The owner of an eighteen-person services firm asked a simple question on a Tuesday morning. How many active clients do we have? Her finance lead said 142. Her operations lead said 119. Both were holding spreadsheets they trusted. Both were right under their own definition of active. Finance counted anyone with a live retainer, paid up. Operations counted anyone the team had delivered to in the last ninety days. Neither had ever written down what active actually meant. They had been working with the same phrase and three different numbers for two years, and the gap had only become visible because the owner had started thinking about a forecasting tool that would need to ingest both spreadsheets.
The pattern is everywhere. Every small business runs on shared vocabulary that has never been written down. Active client, in pipeline, delivered, wrapped up, complete, sold, signed off. Humans negotiate the ambiguity in conversation, in real time, by asking. AI tools cannot. They take whatever definition they are given and produce confident answers built on quietly inconsistent inputs. The cheapest and highest-impact AI readiness investment a small firm can make is a one-page business glossary, twenty or thirty terms, agreed by the team, reviewed every quarter. Almost nobody has one.
What is a business glossary?
A business glossary is a short document that defines the terms your business uses to make decisions, in plain language, with a named owner per term and a review date. It answers one question for each term, what do we mean when we say this, and how would two people independently reach the same answer. Alation’s reference is that the glossary belongs to the people who use the terms, not IT.
The shape is deliberately small. For a ten-person services firm, fifteen to twenty terms is usually enough. For a thirty-person firm, thirty to forty. Each entry carries the term name, any synonyms or abbreviations, a plain-language definition of forty to one hundred words, a named owner, a calculation rule if the term is a metric, and a last-reviewed date. OvalEdge’s template approach groups terms by domain, finance, sales, operations, customer. A small firm rarely needs more than two domains to start. The discipline is to keep it short, current and accessible. A two-page glossary that everyone uses beats a fifty-page glossary that nobody reads.
Why does it matter for your business?
The cost of undocumented vocabulary is real and distributed. It sits inside chased invoices when sales and finance disagree about what counts as a closed deal, and inside missed forecasts when active client means three different things across the same dataset. IBM’s research on data quality puts the downstream cost of definitional drift in the millions for a large firm, and proportionately bigger as a share of revenue for a small one.
A single field that means different things to different teams breaks every report it appears in, and every decision that report informs. The AI angle sharpens the cost. Gartner’s prediction is that 60 percent of AI projects will be abandoned through 2026, and the headline cause is that organisations cannot turn pilots into measurable value because they never specified what they were measuring. AtScale’s analysis of semantic drift in enterprise AI puts the same point in operational language, a glossary is the controlled, business-ready interface that AI systems can safely consume.
Where will you actually meet it?
Five categories of term consistently need defining in SME glossaries. Customer and prospect states, prospect, lead, qualified, active client, inactive, lost. Deal stages, in pipeline, proposal sent, verbal yes, signed. Work statuses, in progress, complete, delivered, signed off, invoiced. Performance metrics, project margin, utilisation, client satisfaction, gross retention. Role and responsibility boundaries, where sales ends and account management begins, where delivery ends and support begins.
Each of these is a frequent source of two-people-two-numbers conversations. Salesforce’s reference on sales pipelines is that proposal can mean the sales team drafted a document, or the customer received it, or the customer responded positively, and the same word covering three states is what makes forecasts unreliable. Aspire’s metrics glossary makes the same case for performance metrics, customer acquisition cost calculated two ways will give two different answers, and the team will not notice until a board paper has both numbers on the same page. The fix in every case is the same. Pick one definition, write it down, name an owner, agree a review date. Move on.
How do you actually draft one?
The process that works at SME scale is ninety minutes with the owner and two or three senior people who deal with these terms daily, typically the finance, operations and sales leads. Before the meeting, send a short prompt, what are the three to five terms that come up in meetings where different people seem to mean different things. Atlan’s guide to creating a glossary calls this surfacing the live disagreements.
In the room, spend the first thirty minutes listing terms, not defining them. How many active clients do we have. What counts as delivered. When can we say a deal is closed. What is project margin. The list will run to twenty or thirty items. Spend the next forty-five minutes drafting definitions for the most contentious five or six, one at a time. For each, ask what definition the team can live with for the next quarter. Write it down on the spot in plain language, the way you would explain it to a new hire on their first morning. Name the owner. Note the review date. Commit. Spend the final fifteen minutes agreeing how the rest will be filled in, by whom, by when.
The day after, take fifteen minutes in an all-hands or team sync to walk the team through the glossary. This is the part many owners skip and it matters. It signals that precision on vocabulary is the new norm, that ambiguity has had a cost, and that the leadership team is committing to a shared language. A Google Doc or a Confluence page is fine for storage. The Information Commissioner’s Office expects UK firms of every size to keep records accurate, and a documented glossary is part of how that expectation gets met without drama.
When does this become an AI readiness investment?
The moment any AI tool is sitting on top of the firm’s data. A forecasting model, a knowledge assistant, a customer profiling engine, a process automation bot, each of these ingests whatever vocabulary the organisation is using. When active client has meant three things across the historical data, the model learns three patterns and produces answers that look confident and are silently unreliable.
A human analyst would spot the inconsistency and ask. A forecasting model or an LLM amplifies whatever vocabulary it is handed.
The cost of writing a glossary is a few hours of leadership time and a shared document. The cost of skipping it is harder to see and bigger over time. It shows up in wasted licence fees on the AI tool that produced unreliable output, in the founder pulling the plug on a pilot that never worked because the data underneath was inconsistent, in the slow erosion of trust in dashboards and forecasts. The proportionate response is ninety minutes, three senior people, and a one-page document committed to for the next quarter, not a year-long data governance programme. The day a small business decides to take AI seriously, the glossary is the first thing to write.
If you want help drafting the first version for your own firm, or sense-checking it before the team meeting, book a conversation.
