Most organisations' data still isn't ready for AI, and the gap is quality and ownership.A recent IBM study has 78% of chief data officers ranking proprietary data as a top-three differentiator. Identifying that competitive moat is easy, realising it isn't. The edge only exists if the data is accurate and someone is accountable for it. Pick the one dataset your first use case depends on, give it an owner and fix its quality.
Funding raised by AI security startup Reco to expand enterprise “agent” governance
techcrunch.comBlue Cross Blue Shield found AI-assisted hospital coding added an estimated $942 million in extra costs over 2 years
techcrunch.com3 briefings this month, with 3 new figures that passed our source checks.
subscribers read every briefing in full, and get each one on WhatsApp. subscribe free · how we research and check sources
| check | ready when |
|---|---|
| Availability | The system that needs the data can reach it |
| Ownership | A named person is accountable for its accuracy |
| Quality | Its accuracy is known and good enough for the decision |
| Permission | Using it with AI is allowed by your data rules and contracts |
| Upkeep | Checks keep it right after launch |
Waiting for perfect data is the most expensive way to start. Work back from the priority use cases, fix the data each one depends on, and let the data strategy grow from real demand.
Readiness is also organisational: clear ownership, the skills to use data well and governance that lets AI touch sensitive data safely. The technology is usually the easiest part.
Four foundations decide whether AI scales: data that is fit for purpose rather than perfect, an architecture that lets AI sit inside existing workflows, a workforce that understands why its work is changing, and governance built in early enough to speed things up rather than slow them down.
how we help: define the value. →our frameworks: the Lumo method · the value framework
Look at five things for each priority use case: data availability and quality, technical architecture, skills, governance, and the appetite of the people who will use it. The result is a baseline and a sequenced plan, not a score for its own sake.
Not necessarily. Many valuable use cases run on data you already hold in operational systems and documents. Build shared data platforms when several use cases need the same data, not before.
Models amplify what they are given. Incomplete, inconsistent or out-of-date data produces confident but wrong output, and teams lose trust in the system. Most AI project budgets underestimate data preparation.
Data that is findable, owned, documented, of known quality and permitted for the intended use. For generative AI that includes unstructured content such as documents and emails, not only tables.
Four. Data fit for the use case: catalogued, accessible to the right systems and with clear lineage. An architecture that integrates, through APIs, event streams and identity, without locking you into one model provider. A workforce that knows why its work is changing and what is in it for them. And governance built in early, so an initiative can be defended to a board or a regulator without slowing it down.