Skip to main content
    The Digital Hub

    Private AI

    Private AI systems for UK businesses

    The Digital Hub designs and builds private AI systems for UK organisations that need to use company information without giving uncontrolled access to confidential data.

    Deployments can be UK-hosted, on-premises or hybrid, with permission-aware retrieval across Microsoft 365, SharePoint, Outlook, CRM records, files and operational systems.

    Book a private AI demo
    Download the diagram

    Who builds and supports the system

    Private AI is bought on trust, so it is worth knowing exactly who is doing the work and what is agreed before anything is connected.

    • A UK team based in the York area, at English House, Railway Street, Pocklington.
    • Discovery, architecture, implementation and ongoing support are handled by that team.
    • Experience building connected CRM, document, customer-service and operational systems, not AI alone.
    • The deployment is designed around the systems you already run rather than one fixed model.
    • Accounts, data and the agreed code-ownership position are clear in writing before implementation.
    • A written data-flow and permissions design is produced before any source is connected.

    Company summary

    Company
    The Digital Hub
    UK base
    English House, Railway Street, Pocklington, York, YO42 2QZ
    Telephone
    01904 900201
    Area served
    United Kingdom
    Delivery
    UK-hosted, on-premises and hybrid private AI systems
    Typical starting route
    A defined pilot with measurable success criteria

    From question to answer — inside your own environment

    1

    Your team

    People across the business ask questions and request information in plain English.

    • •Staff
    • •Managers
    • •Departments
    • •Hybrid workers
    2

    Your business information

    Approved sources remain authoritative; access follows the permissions agreed during discovery.

    • •CRM — customer & business data
    • •Files & documents — policies, contracts, plans
    • •Emails — Outlook & communications
    • •Reports & spreadsheets — financials, operations, data
    • •Forms & systems — operational systems & tools
    3

    Your private AI business hub

    Retrieves from approved business sources, answers with citations, drafts content and creates reports within the permissions and guardrails agreed for the deployment.

    • •UK-hosted, on-premises or hybrid
    • •Controlled model gateways
    • •Permission-aware answers and audit logs
    4

    Useful outputs

    Better decisions. Faster. More time for what matters.

    • •Answers — clear, accurate responses
    • •Reports — instant, consistent reports
    • •Summaries — key point summaries
    • •Actions — tasks and next steps
    • •Insights — spot trends, risks, gaps

    Ask it like a person

    The tools your team already uses — indexed and ready to answer

    The hub connects to Microsoft 365, Outlook, Excel, Azure and your CRM, files and emails. Everything is indexed in one place — so anyone can ask a question in plain English and get a clear answer back in seconds. No querying databases, no chasing the right spreadsheet, no "I'll come back to you." And because it runs inside your environment, you can use it every day with the confidence it won't touch, change or break your existing systems.

    Connects to

    • Microsoft 365 logoMicrosoft 365
    • Outlook logoOutlook
    • Excel logoExcel
    • Teams logoTeams
    • OneDrive logoOneDrive
    • SharePoint logoSharePoint
    • Azure logoAzure
    • Google Drive logoGoogle Drive
    • Gmail logoGmail
    • Gemini logoGemini
    • OpenAI logoOpenAI
    • ChatGPT logoChatGPT
    • Slack logoSlack
    • + much,
      much more

    Connects to Microsoft 365, Google Workspace, leading AI models (OpenAI, ChatGPT, Gemini) and your own private AI instance — plus CRM, files, emails and operational systems. Bring your own model or let us set one up.

    "What did we quote this client last year?"
    "Summarise the key risks in this contract"
    "How did each site perform last week?"
    "Find the policy on holiday carry-over"

    Real-world use case

    A week in the life of a private AI hub

    A 90-person construction firm spreads its information across a CRM, shared drives, Outlook inboxes and a dozen spreadsheets. Before the hub, finding a single answer meant three phone calls, two inboxes and an hour nobody had.

    The hub applies agreed source permissions and returns cited answers. Where contracts require it, the model and retrieval layer can run inside the organisation's network boundary.

    1

    Tender documents

    Before

    A bid deadline hits at 5pm. Answers to 'what did we quote last time?' live across three folders and two inboxes — nobody can find them fast enough.

