Guide

Managed AI

AI for your whole organisation, on data you can trust.

4 min read

AI has made generating things cheap: anyone in your organisation can now produce an analysis, a report or an automation in minutes. It has also made verifying things more valuable than ever, because AI presents wrong answers with exactly the same confidence as right ones, and it will never tell you which is which.

Our answer is not to slow your people down. It is to give them full freedom on a foundation that cannot lie. We call it Managed AI, and it rests on three commitments.

One version of the truth

Everything begins in your data warehouse: the one central place where data from all your systems is brought together, cleaned and checked. We build and guard it: structured, tested, documented and verified, so that every number that matters has exactly one definition.

If that foundation is right, an experiment can only ever be harmlessly wrong; if it is wrong, every tool on top of it, human or AI, is confidently wrong at scale. That is where our senior attention lives, and nothing built on top can compensate for skipping it.

Real freedom for your team

Your power users, the people in your team who live in the numbers, connect the AI assistant of their choice through one secure, read-only door: an MCP server we operate. Behind it sits certified data, carrying the definitions the AI needs to understand what it is looking at.

And behind that door, the freedom is real: your people explore, analyse and prototype without waiting in a queue. No requests stuck in someone's inbox, no bottleneck: genuine self-service, on numbers that are correct at the source.

A gate before anything goes live

A good prototype does not creep into daily use; it earns its way in. Your user hands it back together with a specification written by the AI they used: what it does, the logic it applies, the results it should produce.

We review the code, validate the data against the warehouse, and implement it properly on your internal platform: documented, monitored, and giving the same answer every time. Your team keeps the speed. Your organisation keeps one truth.

What stays yours

Everything. The platform, the environment, the code, the documentation: you own all of it. What you keep us for is not a dependency; it is the memory and judgment behind the platform: knowing why every decision was made, deciding what is worth verifying, and noticing when a workload has outgrown its approach.

We have run this model since 2011, through three generations of tooling: first in QlikView, with a protected data foundation the business built its own analyses on; then in Qlik Sense and Power BI, with certified datasets and self-service exploration on top. AI is simply the third generation.

The Managed AI model. At the bottom your sources, which come together in the data warehouse (1): one version of the truth. On the left, free for your team: through the read-only MCP server (2), power users work with AI. On the right, governed by ThinQ: work that earns a place in daily use crosses over as files with a specification (3) to the ThinQ review (4), and then runs on the internal platform. Your team · free ThinQ · governed Power users + AI explore and prototype 2 MCP server one door, read-only 4 ThinQ review code and data verified Internal platform same answer every time 3 1 Data warehouse one version of the truth: modelled, tested, verified Invoicing Planning Accounting Excel ERP

The model at a glance

The numbered blocks match the three commitments: 1 is the truth, 2 is the freedom, 3 and 4 are the gate.

Read it bottom-up: everything stands on the warehouse (1). Your team climbs the left tower through the MCP server (2), one read-only door onto certified data, to explore and prototype freely.

Work that earns a place in daily use crosses over as files plus an AI-written specification (3), passes our review (4), and only then runs on your internal platform, which itself stands on the same trusted foundation.

How we get started

Each step delivers something that works on its own: you see value before you commit to the next.

  1. Scoping

    A short first step: we map your current data landscape (sources, systems, and the definitions your business actually runs on).

  2. The foundation

    We build or harden the warehouse and the shared business definitions behind it, until there is one version of the truth worth building on.

  3. Open the door

    A pilot group of power users gets MCP access and starts exploring and prototyping on certified data, with the AI tools they already use.

  4. The first handover

    Together we run the first full cycle (prototype, specification, review, go-live) and agree the working rhythm: what stays free, and what passes through the gate.

Curious what this means for your team?

References and documentation are available on request. A short conversation is enough to see where you stand.

Book an introduction