Case study 02

Not just answers — an AI that acts in your systems

A production MCP server that lets More Sailing's team query twelve data sources and execute actions directly in their own operational systems through natural language — agentic BI with enterprise-grade guardrails.

Client
More Sailing AB
Type
AI data platform · Agentic BI
Focus
Natural-language analytics · Safe AI actions
The challenge

Dashboards answer the questions you predicted. The team needed answers to the questions they hadn't thought of yet — without learning query languages, and without risking the data.

With the BI platform in place, More Sailing had the data — but every new question still needed an analyst. The team wanted to ask ad-hoc questions conversationally, and go further: read reservations, update crew lists, and adjust charter pricing directly from an AI assistant.

Giving an AI agent access to business data and live pricing systems demands real security: strict access control, read-only analytics, and a full audit trail.

Users
Non-technical staff, analysts & admins
Surface
Any MCP-capable AI assistant
Risk profile
PII, pricing mutations, financial data — zero tolerance for uncontrolled access
The solution

A production MCP server with security at its core

“MCP moresailing” — a production server of roughly 12,000 lines of code, running in the cloud — connects AI assistants to the central data warehouse and the live charter system.

Natural language → answers

Questions over the full warehouse, guarded by a strictly read-only query layer with per-role access to data.

Role-based access control

Viewer / analyst / admin roles plus fine-grained per-user capabilities gating every action tool and PII-exposing interface.

Two-way — the AI takes action

Via the AI assistant, the server executes actions directly in the customer's own systems: managing crew lists and updating charter pricelists (absolute, percentage and bulk dynamic pricing) — not just reading data.

Enterprise auth & audit

A dedicated enterprise-grade authentication layer, a secured admin dashboard, and a scrubbed audit record for every gated action.

In practice

From question to answer — and from decision to done — in one sentence

One conversational layer covers twelve data sources and the operational charter system. Because the AI both answers questions and carries out the follow-up action in the same conversation, work that used to span analyst requests, system logins and manual updates collapses into a single step — a strong efficiency lever.

  • “How is charter occupancy in Croatia tracking against last year?” — answered directly from analytics tables, no analyst required.
  • “Raise shoulder-season prices 5% on underbooked catamarans” — executed against the live charter system, permission-checked and audit-logged.
  • “What did we decide about the spring campaign?” — shared team memory keeps decisions and context available to every colleague's assistant.
The outcome

Self-serve analytics — and an AI that safely operates the business

12+
data sources queryable in natural language
100%
of gated actions audit-logged — across 3 access roles with per-user capability gating
2-way
read and write — the AI acts directly in live systems
Agentic AINatural-language analyticsAccess controlEnterprise authenticationAudit loggingTeam memory

Want your business data to speak your language?

Book a no-obligation intro call and we'll map the opportunities together. We reply within one business day.

Book an intro