Your people are already using ChatGPT, Claude, Gemini and Copilot, usually on personal accounts where nobody can see what company data goes in. As part of our managed IT services, we move them onto business accounts with admin controls, fix the permissions those tools inherit, and give your team a short policy they will actually follow.
Every AI tool your team uses, on accounts you control.
SF3 Technologies is a member of the Microsoft AI Cloud Partner Program and manages Google Workspace daily, and is a member of Anthropic's Cyber Verification Program. We advise on every platform below without a bias toward any of them.
ChatGPT Business or Enterprise workspace setup with SSO, admin roles, retention settings and the consumer accounts retired
Claude Team or Enterprise setup with the same controls, for companies that have chosen Claude or run both
Gemini for Google Workspace readiness: Drive permissions, shared drives and external sharing reviewed before it goes live
Microsoft 365 Copilot readiness assessment, SharePoint oversharing cleanup, Purview sensitivity labels and data loss prevention
A review of Perplexity, Meta AI, Mistral Le Chat and Grok use in your company, and business accounts where they are justified
A practical AI use policy, with a short list of approved tools and what must never go into them
Private or self-hosted AI options for legal, finance, health and engineering work that should never leave your control
Training for a pilot group first and then the wider company
How it works
Find out, put in order, keep in order.
01
We find out what is in use
Which AI tools your people already use, on whose accounts, and what kind of data goes into them. Most companies are surprised by the answer.
02
We move the work onto company accounts
One company workspace per tool, one sign-in, a retention rule, an owner, and a way to remove access on the day someone leaves.
03
We fix what the tools inherit
Copilot and Gemini read whatever a person can already open, so sharing and permissions are cleaned up before they go live, starting with HR, finance and legal.
04
Then a short policy and a pilot
A one-page policy your team will actually read, training for a pilot group first, and a review as the tools change.
Your tenant stays yours
We look after your Microsoft 365 or Google Workspace. You keep full ownership.
Every workspace we set up, from ChatGPT Business to Copilot, is created in your company's name and administered from inside your own Microsoft 365 or Google Workspace, through dedicated admin accounts in your name, protected by MFA and used only on your instruction. Nothing moves to SF3 Technologies' systems.
If you ever move on, we hand the workspaces, the policy and the documentation over to your new team, help during the changeover, and then remove our accounts, so you keep everything.
ChatGPT and Claude
Company accounts instead of personal ones
The most common AI risk in a small company is not a breach. It is a member of staff pasting a customer list, a contract or source code into a personal ChatGPT account, where the company has no visibility, no retention control and no way to remove it when that person leaves.
ChatGPT Business, the plan formerly sold as ChatGPT Team, and ChatGPT Enterprise fix most of that. OpenAI states that business data on these plans is not used to train its models by default. Admins get a console with workspace analytics, identity and access settings, permissions and roles, retention controls and SAML single sign-on, and Enterprise adds domain verification, SCIM provisioning and a Compliance API for audit logs. We set the workspace up, connect it to your Microsoft Entra or Google identity, decide who gets a seat, and then close the personal accounts in an orderly way: the data people need is moved, the rest is deleted.
Claude's Team and Enterprise plans from Anthropic work the same way. Content is not used for model training by default, both plans support single sign-on, and Enterprise adds SCIM, domain capture, audit logs, role-based access and a compliance API. If your engineers, lawyers or analysts prefer Claude, we set it up with the same discipline rather than tolerating it in the shadows.
Gemini and Copilot
AI inside the data you already have
Gemini for Google Workspace and Microsoft 365 Copilot are different from a chat window. They read the mail, files and chats a person already has permission to open, which means they also surface everything that was shared too widely over the years.
On the Google side we review Drive permissions, shared drives that have grown out of control, link sharing defaults and external sharing before anyone turns Gemini on, and we run a pilot with a small group first.
