AI Integration

Connecting AI to the Systems You Already Run

AI that can read your actual data, through your actual permissions. We built an MCP server around the NSW curriculum for NESA.
Talk to us about AI integration
In short

AI integration connects AI tools to the systems you already run, including HubSpot, Microsoft 365, SharePoint, Umbraco and Xperience by Kentico, so AI can act on real data instead of guessing. Webcoda built an MCP server around the NSW curriculum for NESA, one of the first of its kind in Australian government, so an AI tool can query the curriculum directly and accurately.

Who this is for

Scenario 01
A large content library nobody can query
Thousands of structured documents, syllabuses, policies or products that are technically published and practically unsearchable.
Scenario 02
AI tools inventing answers about your own data
Staff ask an assistant about internal information and get a confident answer assembled from nothing in particular.
Scenario 03
Systems that should share data and do not
Two or three platforms holding overlapping records, reconciled by a person with a spreadsheet.

If there is no AI in it, you want the other page

If you need general systems integration without an AI component, our Systems Integration service is the right page. This one is specifically about connecting AI models to your data and platforms.

System Integration

What we deliver

  1. 01

    Platform integration

    HubSpot, Microsoft 365 and SharePoint, Umbraco, Xperience by Kentico, Kontent.ai, and custom platforms. Named, because entity clarity is part of the job.

  2. 02

    MCP server development where it fits

    A server that exposes your content to AI tools as structured, queryable data with its own permissions, rather than pasted into a prompt.

  3. 03

    API and data-layer work

    The unglamorous part: schemas, sync, rate limits, error handling and the reconciliation logic nobody budgets for.

  4. 04

    Security and permissions review

    Which model sees what, what is logged, what leaves your tenancy, and who signed off on that.

What  MCP actually is

The Model Context Protocol is a standard way for an AI tool to ask a system for information, and for that system to answer in a structured, permissioned way. Instead of copying content into a prompt and hoping, the model queries the source.

The practical difference is accuracy and provenance. An MCP server returns the actual record, so the answer can cite it, and the answer changes when the record changes. Nothing is baked into a model that then quietly goes out of date.

It also puts access control in the right place. The server decides what a caller may see, using your rules, rather than an assistant deciding what looks relevant.

For NESA we built a server around the NSW curriculum, so an AI tool can ask about a syllabus outcome and get the real one. It is one of the first of its kind in Australian government, and it is the reason this page exists rather than a generic integration page.

DefinitionS

Model Context Protocol (MCP)
An open standard for connecting AI models to external data sources and tools through a defined server interface.

MCP server
The component you build and host. It exposes your data and enforces who can read what.

Why it matters here
Retrieval with provenance, permissions kept where they belong, and answers that update when the source does.

What we have built

nesa-ai-case-study-temp
Government · Education · MCP

NESA curriculum MCP server

An MCP server built around the NSW curriculum, so an AI tool can query syllabus content directly and return the actual outcome rather than an approximation of it.

One of the first MCP implementations in Australian government, and the clearest evidence in this whole service suite that the work is real.

How an engagement runs

  1. Step 01

    Assess

    We map what the data actually looks like, not what the documentation says. Sources, owners, permissions and the fields that are quietly free text.

  2. Step 02

    Prototype

    We stand up a narrow integration against real records. One content type, one consumer, measured for accuracy before anything is widened.

  3. Step 03

    Integrate

    We build the production path, with permissions and logging. Auth, rate limits, failure behaviour, and a monitoring view someone will actually look at.

  4. 8.7

    Scale

    We extend to the next system, or stop at one. Integration work compounds, so the second connection should be cheaper than the first. If it is not, something in the design is wrong.

Honest scoping

Shape
Discovery on the data first
Integration estimates made before looking at the data are fiction. The first piece of work is a short, paid look.
Shape
One connection, then a decision
First integration is scoped as a standalone piece of work with its own value, not a phase of a programme.
Ongoing
Hosting and maintenance
An MCP server is a production system. It needs hosting, monitoring and an owner. That can be us or your team.

AI integration, answered

The Model Context Protocol is an open standard that lets an AI tool query a system directly, through a server you control, instead of relying on content pasted into a prompt. It matters because the answer comes from the real record, can be attributed to it, and updates when it changes.
No. Those are the most common, along with Umbraco, Xperience by Kentico and Kontent.ai, but the work is the same shape against any system with an API and a permissions model worth respecting.
Regular integration moves data between systems on rules you define. AI integration also has to handle a caller that asks open-ended questions, so provenance, permissions and accuracy testing carry much more weight. If there is no AI in your project, our Systems Integration service is the better fit.
Not by us, and not by the model providers under the enterprise terms we build against. We will show you the specific boundary for the setup you are on, in writing, rather than reassure you in general.
Often, yes, depending on the platform and the sensitivity involved. That constraint changes the architecture, so it belongs in the first conversation rather than the last.

Talk to us about AI integration

Bring the system nobody can query. That is usually the right place to start.

Contacts

For all your AI Integration, HubSpot, Umbraco, Web Development, Web Design, Intranets and digital marketing projects, connect with us. Tailored solutions for success.

Logo-hubspot Logo-umbraco Logo-B-Corp Logo-xperience Logo-microsoft