    After

    Someone asks the hub in plain English. It pulls the last three quotes, the relevant clauses and the client history in under 30 seconds.

    Bids submitted on time, with the right context — every time.

    2

    Policy lookups

    Before

    HR gets the same question about holiday carry-over every week. Each time it's a search through the staff handbook, then an email, then a follow-up.

    After

    The hub answers directly from the current policy document, cites where it found it, and drafts the reply for approval.

    Repetitive questions handled in seconds, not hours.

    3

    Operational reports

    Before

    The ops manager spends Monday morning stitching together five spreadsheets to answer 'how did each site perform last week?'.

    After

    The hub reads the spreadsheets and produces a consistent, formatted report on demand — with the figures that actually matter.

    Monday mornings back for actual management.

    4

    Contract summaries

    Before

    A 40-page supplier contract lands. Legal is booked out for a week. Nobody else wants to touch it.

    After

    The hub summarises the key terms, renewal dates, penalties and obligations in minutes — flagged for the right person to review.

    Critical dates never missed, risk surfaced early.

    The outcome

    1

    Pilot use case agreed first

    3

    Deployment options

    25

    Users in our cost example

    100%

    Questions available for audit

    This is a worked example of the workflow, not a published client result. The pilot measures search time, answer quality and adoption before a wider rollout or savings claim is made.

    How to choose a UK private AI supplier

    Use these eight questions with any supplier, including us. Each one below states how we answer it in practice.

    1. 1

      Can the supplier show exactly where your data travels?

      You receive a written data-flow and permissions design before a single source is connected: what is indexed, what leaves your tenant, which model processes it, what is retained and for how long.

    2. 2

      Are Microsoft 365 and source-system permissions preserved?

      Retrieval is permission-trimmed against the user's own access before content reaches the model, so a user cannot see through the assistant what they cannot see in SharePoint or the CRM.

    3. 3

      Are UK-hosted, on-premises and hybrid all genuinely available?

      All three are offered and recommended on the basis of your information classes, not on what is easiest for us to run.

    4. 4

      Can refusals and unauthorised access be tested, not just correct answers?

      Pilot testing includes negative tests: restricted users asking restricted questions, with the results reviewed alongside the audit trail.

    5. 5

      Are citations returned with every answer?

      Answers cite the source document, record or message so a person can verify the claim rather than trust the model.

    6. 6

      Is there a named human owner of content quality and support?

      Ownership of content freshness, retention and support is agreed and written down at handover, on both sides.

    7. 7

      Will they recommend a smaller pilot or an off-the-shelf option?

      If Microsoft Copilot or a narrower pilot solves the problem, we will say so in discovery. Most engagements start as a defined pilot with measurable success criteria.

    8. 8

      Are setup, model use, hosting, support and improvement priced separately?

      Each is quoted as its own line so you can see the build cost and the running cost before committing.

    Evidence: what is live, and what is illustrative

    Implemented work and worked examples are kept separate on this page so neither can be mistaken for the other.

    Implemented deployment

    York Racecourse — Ailison AI concierge

    Business problem
    A high volume of repeat visitor questions on race days and in the lead-up — tickets, access, hospitality, timings and facilities — arriving by phone and email into a small team.
    Sources connected
    Approved published venue information, event and ticketing content, and operational FAQ material maintained by the venue.
    Permissions model
    Public-facing scope only. The assistant answers from an approved public content set and has no access to customer records or internal systems.
    Deployment boundary
    UK-hosted, with a controlled model gateway, citations back to source content and an audit trail of questions asked.
    Result
    Routine enquiries answered without staff involvement, with unanswered questions logged and used to improve the source content. Figures are reviewed with the venue before publication.
    See the AI concierge in detail

    Worked example — not a client result

    Construction document retrieval

    An illustrative workflow showing how a permission-aware system would answer questions across drawings, RAMS, variations and site correspondence held in SharePoint. It is published to show the architecture and the permissions logic, not to represent a measured outcome at a named client.

    Pilots in progress are described the same way. Nothing on this page is presented as a live client result unless the client is named and the result has been agreed with them.

    What it connects to

    Source documents remain in their existing systems. The retrieval and model data flow is documented for the chosen UK-hosted, on-premises or hybrid deployment.

