AI frameworks and agent build
Data, not technology, is the constraint. We find out what you actually have
Most ambitious AI use cases are gated by data that is missing, scattered or inconsistent. Our engineers audit what exists, where it sits and whether it is usable, mine it for what it can already tell you, and turn the findings into specified requirements and the foundations the ambitious work depends on.
Only about a third of AI ideas have the data ready today
That was the number in our last full discovery programme, and it is typical. Three situations recur. Sometimes the data does not exist: energy-optimisation ideas with no sensor data to learn from. Sometimes it exists but cannot be reached: information inside devices that is not available in an analysable form. Most commonly it exists but is scattered, held across the ERP, departmental spreadsheets and line-of-business systems in inconsistent formats that must be reconciled before any AI can use it reliably.
The consequence is an inverse relationship between appetite and readiness: the themes that attract the most ideas, forecasting, market intelligence, knowledge, are often the least ready. Which is why master-data quality emerges in every roadmap as an enabler rather than a nice-to-have, and why the first scoping task of every data-dependent use case is a data audit that turns a reported position into a measured one.
Four kinds of data work
Data audit
For each use case on the roadmap: what data it needs, where that data lives, its format, quality, completeness and ownership, and a measured data position: in hand, siloed, partial or absent. The audit decides sequencing.
Data mining and proofs of value
Where data is in hand, we mine it for what it can already show: field-failure patterns against factory test yields, spend and purchase-order anomalies, KPI automation from data that is compiled by hand today. Early proofs of value that need no new foundations.
Master data and traceability foundations
The single contract repository, the supplier master, the product and lot traceability, the trusted knowledge base: the cross-cutting foundations that the strategic and research use cases stand on, started early because so much depends on them.
Requirements specification
The findings turned into specified, buildable requirements: sources, transformations, quality rules, ownership, access, and the evaluation measures, ready for a build team, ours or yours.
Twenty years inside ERP, warehouse and commerce systems
Data work for AI is integration work. Knowing that a forecast needs sales history is easy; knowing that the history lives in three Oracle instances with different part-number conventions, that the warehouse system overwrites it monthly, and that the export the finance team uses has a manual step in it is the job. The group has built and integrated ERP, warehouse, PIM and commerce platforms for manufacturers, publishers and retailers since 2006, and that is the experience this service draws on.
Audits in weeks, foundations in quarters
A data audit across the use cases of a roadmap takes two to four weeks and produces the measured data positions and a sequenced plan. Data-in-hand proofs of value run in four to six weeks each. Foundation work is scoped as a programme, usually in parallel with the quick wins so that visible value and groundwork proceed together. Every piece is scoped and fixed before it starts.
About data mining and requirements
Do we need a data warehouse first?
Usually not. Most early proofs of value run on data that already exists in one system, and the foundations are built around the use cases that need them rather than as a platform project in search of a purpose.
Can you work with our existing data team?
Yes, and we prefer to. The audit and the requirements are theirs to own afterwards; we bring the AI-specific lens and the integration experience.
What is a data position?
A use case's relationship to the data it needs: low dependency, in hand, siloed, partial or absent. It is assessed in discovery and measured in the audit, and it decides when a use case can start, not whether it is worth doing.
Is this the same as a data strategy?
It is narrower and more useful: the data work the AI roadmap actually needs, sequenced by which use cases it unlocks, rather than an enterprise-wide programme with no immediate consumer.
Find out where your data really is
An audit tells you which of your AI ideas can start now and what the rest are waiting for.