Context Visualised home Talk to us
Bring your own AI

Bring your own AI.

CoVi Analytics doesn't bring AI to your data. It brings your data to your AI.

Bring your own AI means CoVi Analytics works with the AI assistant your firm already uses. It adds no model of its own. We are building it so your assistant can read CoVi Analytics and draft a control test, an incident or a KRI commentary. A named person then approves it.

Talk to us Follow Halden

In build Available from late October 2026.

Your AI · CoVi Analytics

Built to work with the AI you already use

  • Copilot
  • OpenAI
  • Anthropic
  • In-house model

Your AI draftsA named person approvesCoVi Analytics records it

The problem

Every new AI tool is another model to approve.

Risk teams want help with the drafting: test results, incident write-ups, KRI commentary. Most tools that offer it bring their own model. That means a new supplier, a new data flow and new questions from IT, legal and the board.

There is a second problem. An AI that writes straight into the record makes it hard to say who decided what. Under the FCA's senior managers regime, a named senior manager is still accountable.

What you get with most AI tools
  • A new model to approve and isolate
  • Your data sent somewhere new
  • Changes that can skip a named approver
What we are building
  • The AI your firm already trusts and pays for
  • Your data read where it already sits
  • Every change approved by a named person
How it will work

It drafts. Your people approve.

Your AI will never change CoVi Analytics on its own. It drafts from what it can see, then asks a named person to approve.

  1. Your AI will read and draft.

    It will read CoVi Analytics through the MCP connector we are building. It will draft a control test result, an incident or a KRI commentary.

  2. The owner will approve in Teams.

    The control owner will get a Teams card with three buttons: Approve, Edit and Open in CoVi Analytics.

  3. CoVi Analytics will record who and when.

    On approval, it will record who approved the change and when. The dashboards will update.

Approval card · in Teams

Concept: in build

CoVi AnalyticsHalden · Controls

Coverholder audit evidence found for C-137. Review and approve?

Control
Coverholder audits
Linked risk
Coverholder binding outside authority
Drafted by
Your Copilot

The card offers three actions: Approve, Edit, and Open in CoVi Analytics.

Concept design of a card we are building. C-137 is Halden's coverholder audit control.
Reporting on demand

Ask why a number moved, and get the answer from approved data.

Today a director who wants to know why a KRI turned red waits for the quarterly pack. With CoVi Analytics connected, they will be able to ask in their own words.

Example · Halden

"Why is the large loss ratio red?"

Halden is our fictitious managing agent. Copilot would answer from its large claims and appetite reports. The ratio is 18.1% against an 18% limit, driven by eight large losses and cat events.

Questions risk teams ask

Can AI help a risk team draft board risk reports?

Yes, if it drafts and a person approves. AI is good at turning approved data into first drafts: KRI commentary, incident write-ups, a committee summary. It should not change the record on its own. We are building CoVi Analytics so your own AI drafts from its data, and a named owner approves each draft.

Do we need a new AI model to use AI in risk reporting?

No. If your firm already uses Microsoft 365 Copilot, it can read other systems through connectors. Microsoft supports MCP servers as plugins for Copilot agents, signing in as that person. We are building our connector that way, so Copilot only sees what that person is allowed to see.

We will show it working at Halden first.

When the connector ships, we will run it at Halden in the open: a draft, a Teams approval and the dashboard updating.

Already use an AI assistant?

Tell us how your firm uses it today. We will show you how CoVi Analytics will connect to it, using Halden as the starting point.

Talk to us Follow Halden