Risk Formulation and Classification Quality Automation: A Vague Risk Never Makes It Into the Register

A risk register invites whatever a consultant is worried about at the moment they open it, and under that pressure, three different kinds of entries tend to blur together: a genuine future uncertainty, a problem that is already happening today, and an assumption being quietly treated as a fact. Each of those needs a different response — a risk needs a probability and a contingency, a current problem needs the diagnosis and an immediate fix, an assumption needs to be tested before it is relied on — but nothing about a blank text field stops them from landing in the same list, worded the same way, with the same apparent urgency.

Distinguishing a risk from a problem, an assumption, or a blame statement requires reading for tense and intent, not just topic — a discipline that is easy to state as a principle and easy to skip under the pressure of writing something down quickly during or right after a client meeting. And a probability rating typically gets chosen by feel in the moment, with no separate step forcing anyone to write down where that number actually came from before it goes into a client-facing register.

Every risk statement is assessed against three required components — a cause ("why might this happen"), an uncertain event ("what exactly could occur"), and a consequence ("what does it cause, and to whom") — each needing a minimum length before it counts as present; missing all three scores the statement "vague," missing some scores it "partial," and only a statement with a genuine cause, event and consequence scores "adequate." Separately, the exact text is checked against seven specific vague-phrasing patterns named as inadequate on sight — including the literal "the project might be delayed" example — and three patterns that attribute incompetence, blame, or unwillingness to a person or team rather than describing a condition anyone could independently verify. A second, independent classifier reads the same text for tense and modality markers: present-tense language ("is already happening," "has been since [month]," "recurringly") redirects the entry toward the diagnosis as a current problem instead of a future risk; "we don't know" or "it's unclear" language redirects it toward critical uncertainties; "assuming that" language redirects it toward the plan's stated premises; a declared budget ceiling or hard limit redirects it toward constraints; and "depends on" language redirects it toward dependencies — each redirect naming exactly where the entry actually belongs and why, not simply rejecting it. On the numbers side, any probability rated above "moderate" is silently capped back down to "moderate" unless a declared basis exists — a named source, a rationale of real length, linked evidence, or a linked anticipatory signal — and "imminent," the single strongest rating on the scale, specifically requires an observed signal, downgrading to "very high" without one. Every financial or numeric estimate is checked for a stated formula, source, assumptions, time period, and limitations, and a single point value with no declared low-high range is flagged and automatically presented with an explicit hedge rather than as a bare number.

A consultant types three notes in a row during a client call. The first — "sales could drop if the main competitor launches a promotion, cutting this quarter's revenue by a meaningful margin" — is accepted as a well-formed risk: it names a cause (a competitor promotion), an uncertain event (a sales drop), and a consequence (reduced quarterly revenue). The second — "we are already losing about three days per approval cycle" — is flagged by the tense classifier as a present-tense problem, not a future risk, and redirected toward the diagnosis with an explanation of why. The third — "assuming the new pricing page ships by month-end" — is redirected toward the plan's premises, because it describes a condition being assumed true for planning purposes, not something uncertain that might occur.

A risk register stops silently absorbing present-tense problems, untested assumptions, and vague worry statements that nobody could actually act on. A probability rating carries a stated, checkable reason instead of sitting on a feeling, especially at the top of the scale, where the language gets used to justify the most urgent client conversations. And a numeric estimate is never handed to a client dressed up as more precise than the data behind it actually supports.

A team that wants a completely freeform notes list, with no structure enforced on how a worry gets written down, will find this formulation gate adds friction rather than value — it is built specifically for a consultant maintaining a formal, client-facing risk register, not a scratch pad for unstructured brainstorming.

A freeform list accepts whatever gets typed, so a present-tense problem, an untested assumption, and a genuinely vague worry all sit indistinguishably next to a well-formed risk. Centriu Oracle's formulation gate instead checks the actual structure and tense of every entry before it settles into the register, redirecting anything that is not really a future risk to where it actually belongs, and specifically naming the vague or blame-based phrasings the methodology considers inadequate on sight.