Telling Verinode a signal was helpful or wrong

What this control is

Some signals carry more than a rule-based observation. Once you are past the free tier, Verinode runs a reasoning layer on top of the raw detector output: it takes a cluster of atomic findings, writes a plain-language recommendation, and hands that recommendation to you as a decision card. That recommendation was drafted by a named specialist agent (the piece of Verinode built to reason about that specific domain, carriers, vendors, cash flow, and so on), not by a static rule. Wherever you see one of these AI-drafted recommendations, a small strip sits underneath it: "Was this helpful?" with three responses, Yes, No, and Inaccurate, and an optional note.

This is not a survey. It is the one place in the product where your read on a specific piece of AI output feeds directly back to the agent that wrote it, tagged to that exact run so the team responsible for that specialist can see, in its own words, what it got right and what it got wrong.

Rule-based findings that never passed through the reasoning layer do not carry this control, there is no drafted language to rate. It only appears once a signal has an AI-authored recommendation attached to it.

Where to find it

This lives on the recommendation itself, wherever Verinode shows you one, inside a decision card in the Feed or the full Decisions log. If you came looking for it under a section literally labeled "Pipeline," see Signals: how Verinode watches your data: that name (and later "Signals") described this same detection engine before it split into Feed and Decisions, and both old routes now redirect straight there. The feedback strip sits directly under the bolded recommendation text, in the same block as the dollar impact, the context, any benchmark comparison, and a pre-drafted email if one exists.

Yes: the recommendation was on target

Click Yes (its tooltip reads "Helpful") when the recommendation was accurate and useful, whether or not you ended up acting on it. Your response is recorded the moment you click, there is no confirmation dialog to get through first. The strip immediately collapses to a single line:

Thanks for the feedback!

No note field appears after a Yes. A positive rating does not need an explanation to be useful, it is the signal itself: this specialist read your data correctly and said something worth saying.

No and Inaccurate: telling Verinode it missed

No (tooltip "Not helpful") and Inaccurate are both there for when the recommendation did not land, but they are not the same complaint:

• No is for a recommendation that was probably true but not useful to you right now, wrong priority, already handled, does not apply to how you run this job.
• Inaccurate is for a recommendation you believe got the facts wrong, a number that does not match your books, a carrier or vendor it named incorrectly, a claim about your business that is not true.

Click either one and, same as Yes, your rating is recorded immediately. But because these two responses tell Verinode something went wrong, an optional note field opens right below the buttons so you can say what.

The optional note

The note field is a single line: "What was wrong? (optional)". It only appears after No or Inaccurate, never after Yes, and it is genuinely optional. Two controls sit beside it:

• Send submits whatever you typed, attaching it to the rating you already gave. If you leave the field blank and click Send, nothing extra is recorded, your No or Inaccurate rating stands on its own.
• Skip closes the note field without sending anything further. Your rating from the previous click is already saved either way, Skip only declines to add detail to it.

Either path ends the same way: the strip collapses to:

"Thanks, we'll improve this."

A word on what the note is for: it is read by the people who maintain that specialist's prompt, not published anywhere, not shown to other operators, and never compared against another operator's business. Keep it specific. "This carrier's actual average is closer to 45 days, not 60" is more useful to the specialist than "wrong."

What your rating changes, and what it does not

Rating a recommendation does not rewrite the card in front of you. Verinode never edits or retracts a recommendation because you flagged it, the decision stays on your log exactly as written, and you still decide whether to act on it, park it, or dismiss it (see acting on decisions). What your rating does change is everything downstream of that one run:

• It is saved against that exact execution, tagged to the specialist that wrote it, the model it ran on, and which version of that specialist's instructions produced the output.
• Ratings roll up per specialist into an overall helpful rate that the team watches over time, not per-operator, per-specialist.
• Notes are read verbatim by the team responsible for that specialist. When a version of a specialist's instructions collects a run of No or Inaccurate ratings with notes pointing at the same kind of mistake, that is the signal to rewrite the instructions and ship a new version. Every future run carries the updated version, and the pattern the notes described should stop showing up.

That is the "teaching" part: Verinode does not retrain a model on your feedback in real time, it is the team's evidence for what to fix in a specialist's instructions, and the fix ships to every operator using that specialist once it lands.

Best-practice example

You open a decision on the Feed that says a specific carrier's average time-to-pay has drifted past its own historical norm and recommends escalating a batch of invoices. You know that carrier just switched adjusters and the drift is a one-time transition blip, not a real pattern. Click Inaccurate, then in the note write "Carrier changed adjusters last month, this is a one-time transition delay, not a real slowdown." Click Send. Your read is attached to that exact run for the carrier specialist's team to see, the card itself stays on your log so you can still decide whether to escalate anyway, and if this same kind of miss keeps showing up in other operators' notes, it becomes the specific thing the team fixes in that specialist's instructions.

1. 1Open a decision card that has an AI-drafted recommendation on it, in the Feed or Decisions.
2. 2Read the recommendation, then look under it for "Was this helpful?"
3. 3Click Yes if it was accurate and useful. Click No if it was accurate but not useful to you right now. Click Inaccurate if it got something wrong.
4. 4On No or Inaccurate, optionally add a specific note in the field that appears, then Send it, or Skip to leave your rating as is.
5. 5Keep acting on, parking, or dismissing the decision itself, your rating does not change what is on the card.

Data sources