Confirming patterns Verinode spots

What a "pattern to confirm" is

As your jobs move through Verinode, agent specialists quietly log small observations about how work actually gets done: a step that keeps happening before drying starts, a sign-off that recurs on every mold job, a habit around scope changes. Each observation is captured on its own, tagged with a category, a work type, an inferred step, and a snippet of the evidence behind it. On their own, single observations do not mean much. A step your crew happened to do once is not a pattern.

Every night, a clusterer job reads through the observations still waiting to be reviewed and groups the ones that agree with each other: same category, same work type, same inferred step, same kind of pattern. Once the same behavior has been captured across several separate jobs, the clusterer turns that group into a "pattern to confirm," a claim about how your operation actually works, backed by real evidence pulled from your own jobs. That claim shows up as a tile in the Explore row on your Processes home, waiting for you to say whether it is right.

Verinode does not decide this is one of your standard operating procedures. It surfaces the pattern it noticed and asks you to confirm, correct, or dismiss it. Nothing becomes evidence for anything else, including the Business Analyst's process-mining reads, until you have acted on it.

Where to find it

Open Processes from the sidebar at iq.verinode.ai/processes. The patterns waiting on you show up in two places on that home:

• The money-and-status band at the top counts them under To Confirm, alongside SOPs written and lines without an SOP. When nothing is waiting, it reads "Nothing Waiting"; when something is, it reads "Patterns Waiting On You."
• The Explore row, right under Take Action, carries four KPI tiles (My SOPs, To Confirm, your state's standards, Lines Without An SOP), and then, appended after those four, one tile per pattern still waiting on you. That is where you actually confirm, edit, or dismiss.

Up to eight pattern tiles show at once, oldest first. Clearing the longest-standing pattern first works down the backlog instead of only ever reacting to the newest capture.

What each tile shows

Every pattern tile carries the same layout:

• "Pattern to confirm" in the top-left corner, a fixed label so you always know what kind of tile this is.
• "N× captured", top right, in copper. This is how many separate observations the clusterer grouped together to build this pattern. It only shows up once it has been captured enough times to be a real signal, not a one-off.
• The inferred step, in bold: the behavior Verinode thinks it found, in its own words (for example, "Photograph moisture readings before setting equipment"). If nothing specific was inferred, the tile falls back to the evidence snippet, and if there is neither, it reads "Unnamed pattern."
• A quoted evidence snippet underneath, in italics: the actual wording Verinode pulled from your jobs to support the claim, so you are never confirming a black box.
• A category and work type line at the bottom (for example, "Water Mitigation · Residential"), so you know which part of your business the pattern is describing.

The three ways to respond

1. 1Confirm. Tap Confirm if the pattern is accurate as written. The tile shows "Confirming…" while it saves, then toasts "Pattern confirmed." Every observation that fed this pattern flips to confirmed and becomes usable evidence for the Business Analyst's process reads.
2. 2Edit, then confirm. Tap Edit if the pattern is close but not quite right. The tile swaps to a text box pre-filled with the inferred step, so you can refine the wording. When you are ready, tap Save & confirm (disabled until you type something); the tile toasts "Confirmed with your edit." Your phrasing is stored as the record of what actually happens; Verinode's original inference stays on file underneath so future clustering can still recognize the same behavior. Cancel discards your edit and returns to the original view without saving anything.
3. 3Dismiss. Tap Dismiss if the pattern is not real, or does not matter. The tile toasts "Pattern dismissed." The observations behind it are not deleted or marked wrong, they are simply unlinked from this cluster and go back to waiting. If the same behavior keeps showing up on later jobs, the clusterer can group it into a fresh pattern to confirm down the line.

Where these patterns come from

Patterns are not manually logged, they appear on their own as your specialists work through your jobs, documents, and communications. Nothing in this flow is something you create or track by hand; you are only ever confirming, correcting, or dismissing what already surfaced.

A pattern needs to be seen across more than a couple of jobs before it earns a tile, a single coincidence is not treated as a real behavior. Patterns you never act on do not sit in your queue forever, either: one that has gone unconfirmed and unedited for a long stretch quietly drops off the list on its own rather than nagging you indefinitely. It is not lost, the underlying observations stay on file, they simply stop asking for a decision until a fresh wave of the same behavior brings the pattern back.

Empty states and edge cases

• No patterns waiting. The Explore row simply shows its four KPI tiles and nothing more, there is no placeholder message for zero pending patterns. The To Confirm KPI tile itself reads "0" with "Nothing Waiting" underneath, and its dot preview (which otherwise shows one dot per pattern still open) is empty.
• No evidence snippet. If a pattern was inferred without a usable quote, the tile skips the italic snippet entirely and shows only the inferred step and the category line.
• A confirm or dismiss fails. If the save does not go through (a connection hiccup, for example), the tile toasts "Couldn't confirm, try again" or "Couldn't dismiss, try again" and leaves the pattern in place so nothing is lost.
• You want to see everything being captured, confirmed or not. Open the To Confirm KPI tile or the Most recent row at the bottom of the Processes home. That list, labeled Pending Confirmations, shows every individual observation on file (not grouped into patterns), each with its own inferred step and a quoted snippet, or "Captured by [specialist]" when no snippet was kept.

Best-practice example

Say the Explore row shows a tile reading "Pattern to confirm, 4× captured, Photograph moisture readings before setting equipment, Water Mitigation." Four separate jobs backed the same behavior. If that is genuinely your team's habit, tap Confirm, it is now on record as a real, evidence-backed part of how you run water jobs, and it feeds directly into how the Business Analyst reads your process going forward. If the wording is close but your crew actually photographs readings after setting equipment, tap Edit, fix the order, and Save & confirm, your phrasing becomes the record. If it was a one-off habit on a single crew that is not how you actually want the job run, Dismiss it, nothing is lost, and if the behavior becomes a genuine pattern later, it will resurface.

Related reading

• The decision workspace
• Acting on decisions
• Connecting your data
• Forwarding documents

Data sources