JENKS ENG

Case study / Professional services

Safe, audited AI access to the systems a business runs on

A server that brings systems with no AI integration into reach of Claude, so staff can fetch data and automate routine work, with sign-in, permissions and an audit trail the business controls.

Industry
Professional services
Used by
Staff, through Claude
Our part
Design, build and manage
Built with
Cloudflare Workers, TypeScript, Terraform
Status
Live, in daily use

01 / The industry

Many systems, each with its own console

A growing services business runs on a stack of cloud systems: accounting, a CRM, a project tracker, Microsoft 365 for email and calendars, and the specialist tools its teams use every day.

Each has its own console and login. A simple question, such as what is overdue across every project, or how a client's account stands, means opening several of them and piecing the answer together by hand.

02 / The problem

AI access without losing control

AI assistants are good at exactly this kind of lookup, but most of these systems had no way to connect to one, and the ready-made connectors that did exist were too blunt for everyday work.

Connecting AI to accounting and client systems also raises hard questions. Who can see what? Who can change what? And how would anyone know afterwards? The brief was one connection point that:

  • lets staff sign in with the accounts they already use;
  • gives each person only the tools their role allows;
  • never changes anything without an explicit confirmation;
  • records every change against a named person.

03 / What we built

One server, many systems

We built a server on the Model Context Protocol (MCP), the open standard AI assistants use to call tools. Staff use it from Claude on the web, desktop and phone, and the business's own automations use it too.

  • Finance. Profit and loss, cash position, invoices and quotes, plus draft quotes and invoices for a person to check and send.
  • Operations. Read-only lookups in the specialist systems its teams work in, for faster answers to everyday questions.
  • Email and calendar. Send or draft email as the person asking, in the house style, and book events.
  • Project health. One tool runs fixed checks across the project tracker and the CRM, and returns only what needs attention, instead of pushing hundreds of raw records through the model.

Each system is a module. Adding one never changes the core, so the next system plugs into the same sign-in, permissions and audit.

04 / How it works

Control, security and an audit trail

  • Who can use what. Staff sign in with their existing work accounts. Each role maps to a set of permissions, the tool list is filtered for each person, and the permission is checked again on every call.
  • Changing things safely. Every change to a business system is previewed first and only made after an explicit confirmation. Accounting records are always created as drafts, and only tools on an allow-list can write at all.
  • Knowing what happened. An audit trail records who asked for what, when, and the outcome.

It runs on Cloudflare Workers, written in TypeScript, with the infrastructure in Terraform and automated tests run before every deployment.

05 / Where it is now

Live and in daily use

The server has been live since its first modules shipped and is used every day, and new systems are added as modules when the business needs them.

  • 50+tools across the business's systems
  • 1connection point for staff, assistants and automations
  • 3 monthsto build, a module at a time
  • 380+automated tests

06 / More work

Other case studies

Manufacturing

Keeping machine know-how when engineers move on

Machine knowledge lived in manuals nobody could search and in a few senior engineers' heads. We built an assistant that answers from the manuals and breakdown history, cites the exact page, and refuses rather than guesses.

AI · Search · AWS

Vehicle leasing

Salary sacrifice car leasing, joined up end to end

A vehicle leasing provider ran salary sacrifice schemes on a CRM, with manual steps at every stage. We built the pricing engine, portal, partner integrations and automation that take a deal from first quote to monthly mileage.

Platform · Integration · AWS

Commercial fit-out

Live dashboards from a decade-old project system

A commercial fit-out contractor's project and finance data sat in a decade-old system with fixed reports. We built live dashboards into it for forecasting, work in progress and daily changes, plus wallboards in its offices.

Data · QuickSight · SQL

See case studies

07 / Contact

Want AI to reach the systems you already use?

Tell us which systems your people work in every day, and we will say what could be connected and how it would be controlled.

Start a conversationEmail us