On the Microsoft side we run a Copilot readiness assessment: where your data lives, who can open it, and what would appear on day one. Then we clean up SharePoint and OneDrive oversharing, starting with HR, finance and legal, apply Microsoft Purview sensitivity labels so confidential content stays confidential even when Copilot can see it, and set up data loss prevention so people cannot paste or send what the policy forbids. A tenant where permissions are correct is safer whether or not you ever buy a Copilot license.
Your staff will not stop at four tools, so neither do we. For each tool we answer the same three questions: does a business account with admin controls exist, do you need one, and what do staff who keep using the free version need to know.
Perplexity: popular with sales and research teams. Perplexity Enterprise Pro and Enterprise Max add SSO and SCIM, user management, audit logs and configurable file retention, and Perplexity states that it does not train its models on enterprise customer data. We move heavy users onto the enterprise workspace and leave casual users on the policy.
Meta AI: built into WhatsApp, Instagram and Facebook. Personal WhatsApp chats stay end-to-end encrypted and are not seen by Meta AI, but anything a person sends to Meta AI is received by Meta. In companies where WhatsApp is the main customer channel, we set a clear rule: Meta AI is not for customer data, contracts or financial information, and work chats run on the WhatsApp Business tools your company controls.
Mistral Le Chat: Mistral is a French company, which matters to buyers in Germany and the rest of the EU. Le Chat comes in Free, Pro, Team and Enterprise versions; Team adds role-based access and central billing, and Enterprise adds SAML single sign-on, audit logs, central user management and the option to run it self-hosted or in a private cloud. For companies that want a European provider under European law, it is a serious option and we set it up accordingly.
Grok: xAI's assistant, built into X. The consumer version is tied to a personal X account, so it belongs on the "not for company data" list unless your company has set up one of xAI's business or enterprise workspaces with licenses and organization controls.
Policy and private AI
An AI use policy people will follow, and private AI when it matters
Most AI policies are three pages long and nobody reads them. Ours fits on one page: the approved tools and accounts, the data that never goes into any AI tool, how AI output is checked before it reaches a customer, and who to ask. We write it with you, adapt it for GDPR in Europe and local data protection laws such as Colombia's Ley 1581, and keep it current as tools change.
Some work should not go to any public AI service at all: privileged legal matters, unreleased financials, patient data, source code under contract. For those cases we set up private or self-hosted AI, either an open model running on hardware or cloud you control, or an enterprise deployment with no data retention, and we document where the data is processed so you can show a data protection officer or a customer.
ChatGPT Business is enough for most companies under about 150 people, because it already excludes your data from model training by default and gives admins SSO, roles, analytics and retention controls. Enterprise adds domain verification, SCIM provisioning and a Compliance API for audit logs, which matters once an auditor, a regulator or a large customer asks for them.
Give them a company account that is better than the personal one, then make the rule clear and enforce the parts that can be enforced. We set up the business workspace, retire personal accounts, add data loss prevention in Microsoft 365 or Google Workspace for the most sensitive content, and write a one-page policy that says exactly which tools and data are allowed.
Yes, Anthropic offers Claude Team and Enterprise plans that do not use your content for model training by default and support single sign-on, and Enterprise adds SCIM, domain capture, audit logs and role-based access. We set it up and manage it alongside ChatGPT, Copilot and Gemini rather than making your team choose one for our convenience.
Copilot is only as safe as your existing permissions, so review SharePoint and OneDrive sharing, apply Purview sensitivity labels and pilot with a small group before a wide rollout. Done in that order it is safe; switched on blindly it will surface files people should never have been able to open.
Personal WhatsApp chats stay end-to-end encrypted, but anything a person sends to Meta AI is received by Meta, so customer data, contracts and financial details should never go into it. We set a clear rule for Meta AI, Grok and other consumer assistants and move the real work onto accounts the company controls.
Yes, if every tool runs on a business account covered by a data processing agreement, your data is excluded from training, and the content that must stay out of scope is labelled and blocked. We document where each tool processes data, which is also where a European provider such as Mistral or a private deployment can be the right answer.
Not sure which AI tools your team already uses, or on whose accounts?