Compass Adds Custom Call Drivers That Must Pass Testing Before They Go Live

November 18, 2026

Many call-category tools stop at letting customers add their own labels. Compass goes further: it drafts a customer's call drivers, the reasons customers call, from their own conversations, proves them against real calls before they go live, and carries them through scoring, coaching and Insights.

Why generic categories stop sounding like the business

Every contact center has its own vocabulary for why customers call. A tire retailer tracks warranty claims, trade-in appraisals and service reschedules. A credit union tracks dispute filings, loan status checks and account-access issues. A legal services firm tracks intake, retainer disputes and matter updates. Force all of them into one generic taxonomy and the reports stop sounding like the business.

Drafted from your own calls, not a template

An admin opens Insights and asks Compass to define a driver taxonomy. Compass reads the customer's recent call history, drafts a taxonomy grounded in what customers are actually calling about, and presents it for review. The admin edits, merges, renames or adds drivers in the same conversation, in plain language. No engineering ticket.

This matters because taxonomy drift, the gap between the categories a system tracks and the reasons customers actually call, is a common reason QA data loses its value. If drivers were defined two years ago and the business has changed since, your conversation quality reports are measuring yesterday's business.

Proven before it goes live

A taxonomy that reads well on paper can classify badly in practice.

Compass runs a batch of recent conversations through the draft and shows the admin how each call would have been classified. Drivers that misfire are visible immediately, and the admin adjusts and reruns in the same session. This is the same proof-before-production pattern Compass applies to custom scoring objectives: test against the customer's own data before anything ships.

Activation is gated. A taxonomy can move to production only after passing a fresh backtest. Taxonomies are also versioned as draft, active or archived, so teams can iterate safely and roll back if a change causes a regression.

Drivers that feed every Compass output

In many platforms, call reasons are a reporting dimension: you can slice a dashboard by them, and that is where their job ends.

In Compass, drivers sit underneath everything. They feed paradigm classification, lift scoring, conversation coaching and every Findings Intelligence question an operator asks, and all of that scoring adapts to a new taxonomy without a model retrain. When a customer's taxonomy changes, Compass automatically rescores the historical calls whose lift eligibility flips, so trends stay consistent instead of resetting on day one.

Compass also classifies conversations semantically rather than by keyword matching, so drivers reflect the intent of a call, not whether a specific phrase was said. "I need to reschedule the mechanic coming out next week" is a service appointment reschedule whether or not the customer uses those words. For a fuller explanation of how this works, see How AI evaluates customer conversations.

For regulated verticals the effect is sharper. A credit union's drivers can mirror the exact categories their examiners care about, see NCUA exam readiness. A bank's drivers can map directly to BSA/AML workstreams, see banking BSA/AML. The taxonomy earns its keep because it speaks the vertical's language, not a vendor's.

That is the version of conversation intelligence that actually gets adopted.

Availability

Custom call drivers are available now to all Compass customers. Admins with the settings.edit_scoring permission can draft, backtest and activate taxonomies. Everyone else benefits automatically: their evaluations, lift scores, coaching and Insights answers reflect the drivers the team has defined.