Claude Enterprise
Deploy Claude Enterprise across your company
Softblues puts Claude Enterprise into everyday use inside a business: governed access for the people in scope, the sources and procedures it is allowed to work from, the workflows worth encoding once, and training built around the jobs your teams already have. We deploy it, and we run our own company on it.
Claude is one platform of several we work with. If you have not settled that question, start with the comparison rather than with us.
- OpenAI Select Partner
- Google Cloud Services partner
- Anthropic Partner Network member
- Member of the Microsoft AI Cloud Partner Program
The service
What does Softblues put in place?
Four pieces of work. The first decides whether to do the other three, and it is allowed to conclude that you should not.
Discovery and fit
What the work is, what it costs today, and whether Claude is the right platform for it. This is also where the answer can be that it is not, or that something cheaper would settle the question. See what strategy and roadmap covers.
Governed deployment
Sign-in, the connectors to the systems people already work in, the permissions each person carries into it, and the retention position, agreed with whoever owns IT and data protection in your company. See what deployment and activation covers.
Training around real work
Role-based sessions built on the tasks a team already has, using the company’s own documents and its own way of writing. A feature tour teaches people what the product does; this teaches them what to do on Monday.
Ongoing improvement
A named owner inside your company, a monthly look at what is actually being used, and a decision about what to add, change or stop. See what continuous partnership covers.
Before you commit
How do you establish whether Claude fits?
Four questions worth settling first. They are the same four whichever platform you end up on, which is the point: answering them is what makes the choice defensible rather than a preference.
- What would people actually do with it
- Name the tasks, not the department. Reading long documents, drafting to a house standard, analysing a dataset and answering the same policy question forty times a week are four different jobs, and a platform that is strong at one is not automatically strong at the next.
- What it has to be able to read
- The sources that hold the answers, where they live, and whether they are current and correctly permissioned today. Auditing them, fixing what is stale and correcting the permissions is work in its own right, and it is work whatever platform you choose.
- Which controls you have to be able to show somebody
- Who can use it, what it can reach on their behalf, how long anything is kept, and what evidence exists afterwards. Write these down before you compare products, because they narrow the field faster than a feature list does.
- Whether you have compared it with the alternatives
- Claude is one route of several, and choosing by brand is the expensive mistake. The platform selection page sets out the criteria and the routes worth investigating, the adoption hub covers the wider question of becoming AI-native, and a consulting assessment ranks the candidates across a business before committing to one.
How it fits together
What does a governed deployment actually consist of?
Four parts, holding at the same time rather than one after another. Three of them are work somebody does rather than a setting somebody switches on, so budget them as work.
Who can use it
Access
Sign-in configured for the people in scope, and then a connector at a time: check what each one actually lets those people reach, and set it to what you intend rather than to what it defaults to.
What has to be settledWhat each connector permits, one by one, and whether your existing permissions express what you want today. The Microsoft 365 connector, for instance, passes the signed-in person’s own access through, so it inherits whatever is already there, including the shared drive nobody has tidied since 2019.
What it may read
Approved knowledge
The procedures, templates, policies and house language the assistant is allowed to work from, so its answer is your company’s answer rather than a generic one.
What has to be settledWhich sources are current, who owns each of them, and what happens when one changes. A confident answer drawn from a superseded policy is worse than no answer.
What it does repeatedly
Workflows
The jobs worth encoding once rather than re-explaining every time: the weekly pack, the standard document, the first pass over a long file.
What has to be settledWhich work is repetitive enough to be worth encoding, and which is better left to a person asking in their own words. Not everything that can be encoded should be.
Who answers for it
Human ownership
A named person approves anything that leaves the building or goes on a file, and a named person owns whether the whole thing is still working a quarter from now.
What has to be settledWhich of those two people is which, and what each of them is accountable for. Settle it before the rollout rather than after it: the approver and the owner are different jobs, and neither is the IT lead by default.
What the four produce
A record of what happened
Audit logs and an agreed retention period turn “staff are using an AI tool” into something you can answer questions about afterwards. Which controls are available to you depends on the plan your company buys, and the current position is worth confirming against Anthropic’s own documentation rather than taking it from a page like this one.
Evidence
What does this look like when it has been done?
Two pieces of work, and it is worth being precise about what each one establishes.
