H&P Exec AI Data Platform

H&P Exec is an executive search firm whose most valuable asset is the knowledge in its Bullhorn database: candidate histories, consultant notes and years of market intelligence. Its team wanted to put that knowledge to work with AI assistants like Claude, but the data sat outside those tools, and the one-off integrations they had tried were fragile, costly to run and confined to a single chat.

The brief was to connect that data to AI properly: securely, with GDPR in mind, and without candidates’ personal information reaching an external AI provider by default.

We designed and built a multi-user MCP platform: an AI gateway between H&P Exec’s systems and the AI tools its team already uses. Staff sign in with their Microsoft account, connect Claude through a single secure link, and can then search, report on and update their Bullhorn and Xero data in plain English.

The platform gives Claude 38 purpose-built tools. On the recruitment side, staff can search across candidate notes, ask read-only questions of their Bullhorn data, from candidates and job orders to placements, and look up live records. On the finance side, they cover Xero invoices, bills and bank transactions, plus profit and loss, balance sheet and trial balance reports.

An admin dashboard lets H&P Exec decide who has access, which records each person can see, which tools they can use and what is masked. Bullhorn and Xero logins stay securely on the platform and are never shared with the AI, and the platform is hosted in London on Google Cloud.

Every safeguard is enforced by the platform rather than by asking the AI to behave.

Personal data masked by default. Names, emails and phone numbers are anonymised before results leave the platform. Staff can only reveal personal details if they have been given permission, and every reveal is logged.

The right access for each person. Each person only sees the records they’re allowed to, such as their own candidates or specific clients. These rules are enforced on the data itself, so they apply to every search and question, whatever the AI asks for.

Read-only by default. Questions of Bullhorn data can only read it, never change it, and access to Xero finance data is read-only.

Human-in-the-loop writes. Changes such as adding a note or updating a candidate’s status follow a preview-then-confirm step. The AI shows exactly what will change, and nothing is written until the user confirms. Bulk changes stay switched off unless an admin turns them on, and are limited in size.

A full audit trail. Every AI action, reveal and admin change is recorded in a permanent log that can’t be edited, so 100% of AI activity is audited.

H&P Exec’s data is now available to its team inside the AI tools they already use, under rules the business controls. The platform handles around 1,700 AI requests a month and saves 58 staff hours a month, as consultants ask complex questions of years of recruitment and finance data instead of compiling reports by hand. Leadership keeps oversight of who can see what, with an auditable record of how AI is used.

Because the platform is built to connect more systems over time, Xero was added as a second one without rebuilding it. That makes it a foundation for H&P Exec’s wider AI strategy, not just a single integration.