Data readiness first, then one use case at a time
Everyone will sell you AI. We will tell you if your data is ready.
Most AI projects fail on the data, not the model. We settle which figures are right first, then build one thing that works.
Most AI spend has not paid back yet.
25% of AI initiatives have delivered the return expected of them over the last few years.
IBM Institute for Business Value, 2025
16% of AI initiatives have scaled across the whole enterprise.
IBM Institute for Business Value, 2025
26% of companies have the capabilities to move beyond proofs of concept and generate tangible value. The other 74% have yet to show it.
BCG, 2024
4% of companies have built cutting-edge AI capabilities across their functions.
BCG, 2024
60% of AI projects unsupported by AI-ready data will be abandoned through 2026, Gartner predicts.
Gartner, 2025
68% of chief executives say an integrated data architecture across the enterprise is critical to cross-functional work.
IBM Institute for Business Value, 2025
The drawing is to scale: a hundred goes in, and each stream out is as wide as its share.
So we check the data first, and measure one task against where it started.
What we take on.
What we report back
We report the task before and after, measured against its baseline.
- We turn the AI question into one task somebody does every week.
- We check whether the data behind it exists, has an owner and agrees across systems.
- We build the platform in your own tenant, on Databricks.
- We set who may read which table, and trace every figure to its source.
- Your board pack and any assistant read the same governed figures.
- We fix permissions before anything is allowed to retrieve on your behalf.
- We measure answers on questions we did not choose.
Bronze, silver, gold.
This is a published pattern that we build to. Raw data is kept apart from the figures a director quotes, and each layer is made from the one below it.
Bronze
What it holds
Everything arrives exactly as it was sent, wrong rows included, and nothing is corrected or thrown away.
What you can trust it for
It proves what a source system said on a given day.
Silver
What it holds
The rows from bronze are joined, tested and de-duplicated, and the rules that reject a bad row are written down.
What you can trust it for
It is safe for analysis, but not for a board pack, because no definition is agreed yet.
Gold
What it holds
The figures the business has agreed on are built once from silver, with their definitions attached.
What you can trust it for
Reports and any model that answers questions read from this layer only.
Building the chat window first.
The assistant is built first because it demonstrates well, and it is signed off on the demonstration.
Asked for last quarter's revenue, it picks one of four figures in four systems and shows no sign of having chosen.
We get the definitions agreed and written down first, so it answers from figures the business has already argued about.
We do not publish a filled-in copy, because its contents belong to the client. Get in touch and we will walk you through a real one.
Contact usQuestions we get asked.
Your assigned Solutions Architect oversees the work from the first workshop to handover, supported by secondary support staff working from the same documentation. Client Services help with scheduling and account queries. We confirm who is assigned before work starts.
You can, and everyone finds out what they could already open. An assistant sees what its user sees, so the permissions come first.
It is the reason to talk to us. We find the smallest part one useful thing depends on, and start there.
Not when the platform sits in your own tenant, which is what we usually recommend. If a use case needs an outside service, we name it and what it could see first.
Where this page sits.
Our work sits in three layers, and each one depends on the one beneath it. The page you are reading is marked.
You are hereAI Adoption & Data Strategy
AI
- You are here AI Adoption & Data Strategy
- AI Solution Packages
- Document Extraction
- Document Analysis
- Proposal Workflows
Start with whether the data can carry it.
The Fortnight looks at the estate this would sit on, and the findings are yours either way. If the honest answer is a year away, the report says so.
See AI Studio review documents with source citations
You get a written scope before anybody commits to anything.