If there is a single finding that shapes every AI roadmap we produce, it is this: data, not technology, is the constraint. The tools to deliver most of the opportunities a business identifies already exist and are proven. What determines whether an opportunity can actually be delivered, and how soon, is whether the data it depends on exists, and whether it exists in a usable form. In our most recent full discovery programme, across 250 ideas from five functions, only around a third were supported by data that was ready today.
Three situations that keep recurring
The data does not exist. Energy-optimisation ideas with no sensor data to learn from. Predictive maintenance for machines that are not instrumented. The opportunity is real and the first step is to create or capture the data, which is a project in its own right.
The data exists but cannot be reached. Information inside devices that is not available in an analysable form off the device. Records in a system with no export. Knowledge in people’s heads and in email threads.
The data exists but is scattered. The most common case by far: sales history across three ERP instances with different part-number conventions, quality data in a test system and a spreadsheet, supplier records in the ERP and in procurement’s own files, in inconsistent formats that must be reconciled before any AI can use them reliably.
Appetite and readiness pull in opposite directions
The themes that attract the most ideas in a discovery programme are frequently the least ready. Market, competitor and supplier intelligence drew the largest field of ideas in the programme above and sat furthest from ready, because there was no market-data subscription behind it beyond public sources. Forecasting and knowledge management were similarly rich in ambition and light in ready data. By contrast, the themes the business could act on immediately, content and creative tooling, a productivity assistant, quality and test-data analysis, were precisely those where the data was already to hand.
Plotted, appetite against readiness, the picture is clear and uncomfortable: enthusiasm and readiness point in opposite directions. A roadmap built on enthusiasm alone funds the least deliverable work first.
Five data positions
Every use case in a discovery programme is assigned a data position, assessed from what participants report and confirmed by audit as the first scoping task:
| Position | Meaning | Consequence |
|---|---|---|
| Low dependency | The use case runs on tooling, not on a body of data | Ready when access is |
| In hand | Data exists, accessible, usable | Proceed now |
| Siloed | Data exists but sits unjoined across systems or devices | Integration first |
| Partial | Some strands have data, others do not | Phase by strand |
| Absent | The data does not yet exist and must be created or instrumented | Data-foundation or research pathway first |
The position decides sequencing, not feasibility. A high-value use case with absent data is still high value; it simply starts with a different first step.
What follows for the roadmap
Two things. First, a data-collection and pooling initiative runs underneath the ambitious opportunities as a genuine precondition. That is why master-data quality emerges in the findings as an enabler in its own right rather than a by-product of any one use case, and why a single, trusted knowledge foundation is recommended as an early, cross-cutting investment: it is the ground the majority of the more ambitious use cases stand on.
Second, the work is sequenced by readiness, not enthusiasm. The opportunities with data in hand proceed now, so the first quarter shows visible value; the data-dependent ones enter a research or data-foundation pathway first, so effort is spent building the ground the ambitious use cases will later stand on. Different teams can progress different groups in parallel; the data foundation is not a group to be done “later”.
What it means for leadership
The uncomfortable part of this finding is that the exciting ideas are not the ready ones, and the ready ones are the unglamorous ones. The useful part is that it gives a leadership team a defensible reason to fund the unglamorous work, the master data, the traceability, the knowledge platform, because every ambitious use case on the roadmap can be shown to depend on it.
Data positions are part of the convergence scorecard, the audit is described under data mining and requirements, and the four adoption pathways that follow from them are under our approach.