Fixed-fee assessment · 2 to 8 weeks
Will your AI use cases hold on your data?
A fixed-fee check of whether your priority AI use cases can ship on the data, platforms, controls and people you have today: what blocks each one, what fixing it costs, and who should fix it.
01 / Why pilots stall
Pilots rarely fail on the model.
First pilots often demo well on a curated sample, then stall on production data. The same customer sits under three IDs. Finance and sales define an active customer differently. Contracts have no metadata, and an assistant can see more than the user may see. Nobody has time to fix any of it.
The model is rarely the cause. The foundation is: entities, quality, definitions, documents, controls, and the people to run them.
02 / What we test
Twelve questions, tested on your own records.
We score each use case across twelve dimensions, deepest where it depends. Scores rest on evidence from your systems. An interview alone can't lift a score.
Use cases and value
Which AI use cases are worth doing first, and who owns the result?
Data sources and rights
Can we get the data each use case needs, and may we use it this way?
Platform and architecture
Will our platforms carry it at a cost we can predict?
Data engineering and operations
Will the data arrive on time, tested, every day?
Data quality and master data
Do we know who our customers, suppliers, products and counterparties are, in every system?
Semantics and metadata
Do we agree what our numbers mean, and can we trace them to source?
Documents and retrieval
Can AI find the right document, read it correctly, and respect who may see it?
AI and ML engineering
Can we build, measure and run it, and know when it's wrong?
Security, privacy and access
Will AI see only what the user may see?
Governance, risk and compliance
Can we show a regulator, auditor or buyer how each AI decision was made?
People and operating model
Do we have the people and owners to build and run it?
Roadmap and economics
What does it cost, in what order, and what does it return?
In scope
One to three named use cases (more for a multi-entity platform), evidence tests on real data, a roadmap, a 90-day pilot plan and a fixed-fee phase-one quote.
Out of scope
Building the pilot, remediation, platform RFPs, legal opinions, penetration tests and AI strategy without named use cases.
Data handling
Tests run inside your environment, under your access controls. We take no copies of personal data; only aggregated results leave.
Routing
Where the fix belongs to your security team or a partner, the readout says so.
03 / How it runs
Three sizes, one method.
Focused · 2 to 3 weeks
One use case in one business unit.
Standard · 4 to 6 weeks
Two or three use cases, or one acquisition.
Platform · 6 to 8 weeks
A buy-and-build platform with several entities, plus a cross-entity view of customers, suppliers and products.
Fixed fee for each size. We share pricing on the first call.
04 / What you get
A decision, not a binder.
- A verdict per use case: ready, ready with conditions, or not ready.
- A twelve-dimension scorecard, with the evidence behind each score.
- Blockers ranked by cost to fix, each with an owner and a fixer. Some fixes will be your own team's or a partner's, and we say so.
- A 90-day pilot plan with a success metric and kill criteria.
- A 6 to 12 month foundation roadmap with investment and run cost.
- A roles plan: hire, augment or partner, for each role the roadmap needs.
- A fixed-fee phase-one quote.
Who runs it.
A data and AI architect leads, with a master data and data quality lead. Master data and quality are where we're deepest, and where AI most often fails in companies built by acquisition. We're independent of every platform vendor.
Questions
Our platform vendor will do an AI assessment for free.
Take it; it will show you how to use their platform. Ours tests whether your data can support any platform, on your own records, and names who should fix each gap, including when the answer isn't us.
We already have an AI strategy.
Good, we start from it. Most strategies stop at the use-case list. We test whether the data under the top three will hold, and we price the fix.
I need results in 90 days, not a study.
The assessment is the first weeks of your 90 days. It ends with a pilot plan and a fixed quote, so the pilot starts with its blockers already known.
Security won't give outsiders access.
The tests run inside your environment under your controls, and only aggregated results leave. We'll complete your security questionnaire.
Can you guarantee our AI will work afterward?
No honest firm can. We stand behind the assessment's deliverables, and we design the pilot with a measured metric and kill criteria.
Which AI initiative is your board expecting this year?
In 30 minutes, we'll tell you which size fits and what we'd test first.