    CRM

    Customer & business data

    Files & documents

    Policies, contracts, plans

    Emails

    Outlook & communications

    Reports & spreadsheets

    Financials, operations, data

    Forms & systems

    Operational systems & tools

    Useful outputs

    Better decisions. Faster. More time for what matters.

    Answers

    Clear, accurate responses

    Reports

    Instant, consistent reports

    Summaries

    Key point summaries

    Actions

    Tasks and next steps

    Insights

    Spot trends, risks, gaps

    Deployment options

    Private does not mean one fixed architecture

    The right boundary depends on your contracts, information classes, existing infrastructure and appetite for operating local models. We document the data flow before anything is connected.

    UK-hosted private cloud

    A dedicated UK-region environment with identity integration, controlled model gateways and an audit trail. This is the usual balance of governance, speed and running cost.

    On-premises

    Open-weight models and retrieval run on infrastructure you control. This suits contracts or information classes that prohibit data leaving your network boundary.

    Hybrid

    Sensitive sources stay on your network while approved, permission-filtered extracts use a managed model through a controlled gateway.

    Pilot first. Roll out with evidence.

    1. 1–2 weeks

      Discovery and data audit

      Agree the questions, authoritative sources, information classes, permissions and a measurable pilot outcome.

    2. 2–4 weeks

      Pilot

      Connect a limited source set, test retrieval and refusals with a small user group, and review the audit trail.

    3. 2–6 weeks

      Controlled rollout

      Add approved sources, tune permissions, train users and document ownership, retention and support.

    4. Monthly

      Ongoing operation

      Review answer quality, source freshness, access, model use and the next safe workflow to add.

    Indicative investment

    What a private AI rollout costs

    These are planning ranges, not quotes. A 25-user internal knowledge system usually sits between £12,000 and £30,000, with ongoing hosting, model use and support commonly between £250 and £900 a month.

    Focused pilot

    £6,000–£12,000

    One use case, a limited source set, a small user group and clear success criteria.

    25-user knowledge system

    £12,000–£30,000

    Microsoft 365 sources, permission-aware retrieval, citations, audit logging, training and rollout.

    Complex or on-premises

    £30,000–£60,000+

    Several data classes, bespoke integrations, local model infrastructure or stricter assurance work.

    Who it suits

    • Teams repeatedly searching SharePoint, Outlook, shared drives or a CRM.
    • Organisations that need source citations, permissions and reviewable logs.
    • Businesses with an owner for content quality and a measurable pilot problem.

    Who it does not suit

    • Teams without reliable source material or named information owners.
    • Businesses expecting AI to replace judgement or guarantee correct answers.
    • Low-volume use cases where existing search or a process change is enough.

    Documented security and governance

    Source documents stay in their existing store. A managed deployment sends only permission-filtered extracts under the provider terms agreed for that project; an on-premises option is available where contracts require processing to remain inside your boundary. Questions, answers and sources can be audited.

    Private by design

    Deployment and data flows are chosen for the information being handled.

    Strict access control

    Role based access and full audit logging.

    Data protection

    Encryption and retention controls documented for the chosen deployment.

    Controlled gateways

    Only approved connections for specific business needs.

    You stay in control

    Your data, your rules, your AI.

    Built for organisations that need documented permissions, data flows, auditability and a deployment matched to the sensitivity of their information.

    Next step

    Get in touch

    Tell us what your team spends too long looking for and we'll show you how a private AI hub handles it — hosted in your environment, while you maintain control over your data architecture.

    • ✓ Reply within one working day
    • ✓ No obligation, no sales pressure
    • ✓ Clear scope before anything starts

    We'll only use your details to reply to this enquiry.

    Private AI questions

    Costs, compliance and how it works

    How much does a secure private business AI system cost?

    What a private, UK-hosted AI system actually costs to build and run — with the numbers from a live deployment.

    Read

    Can a company use AI without sharing confidential information?

    The architectures that let you use AI on sensitive material while maintaining data sovereignty, and what each one costs.

    Read

    UK-hosted AI and GDPR: what businesses need to get right

    Data residency, lawful basis, DPIAs and transfers — the compliance groundwork behind a defensible AI deployment.

    Read

    Building an internal AI knowledge system with SharePoint and Outlook

    What it takes to make your Microsoft 365 content answerable — permissions, indexing, costs and the pitfalls nobody mentions.

    Read