Data Readiness for AI
The gap between having data and AI using it.
Every stalled AI pilot has the same autopsy: the data wasn't ready. We measure the gap — quality, structure, access — and build what closes it, in the order that unblocks your first use case.
What we build
- Readiness assessment against your actual AI use cases
- Cleaning and structuring pipelines for the data that matters first
- Retrieval preparation — chunking, indexing, and evaluation where it earns its place
- Access shaped around your permissions and controls
When teams come to us
- An AI pilot stalled on data quality
- A vendor said “just connect your data” and it was never that simple
- The Discovery Sprint found data as the real first step — this is that step
How we build it
Assess against the use case
readiness is relative to what you want to build
Fix in value order
the subset that unblocks first value, first
Prepare access and retrieval
permissions, structure, indexing
Verify with the AI itself
evaluation on real queries, not assumptions
The four questions
Readiness is not a property of a database. It is the answer to four questions, asked about one use case at a time:
Access
Can the system that will use the data actually reach it, under the permissions the data already has? "It's in the warehouse" and "the model can read it" are different states, and the gap is usually a month of work nobody budgeted.
Quality
Is it true? Duplicates, stale records, fields that mean different things in different systems. A model trained on it will be confidently wrong in exactly the ways the data is.
Structure
Does it have the shape the use case needs? Retrieval wants chunks and an index; a predictor wants a table with a label; an agent wants records it can act on. The same data is ready for one and not the other.
A path
Is there a way from where the data is born to where the model reads it that runs without a person? A pipeline someone re-runs by hand is a demo, not readiness.
Most stalled pilots fail one of the four. Almost none fail all four — which is why the fix is usually smaller than the fear.
What “ready” turned out to mean
The mistake we see most is treating readiness as a project that finishes before AI begins: clean everything, structure everything, then start. It never finishes, and the pilot waits a year for a foundation it needed a tenth of. Readiness is relative to a use case. An assistant that answers questions from your documentation needs those documents indexed and current, and nothing else. A model that predicts churn needs one table, joined, labelled and refreshed — and nothing else.
So the work runs in value order: the subset of data that unblocks the first use case, made reliable first, with the AI itself as the test. If the assistant answers real questions correctly from the data as prepared, the data was ready. If it doesn't, the evaluation tells you which of the four questions is still open — and that is the next week's work, not a rewrite of the plan.
Works well with
FAQ
How long does readiness take?
Depends what the use case needs — readiness for one assistant is not readiness for everything. That’s why we scope to first value.
Is our data too messy for AI?
Messy is the default state of real companies. The question is which subset matters first.
Do we need everything ready before starting?
No. Readiness and the pilot advance together.
What do you actually deliver?
For the use case in scope: an assessment that names which of the four questions are open and what closing each one costs; the pipelines and structures that close the ones on the critical path; and an evaluation set of real queries with known-good answers, so "ready" is a measurement rather than an opinion.
We have a data warehouse. Isn't that enough?
It is a good start, and often it is not the problem. Warehouses are built for reporting: aggregated, refreshed nightly, shaped for dashboards. Retrieval needs documents indexed at query time; agents need records they can act on. The warehouse tells us the data exists; readiness is about the shape it needs for the thing you want to build.
Does our data have to leave our environment?
No. Readiness work happens where the data already lives — your cloud, your permissions, your audit log. The pipelines we build run there, and what we take away is the assessment and the evaluation set, not the data.
Find out exactly where your data stands.
The sprint includes technical feasibility and data requirements.