Services · AI & MCP
AI interfaces for the systems your business runs on.
Connect your AI assistant to selected business information and tools. We handle the integration: matching records, applying business rules and defining who can access what.
Start with a question your team already asks.
What should we know before a customer review? Which support issues relate to an upcoming renewal? Where is the information needed to investigate an escalation?
To answer these questions, the assistant needs to find the right records and understand how they relate. We work out which information it needs, who may use it, and how to connect it through Model Context Protocol (MCP) or another suitable interface.
We check what is possible with your systems and chosen assistant before proposing the work.
How we get started.
Tell us what you want the assistant to do and who will use it. If detailed discovery is needed, we agree its scope and fee before it begins. Implementation follows a separate written proposal with agreed acceptance criteria, price and delivery plan.
Choose the workflow
Agree the question, the people who need the answer, the source systems and what a useful result looks like. Check whether an existing connector already covers the need.
Build the interface
Start with read-only access to the information needed for the task. Set up sign-in and access checks, link the records, and show the source of each answer and when it was updated.
Review the result
Test normal questions, missing information and denied access with the intended users. Agree any additional connectors, production rollout and ongoing support separately.
What we deliver.
The proposal sets out what we will build and how we will test it:
- A description of the task, users, connected systems and access rules.
- A working interface for the agreed systems and assistant.
- Test results covering expected answers, missing information and denied access.
- Documentation and an agreed plan for rollout and support.
The proposal also identifies any new connectors, data preparation, hosting and ongoing support needed for your setup.
Where ClientShape fits
Already using ClientShape?
We can build on your existing ClientShape integrations to give an AI assistant access to selected customer information and tools. We agree what it may access and check compatibility with your chosen assistant.
You can also work with us on AI integration without ClientShape. A direct connection to one of your systems may be all you need.
Before we start.
What is MCP?
Model Context Protocol is a standard way for AI applications to discover and use tools provided by another system. We design those tools around the business task, with defined inputs, outputs and access rules.
Will it work with our assistant and access requirements?
We check the chosen assistant, connection method, identity provider and deployment requirements during scoping. Compatibility and access rules are verified for the agreed environment.
Where will our data go?
We map the data flow during scoping: where the interface is hosted, which assistant and model provider you choose, and which records and fields may be passed to them. We review authentication, access, logging and retention requirements for those services before implementation. MCP defines the connection; the chosen services and deployment determine where information is processed.
How is the work priced?
Any paid discovery has an agreed scope and fee before it starts. A separate implementation proposal follows. Extending an existing integration is different from connecting a new customer's systems; new connectors, data cleanup and production rollout are identified explicitly in the scope.
Can an assistant change business records?
We recommend read-only access first. Actions that change records need a separate design for permissions, review, audit and recovery before they are introduced.
Tell us what your AI should be able to help with.
Tell us about the task, the systems involved and the people who will use it. We'll help you work out a sensible place to start.
Talk to us about a use case