A fixed-scope audit and sprint that scores your organization against the 17 components that make AI knowledge systems actually work — then builds the priority ones, so every AI system you build after starts on solid ground instead of a guess.
A taxonomy, a source-of-truth map, and a permissions layer aren't specific to one project. Build them once and every AI system after inherits them — which is why Data Readiness sits at Stage 0, underneath everything else.
Score the 17 components, build the priority ones.
Your first working AI system, built on the readiness artifacts.
Run and expand it monthly against the same rubric.
New departments become new Activations, not new discovery.
The second use case starts at implementation — discovery is done.
Taxonomy and permissions get paid for once and reused, not re-billed.
This is the work that keeps a project off Gartner’s 60% list.
One glossary means Sales, Legal and Finance AI don’t quietly disagree.
Lineage and versioning are what auditors ask for later — built once.
Both are fixed scope, fixed price, fixed duration — not open-ended advisory. The Sprint contains the Audit, plus construction.
Answers one question: is your organization's knowledge ready for AI? We inventory your systems, interview the people who own the data, and score you against the 17-component model.
Goes further: we build the highest-priority components ourselves — taxonomy, source-of-truth map, metadata schema — and validate with a retrieval test on real business questions.
The signal is consistent even when the language differs — regardless of company size.
Process knowledge lives in spreadsheets, shared drives, and people’s heads — with no single version anyone trusts.
They can see the systems — CRM, ERP, SharePoint, Drive — but not how they relate or which one is authoritative.
They need to know exactly what an AI system can and can’t see before they’ll approve it — mixed-access records included.
Most “data readiness” checks one thing: is the data clean. That catches typos and duplicates. It doesn't catch why AI systems actually give wrong or untraceable answers. We audit — and where needed, build — against four groups.
So the AI knows “qualified lead” means the same thing in Sales and Marketing.
So every answer is current, permitted, and traceable to its source.
So the right retrieval method is matched to each type of question.
So the system improves after launch instead of quietly drifting.
The Audit grades each of the 17 components on a simple maturity scale — absent, ad hoc, documented, or fully operational — so priorities are obvious at a glance. The Sprint builds whichever rank lowest and matter most to your first use case. You don't need all 17 before you start — you need the ones that decide whether your first AI system gives correct, sourced answers.
Every engagement opens with a one-page charter signed before work begins: scope, what “done” looks like, and named owners on both sides.
We map every system, document store, and knowledge owner relevant to your target use case, and interview the people who maintain each one.
Each of the 17 components is graded on a clear maturity scale, backed by evidence rather than opinion.
A scored, prioritized backlog: what to fix first, business impact in euros and hours, and the delivery effort required.
We draft the taxonomy and glossary, define the source-of-truth map, and set up the metadata schema and permissions mapping.
We set the retrieval and chunking strategy for your document types, and add citation logic so every answer traces to its source.
We test against real business questions from your team, score the results, and hand over what passed — plus what still needs work.
Useful on its own, and the first artifact of the next stage. Nothing you paid for gets thrown away — even if you choose a different partner, or none at all.
Every one of the 17 readiness components graded on a clear maturity scale.
Scored in euros and hours, ranked by value ÷ effort.
The official system identified for each knowledge domain.
Working definitions for your priority knowledge domains.
A schema and permissions mapping ready to apply in your systems.
Pass/fail against real business questions, with sources cited.
Does the system retrieve the right answer, from the right source, for a real question your team actually asks? We define the expected answer and source before building anything, then grade the result on the same clear maturity scale used across every Liorant service.
The system can’t answer, or answers incorrectly.
A partial answer with no traceable source.
The right answer, but the source is unclear.
Right answer, clear source, and current.
The lineage, permissions, and sources built during readiness are the same artifacts regulators ask for later — built once, not assembled under deadline. This ties directly to the Trust & control group above.
Article 10 requires documented data governance for high-risk AI — lineage, quality checks, and bias evidence. Most business knowledge systems won’t be high-risk, but the discipline is the same: show where an answer came from, or you can’t show the system is under control.
DL 234 created ANIA and a National AI Development Registry. Systems for certain consequential decisions — credit, insurance, biometrics, public services — must register. Clean lineage and documented sources are the foundation of that registration.
The NIST AI Risk Management Framework is voluntary, but its Map and Measure functions — system context, data lineage, testing against defined criteria — are exactly what a Data Readiness engagement produces, and already function as an affirmative defense under some state laws.
| Liorant Data Readiness | In-house hire | Freelance consultant | Self-managed tools | |
|---|---|---|---|---|
| Method | Scored against a 17-component model, evidence-based | Depends on the individual’s background | Depends on the individual’s background | Self-serve; no external scoring |
| Ties directly into an AI build | Yes — deliverables become the Activation knowledge base | Only if the hire also builds the AI system | Rarely — most stop at documentation | No — surfaces data issues, not AI-ready structure |
| Best for | A scored, AI-specific readiness result, fast | Sustained, large-scale data governance | One-off documentation projects | Teams that already know their gaps |
The backlog, taxonomy, and source-of-truth map you receive become the knowledge base for a Rapid AI Activation — Liorant's flagship that builds your first working AI system on what Data Readiness just structured.
Move into Activation within the agreed window and the artifacts carry forward with zero rework. Build elsewhere, or not yet, and everything you received is yours to use as-is.
The Audit assesses and scores your readiness across the 17 components and hands you a prioritized backlog — you decide what to build and with whom. The Sprint does that, then builds the highest-priority components and validates them with a retrieval test, so you leave with working artifacts, not just a report.
No. Most organizations only need three to six components built well for their first use case. The Audit tells you which ones matter for yours; building all 17 upfront is rarely the right use of budget.
No. The deliverables — taxonomy, source-of-truth map, metadata schema — are platform-agnostic and work with Copilot Studio, Google Gemini, Claude, or a custom build. If you move to a Liorant Activation they carry over directly; if you build elsewhere, they’re still yours.
Every engagement runs a retrieval test — real business questions from your team, scored against expected answers and sources defined before we build anything. That score, not our opinion, is what “ready” means.
We'll walk through where your organization stands on the 17-component model and tell you honestly whether an Audit, a Sprint, or something else is the right first step — no slides, no pitch.