Rollout and training delivered
Key Capital's company-wide Claude rollout
Key Capital is an Irish corporate finance and investment management firm. Softblues ran the discovery and licensing fit, configured Claude inside the existing Microsoft 365 environment, and led nine weekly training sessions. The rollout went live in April 2026.
Read the case study
Running in-house
The Softblues Claude Operating System
Softblues runs its own company on Claude: six connected agents, one per function, each with its own operating manual, tools and routines, joined through a shared hub. We use it before we sell it, so we can show it live rather than in a slide deck.
Read the case study
Images on this page are illustrative. Each card names the stage its work reached.
Where is Claude put to work?
Seven places we are asked for most often. Each page covers what the work looks like in that function and what it takes to set up.
Common questions about Claude Enterprise
What is Claude Enterprise?
Claude Enterprise is Anthropic’s plan of the Claude platform for organisations. Alongside Claude itself, it is where the administrative controls a company needs sit: single sign-on and SCIM provisioning so joiners and leavers are handled centrally, role-based access control, audit logs, a custom retention period and a compliance interface for retrieving what was asked and produced. Anthropic does not train its models on enterprise customer data by default, and that is worth having stated in the contract rather than relied on as a setting. This describes the position as at September 2026; plans and features move, so confirm the current detail against Anthropic’s own documentation before you buy on the strength of it.
What is shadow AI, and why does it matter?
Shadow AI is people using consumer AI accounts for work because nothing approved does the job. The material risk is not the tool, it is that confidential material ends up somewhere with no retention policy, no access control and no record that it happened, so nobody can answer what was shared or by whom. Giving everyone a governed assistant that is genuinely good enough removes the reason to reach for the unapproved one, which works better than a policy telling people not to.
What can Claude see once it is connected to Microsoft 365?
The Microsoft 365 connector uses delegated permissions, so Claude acts as the signed-in person and reads what that person can already read. Connecting it needs one-time tenant consent from an administrator, and write actions are a separate, later decision that needs its own approval. The part worth planning for is that delegated access inherits the permissions you already have, including the ones nobody has reviewed for years, so a permissions clean-up is usually real work in its own right rather than something the licence settles. This is the position for that connector, verified September 2026; it is not a general statement about every system a company might connect.
Where is our data held?
Anthropic’s first-party surfaces, which is the API, claude.ai and the Office add-ins, store data in the United States, and the workspace geography is not something a customer changes. For most UK and Irish companies that is a documentation task rather than a barrier: the transfer mechanism exists in Anthropic’s data processing terms, and what it needs is a transfer risk assessment, a record in your third-party register and an update to your privacy notice. Where a client contract or a public-sector framework specifies UK or EU-only processing for particular data, that data needs a different route and we would design around it. We do not present any of this as a regulatory clearance: it is a position to agree with whoever owns data protection in your company, on current documentation.
How is Claude different from ChatGPT Enterprise or Microsoft 365 Copilot?
They differ on where they sit rather than on a score. Copilot is inside the Microsoft applications people already have open, which matters most when the work is mail, documents and meetings. Claude and ChatGPT are their own surfaces with their own ways of packaging a company’s procedures and reaching its systems, and they suit longer reading, drafting and analysis where the work is not already inside an Office file. Which one fits follows from the tasks, the sources it has to reach, the controls you need and who will use it. The platform selection page works through that comparison properly, and we would rather you read it than take a preference from a page about one of them.
How long does a deployment take?
It depends on how many teams are in scope, how much preparation the sources need and how quickly permissions and approvals can be agreed inside your company, so the honest answer is that discovery settles it and the plan carries a timeline you can hold us to. The programme runs in three parts: strategy and roadmap, deployment and activation, then continuous partnership. We are not going to print a number of days here for an engagement nobody has scoped yet.
Is Softblues an Anthropic partner?
Softblues is a member of the Anthropic Partner Network, and also a partner of OpenAI, Google Cloud and Microsoft; the exact status of each is stated at the top of this page. Membership of a partner programme is a commercial relationship. It does not certify a person, approve a product or say anything about what any platform will do for your company, and we would not want it read as though it did. What is worth weighing instead is that we deploy this platform and run our own company on it.
Start with what you want to improve
Tell us the work you want Claude to take on and how it runs today. We can discuss whether this platform suits it, and whether a scoped discovery is the right next step.
We answer enquiries within 24 hours. The first conversation covers your objective, the tools you already run, and whether Claude is the right platform for it.