Words Before Work: A Boardroom Glossary Test for UK Data Programmes
Board papers rarely fail because directors disagree on the principle of better data. They fail because directors agree to the same words while imagining different projects. A paper asks for funding to deliver 'real-time reporting', create a 'single source of truth' and appoint a 'data owner' so the organisation becomes 'AI-ready'. The vote is unanimous. Six months later, the scope debate begins.
That is the glossary gap. It is not a failure of intelligence or intent. It is a failure to convert familiar language into decisions the board can actually govern. For UK mid-market organisations, where data programmes often sit alongside lean management teams and stretched subject-matter experts, the gap is expensive. It widens timelines, complicates supplier negotiations and turns ordinary change control into a dispute about first principles.
Why vague language survives the boardroom
Ambiguous terms persist because they are socially useful. They allow a paper to reach consensus without forcing uncomfortable choices. 'Real-time' sounds ambitious without committing to a service level. 'Single source of truth' signals control without naming the system, the reporting layer or the political compromise involved. 'AI-ready' suggests future competitiveness without specifying whether the data is labelled, licensed, documented or lawful to use.
Vendors and internal sponsors also inherit language from previous programmes. Terms are repeated because they appeared in a prior strategy, a tender response or a conference presentation. Nobody deliberately hides meaning, but nobody is asked to define it either. The board approves a direction of travel, not a set of obligations.
The result is a governance problem disguised as a communications problem. If the board cannot define the term, it cannot test whether the project has delivered it. If the term is defined differently by the finance director, the chief information officer and the head of operations, the programme has multiple acceptance criteria before it has written a single requirement.
Four terms that quietly widen scope
Real-time
In one organisation, real-time means streaming events within seconds. In another, it means a dashboard refreshed overnight. Both can be described as timely by the people who own them. The board should ask which decision depends on the data, how often that decision is made and what delay would materially harm it. For a manufacturer in the West Midlands, real-time production reporting may justify new integration work. For a professional services firm in Leeds, a daily refresh may be sufficient for utilisation reporting. The word alone does not settle the architecture or the cost.
Single source of truth
This phrase often means 'the system I trust most'. It can refer to a system of record, a data warehouse, a master data hub or a reporting layer. It can also mean a political settlement about which department stops maintaining its own version. The board should ask which entity is in scope, which system is authoritative, who may change the record and how conflicts are resolved. Without those answers, a single source of truth becomes a slogan that every function can support while continuing to use its own spreadsheet.
Data owner
Ownership is one of the most overloaded words in data governance. It can mean an accountable executive, a data steward, a technical custodian, a system owner or the person who happens to know the dataset best. In UK organisations with small teams, these roles often collapse into one exhausted individual. The board should distinguish accountability from maintenance. A data owner may be accountable for quality and permitted use, while a steward performs day-to-day checks. If the paper does not make that distinction, the project may appoint a name without creating a workable role.
AI-ready
AI-ready is not a property that can be certified in the abstract. It depends on the use case. A customer service summarisation tool may need clean transcripts and retention rules. A credit decisioning model may need documented lineage, bias testing and evidence for regulators. A forecasting tool may need consistent definitions across regions. The board should ask which AI use case is being prepared for, what data is required, what quality threshold applies and what legal or ethical constraints affect reuse. Otherwise, AI-ready becomes a budget line that expands every time a new model appears.
A one-page exercise before approval
A practical remedy is a board glossary session before the programme is funded. It should take less than an hour and produce one page. The exercise is not a dictionary for the whole organisation. It is a project-specific test of whether decision-makers mean the same thing.
For each material term, record five items: the decision it affects, the board-agreed definition, the evidence that would prove it, what is explicitly out of scope, and who is accountable for maintaining the definition. A simple table works. The value lies in the argument, not the formatting.
| Term | Decision affected | Board-agreed definition | Evidence | Out of scope |
|---|---|---|---|---|
| Real-time | Operational reporting frequency | Data available within 15 minutes of the event | Timestamped test in pilot | Historical backfill beyond 12 months |
| Single source of truth | Customer record authority | CRM is authoritative for contact data; finance remains authoritative for billing | Lineage map and conflict log | Product catalogue data |
| Data owner | Quality accountability | Executive accountable for definitions and permitted use; steward maintains checks | Named role in terms of reference | System administration |
| AI-ready | Pilot readiness | Documented, lawful and labelled data for one named use case | Data quality report and legal sign-off | Enterprise-wide AI platform |
The table exposes disagreement early. If two directors cannot complete the evidence column in the same way, the project is not ready for approval. If the out-of-scope column is empty, the term is probably still too broad. If the definition requires a paragraph of caveats, it may need to be split into separate commitments.
What UK mid-market teams discover
When teams run this exercise, the most common finding is not that the board lacks ambition. It is that the board has been asked to approve outcomes that were never translated into operational language. A retailer in Bristol may discover that 'single source of truth' means one thing to the ecommerce team and another to finance. A business services firm in Manchester may find that 'AI-ready' was assumed to cover a use case that legal has not reviewed. These are not failures of data strategy. They are failures of definition at the point of funding.
The exercise also improves supplier conversations. A statement of work that references 'real-time' without a latency threshold invites change requests. A data owner role without a defined accountability invites confusion. A glossary agreed by the board gives procurement and delivery teams a reference point that is hard to reinterpret later.
Keep the glossary alive without bureaucracy
The risk is that a useful one-page exercise becomes a heavy governance artefact. Avoid that by keeping it tied to decisions. Review the glossary at project gates, not every week. Version it alongside the business case. Retire terms that no longer affect funding. The aim is not linguistic purity; it is a shared test of what the money is meant to buy.
Data governance is often described in terms of policies, catalogues and controls. Those matter. But before any of them can work, the board needs a common vocabulary for the choices it is making. In 2026, UK organisations face pressure to adopt new tools and demonstrate progress. That pressure makes the glossary gap more dangerous, not less. A project approved without definitions is not a fast start. It is a deferred argument.
The test is simple enough to run before the next board meeting. Ask two directors to define the same term in the paper. If their answers differ, the board is not yet funding one project. It is funding several.