and meetings. |
| We want to connect our own systemsInternal documentation, CRM or databases. With all three this is a separate piece of work. | WorksCustom apps via MCPMCP (Model Context Protocol) is an open standard for connecting AI to systems. The workspace administrator enables the apps. | WorksCustom connectors via MCPClaude inherits the user's permissions in the connected service, and the administrator can limit a connector to read-only. | WorksConnectors and Copilot StudioReady-made connectors, for example for Google services, Salesforce or Confluence. The administrator decides which agents people may use. |
| We need administration, SSO and auditSign-in with the company account (SSO), creating and removing accounts, audit records. | WorksSSO in Business, SCIM in EnterpriseBusiness has user and role management, SSO and domain verification. Automatic account provisioning (SCIM) and the Compliance API come only with Enterprise. | WorksSSO in Team, SCIM in EnterpriseTeam has SSO and account creation on first sign-in. SCIM, audit logs and the Compliance API come only with Enterprise. | StrengthAdministration and audit from Microsoft 365It takes accounts and permissions from Microsoft 365. Work with Copilot is recorded in the Microsoft Purview audit, where the retention period is also set. |
| We want data processed in the EUA requirement from customers, your data protection officer or internal policy. | WorksEU storage only in EnterpriseEligible new Enterprise customers can choose data storage in Europe (EEA and Switzerland); the Business plan does not have this option. | LimitedData stored in the USBy default, processing may take place in the US, Europe, Asia and Australia. | StrengthWithin the EU Data BoundaryApplies to EU customers. Anthropic models are excluded from it, so they are switched off by default in the EU. |
| Best fit when | the company works in Google Workspace or wants custom GPTs shared across the company and @ChatGPT in Teams | the team works with long documents and the administrator wants tight control over what connectors may do | the company runs on Microsoft 365 and wants administration, audit and data processing in the EU |
We checked these features on the official OpenAI, Anthropic and Microsoft websites on 30 September 2026. Vendors change their plans and features often, so we check them again before you buy. We do not list licence prices; the vendor sets them.
Sources:OpenAI: business data (opens in a new tab)OpenAI: data residency (opens in a new tab)Anthropic: model training (opens in a new tab)Claude: SSO and account provisioning (opens in a new tab)Microsoft 365 Copilot: licences (opens in a new tab)Microsoft 365 Copilot: data and privacy (opens in a new tab)
Most companies do not start from zero. Someone already uses AI and management knows only part of it. You will recognise these signs.
One person uses free ChatGPT, another pays for Claude out of pocket, a third is trying Copilot. The company does not know who uses what and has no control over personal accounts.
On a personal account, each person decides on the data settings. The company cannot see what went into the tool and cannot delete it.
There are no simple rules on which tools the company allows, what data does not belong in them and who is responsible for the output. So people either hide their AI use or do not use it at all.
Whoever has fine-tuned their prompts for quotes or meeting notes keeps them in their own account. Everyone else starts from scratch, and when that person leaves, the know-how leaves too.
A business licence covers the contract with the vendor and account management. What people put into AI and what they do with the output is for the company to decide.
OpenAI states that it does not use data from ChatGPT Business and Enterprise to train models by default. Anthropic says the same about its commercial Claude products, and Microsoft about prompts, responses and Microsoft Graph data in Copilot. Free and personal accounts have different rules, which is why company work belongs only in the company account.
Microsoft Copilot only surfaces company data the person has at least view permission for. ChatGPT and Claude connectors inherit the permissions the user has in the connected service. If a payroll folder is accidentally shared with the whole company, AI will answer from it for anyone. That is why we review permissions on sensitive folders before launch.
A short document people actually read. It starts from your tasks and from the data you hold.
All three vendors have a data processing addendum (DPA): OpenAI signs one for ChatGPT Business and Enterprise, Anthropic's is part of its commercial terms, and Copilot is covered by Microsoft's Products and Services DPA. Where data is processed differs from tool to tool; the matrix above has the overview. We do not replace a legal opinion: we prepare material for your lawyer or data protection officer.
Six steps, each ending with a concrete output. The tool goes company-wide only after the pilot. The timeline depends on team size and the number of tasks.
With managers we go through what people do repeatedly: quotes, replies to customers, meeting notes, work with spreadsheets. We pick tasks where AI saves time and where mistakes are easy to spot.
Output: A list of tasks suited to AI and a reasoned tool recommendation.
We write down rules for using AI and help set up company accounts and permissions (with your IT, or for you): company sign-in, user management and the data settings the chosen plan offers.
Output: AI usage rules and a configured company workspace.
A few people from different teams get the tool first. We watch where AI helps them, where it makes mistakes and what needs adjusting in the rules or templates.
Output: A pilot review and updated rules.
Company-wide only after the pilot review. If the tool does not prove itself, we switch it before you buy licences for everyone.
For the selected tasks we prepare prompt templates and shared projects with source material, so nobody starts from scratch. For example a Quotes project with the price list and samples, or a Meetings project with a notes template.
Output: A library of templates and shared projects for your team.
If AI needs to search company documents or work with other systems, we enable ready-made connectors (SharePoint, Google Drive and others) and check who can see what through them. Your own systems are connected via MCP (Model Context Protocol, an open standard for connecting AI to company systems). This is a separate scope of work that we quote separately.
Output: Enabled connectors with verified permissions, or a design for a custom connection.
We train the team hands-on, on its own tasks and in Czech. This includes AI literacy, which the AI Act talks about: how the tool works, where it makes mistakes and how to check its output. For companies in the Czech Republic: a company can draw a contribution for the 50-hour programme from the Vzdělávání pro firmy scheme of the Czech Labour Office (ÚP ČR). That programme is delivered by our partner PECKA.
Output: A trained team that uses AI on its own tasks.
The outputs stay in your company even after our work together ends.
We use AI assistants daily in development and build our own products on them, so we know where they help and where they get things wrong. We run AI training with our partner organisation PECKA.
Tell us how many people would use AI, where your email and documents live and what you want it for. In a consultation we go through your tasks and data rules and prepare a non-binding quote for the rollout. We communicate in English or Czech.
