AI Data-Readiness Check
Will an AI answer your questions correctly from your warehouse?
The AI Data-Readiness Check takes three of your own questions, asks them of an AI on your warehouse, and scores the answers against the ones you trust. Every wrong answer is traced to a fix, with the effort it takes.
For SaaS companies at Series A and B that run a cloud warehouse and a dbt project and plan to switch an AI on over their data.
1 · what you get
A verdict, a scorecard and a fix list, in five working days.
Everything is delivered as a report and a pull request to your repository. It is a check, not a build: nothing is installed in your cloud.
- 01
The verdict, on one page
Ready, Ready with fixes, or Not yet, with a scorecard for your three questions and the top three fixes.
- 02
Your three questions, scored
For each question: the agreed answer and its source, then what an AI returned when it saw only your raw schema, and again once it had the definitions we wrote. Every wrong answer comes with its SQL and its cause.
- 03
Metric definitions
Every metric the three questions need: where it is defined today (a dbt metric, a BI tool, a spreadsheet, or nowhere), where definitions conflict, and the one we propose.
- 04
Model and documentation coverage
For the models your questions touch: descriptions, tests, documented grain, freshness and naming.
- 05
The fix list
Each fix with its effort (S is half a day or less, M is one to two days, L is three to five), the questions it fixes, who owns it and where it sits in the order.
- 06
The test kit, in your repositoryyours to keep
The questions, the agreed answers and a runner, delivered as a pull request, so you can re-run the check after every fix.
What the report does not say: it covers these three questions in the wordings we agree, not every question an AI might be asked. It is not a guarantee of any accuracy figure, and it says so on its first page. Every query we run is logged, and the log is yours.
2 · what we need from you
Three questions, a read-only role, and a little of your team's time.
✓ From you
- Three questions your business asks of its data, and the answer you trust for each, with the date window and who or which report defines it.
- A read-only role you create from our script: SELECT on the agreed schemas only, personal-data columns masked or excluded, and a cost guard. You switch it off at the readout.
- Read access to your dbt project, or a zip of it with its manifest and catalog.
- A 45-minute kickoff, a named contact for up to 30 minutes a day, and a 45 to 60 minute readout.
- A signed agreement and data-processing agreement before any access is granted. Invoices come from Reventus Services Private Limited, trading as tessDeck.
✕ Not for you if
- You have no warehouse or dbt project yet: there is nothing to check, and we will say so on the first call.
- You need a guaranteed accuracy figure, a generic AI-strategy deck, or your data copied out to us.
- Your data is patient-level health data. We can run a metadata-only check; ask first.
- Your purchase needs seven or more approvers and a security review. We are too new for that today.
3 · the price
$2,500 for the first 3 checks. $4,500 after that.
One fee, fixed before we start, paid on delivery.
The founding price carries the same credit terms. No countdown and no reservation: we take three starts a month, and if a month is full we will say which month is not. All prices.
4 · your data
What leaves your warehouse, stated plainly.
5 · questions
Before you write.
- Who does the work?
A person does the check and reviews every finding; software does the inventory, the runs and the first draft of the scoring. We do not name past employers or clients.
- Do you have case studies?
No, and we would rather say so. We have no client results we can attribute yet, and inventing some would make every other number on this site worthless. When a public sample exists it will say plainly that it runs on synthetic data.
- What happens after the check?
You can fix everything yourself from the list, ask us to do the fixes as a fixed scope, or stop. Nothing further is sold until the check is delivered.
Start with a 20-minute call.
Write to us with your name, your company and the AI you plan to switch on. We reply with three times that suit you.