Can my AI assistant prepare a board pre-read on AI accountability?

Yes. The PulseAI MCP server exposes your Compass score trajectory and open findings, so Claude, ChatGPT, Copilot or any MCP-aware assistant can summarize quarter-over-quarter movement and surface the top findings the board should know about.

Can my AI assistant help with ISO 42001, EU AI Act FRIA or Canada AIA audit prep?

Yes. Your assistant can list every open finding mapped to a specific clause, with owner and due date. The same MCP endpoint exposes the underlying AI system metadata and threat-to-control mappings.

Can I track AI KPI progress through ChatGPT or Claude?

Yes. The KPI plan and current trajectory are read tools on the MCP server. Ask your assistant where you are tracking and where you are slipping; it pulls live from PulseAI.

Can my AI assistant recommend what to do next on AI accountability?

Yes. The recommendations tool combines your Compass findings with matching use cases from your use-case intelligence, ranked by feasibility and impact. Ask for the top three and your assistant returns them with rationale.

Can I see every AI tool registered in our organization from my AI assistant?

Yes. The AI Tool Registry is exposed as a read tool. Your assistant can list tools by registration date, approver, data classification tier, or any combination.

Can I compare our AI accountability posture against industry peers from my AI assistant?

Yes. The MCP server exposes anonymized peer benchmarks (k≥5 minimum cohort) by sector and domain. Ask your assistant where you lead and where you trail; it pulls live from the cross-tenant benchmark layer.

AI Assistant Access: Query Your AI Spend, Risk & Outcomes from Claude, ChatGPT or Copilot

What this is

An MCP server for your AI governance posture.

Model Context Protocol is the emerging open standard for connecting AI assistants to external data sources. PulseAI runs an MCP server for every tenant. Once connected, your assistant can read your governance posture the same way it reads your email or your calendar.

The server is read-only. It exposes Compass scores, score trajectories, findings, AI system metadata, KPI plans, the tool registry and the recommendations engine. It cannot register systems, close findings, change scores or modify anything in PulseAI.

Per-tenant authentication. Structural data residency in your region. Queries are never used to train any model.

What you can ask

Six questions, answered by the assistant you already use.

board pre-read · 60 seconds

Board pre-read in 60 seconds

Board chair · CEO · Corporate secretary

"Summarize how our Compass score moved this quarter and the top three findings I should mention to the board."

Replaces the manual briefing pack. Pulls directly from score trajectory and open findings.

audit prep

Audit prep without the spreadsheet

CISO · GRC lead · Internal audit

"List every open finding mapped to ISO 42001 Clause 6.1.4, with owner and due date."

The auditor's questions, answered by the auditor's own AI assistant.

kpi tracking

Are we hitting our AI KPIs?

CFO · COO · Department leads

"Show me where we're tracking against the AI KPIs we set last quarter and where we're slipping."

Closes the loop on 'we set targets, did we hit them?' without opening the platform.

recommendations

What should we do next?

CIO · Department leads · PMO

"Based on our current scores, what are the top three recommendations and which use cases match?"

Compass findings to a vetted use-case playbook, in one prompt.

tool registry

What's actually deployed?

CISO · IT · Procurement

"Show me every AI tool registered in the last 30 days, who approved it, and its data classification tier."

The Tool Registry, surfaced through the assistant the team already uses.

peer benchmark

How do we compare?

Board chair · CEO · Strategy

"How does our governance domain score compare to anonymized peers in our sector?"

Same benchmark, conversational interface. No dashboard to learn.

How it works

Three steps. Then ask anything.

01 · Provision

We provision an MCP endpoint scoped to your tenant. The endpoint lives in the same AWS region as the rest of your PulseAI data (ca-central-1 for Canadian tenants, eu-central-1 for EU, us-west-2 for US).

02 · Connect

Add the endpoint as a connector in Claude, ChatGPT, Microsoft Copilot, or any other MCP-aware assistant. Setup is the same flow as adding any other connector.

03 · Authenticate and ask

Authenticate once. Then ask anything across your governance posture, score trajectory, findings, KPIs and tool registry. The assistant calls only the read tools available to your role.

Trust and scope

Built to enterprise governance constraints.

no mutations

The MCP server exposes no write tools. It cannot register AI systems, close findings, change Compass scores or alter anything in PulseAI.

tenant isolated

Every call is enforced at the protocol layer. There is no cross-tenant data path. Your assistant sees only your governance posture.

in-region

Same regional split as the platform: Canada, US, EU. Your MCP endpoint runs in your region. No routing layer, no configuration toggle.

no training use

Your queries and our responses are never used to train any model, ours or a provider's. Connectors do not change that.

Your AI accountability is now a conversation.