Skip to main content
Skip to content

Opportunities

What Opportunities finds in your workspace, how each finding is measured, who can see it, and what you can do with it.

Opportunities reads the records your workspace already keeps — agent runs, approvals, connections, procedures, goals, projects, and, where you have them, the pipeline, customer issues and spend — and lists the patterns in them that are worth a person's attention. Each finding states what was measured, over what period, and what it might be worth acting on.

Findings come from fixed checks run against those records, not from a language model. A check only reports when there is enough evidence behind it: a handful of slow approvals is noise, so a check waits for a minimum number of cases before it says anything. The checks run about once an hour.

The four kinds of finding

Problem
Something is already costing time or money, such as approvals that wait a long time or an agent whose runs keep failing.
Opportunity
Something that could be better, such as routine work done by hand often enough that an agent could take the repetitive part.
Risk
Nothing has gone wrong yet, but the exposure is real, such as most of the open pipeline sitting with one customer.
Prediction
Where something is heading if it carries on as it is, such as a goal on pace to finish well short of its target.

Reading the evidence

Every figure on a finding carries a label saying how it was obtained, so a count is never presented as if it were a forecast.

LabelMeaning
CountedA number of records, counted directly.
RecordedA value read as it was entered, such as a goal's target or a deal's value.
CalculatedArithmetic on counted or recorded values, such as a median wait or a share of a total.
ProjectionAn estimate of where something will end up. Treat it as a direction, not a fact.

Confidence — high, medium or low — reflects how much evidence the check had and how stable the pattern is. Links on a finding open the records it was measured from, on the screens that own them.

Predictions

A prediction lists the assumptions it rests on, usually that the current rate continues. When the period it describes ends, Hawi records whether it came true or did not happen, and the screen shows that track record so you can judge how far to rely on the next one.

Who can see a finding

A finding is only ever shown to someone who could already see the records behind it. Findings about spend and credits are limited to the people who manage the workspace's money. Findings about the pipeline need access to the pipeline and to customer identities, and findings about customer conversations need access to conversation history. A finding you are not allowed to see does not appear in your list, in notifications, or in Ask Hawi.

Stored findings do not contain customer names or message text. Repeated customer issues, for example, are grouped without storing what the customers wrote. Opening a finding that draws on customer records is recorded in the workspace's access log.

What you can do with a finding

Create a project
Starts a project in this workspace from the finding and links the two, so the finding shows the work that was started.
Ask for a decision
Opens a decision request for the people who need to weigh in, linked back to the finding.
Watch
Keeps it on your list without starting anything.
Snooze
Hides it until a date you choose.
Dismiss
Removes it with a reason. Saying the figures are wrong is the most useful reason, because it tells Hawi a check needs looking at.
Ask Hawi
Opens the assistant with a question about the finding already written, for you to edit and send.

A finding does not start work by itself. Projects and decisions are created only when somebody with edit rights in the workspace chooses to, and every action is added to the finding's history and the workspace's work record. Members who can view but not edit see findings without the action buttons.

Tuning what you see

Workspace owners and admins can choose up to three priorities, such as growing revenue or improving reliability, which move related findings up the list. They can also mute categories the workspace does not care about and choose whether owners and admins are notified about important new findings or not notified at all. Notifications are capped per scan, so a busy hour produces a few messages rather than a flood.