AI controls / Preference systems

AI Muting Promotional Calls

Users wanted fewer junk calls, not more settings.

Product: Equal Call Preference
Role: Product design lead
Scope: Call preference IA, category muting, confirmation loops, Mixpanel-backed results

AI Muting Promotional Calls case study cover

The tension

Users did not want to mute every unknown call. They wanted Equal to understand which kinds of calls they were done with.

What was happening

Users were not rejecting every unknown call. They wanted Equal to understand the kinds of calls they were done with.

What changed the brief

Users were drawing boundaries by intent, so recognition had to come before configuration and every mute needed a visible way back.

Success looked like

Success meant high-reliability category muting, strong caller-mute usage, and clearer user control over promotional calls.

The constraint

AI classification had to be explainable enough to trust, and muting needed an obvious reversal path.

993kusers opened the page

Attention waiting to become control.

469kcaller mutes

36.8k credit-card calls muted inside the same control system.

99.2%category mute reliability

Reliability from the source case study.

The product story

The interface, one decision at a time.

01 / Intent first

Start with the kind of interruption.

Promotional, credit-card, and loan calls are recognizable purposes. The system should speak that language first.

Equal Call Preference category control screen — **Category-level muting matches the user goal.**People choose the kind of interruption they are done with before touching any toggle.

Equal AI homepage with a promotional call label — **The classification appears in context.**A promotional label on the conversation list makes the AI decision visible before the user manages it.

02 / Proof

A strong action needs a visible receipt.

The call summary asks whether the AI understood the intent, then keeps reversal one tap away.

Equal conversation showing AI promotional call mute confirmation — **The AI asks if it got the classification right.**A yes-or-no receipt turns hidden automation into a feedback loop the user can inspect.

Equal unmute caller confirmation dialog — **Reversal is explicit before it changes future calls.**The confirmation protects against accidental unmuting while keeping the consequence plain.

03 / Reversibility

Control stays trustworthy because it can be undone.

Muted callers remain manageable, and successful reversals are confirmed in the same conversation context.

Equal muted contacts management screen — **Muted contacts are a list, not a hidden state.**Users can review who was muted and reverse individual choices without decoding another settings page.

Equal caller unmuted confirmation with undo action — **The final state confirms the change and leaves an undo.**Immediate feedback closes the loop; the five-second undo keeps a mistake cheap.

Equal category unmuting confirmation screen — **Category changes use the same proof pattern.**The interaction stays consistent whether the user is managing one caller or an entire category.

The trust test

The happy path is only half the product.

AI mutes the wrong category

Receipts and reversal keep the user in control after automation acts.

A useful caller gets muted

Muted contacts are easy to review, restore, and understand.

The category feels too broad

The system keeps intent visible first, then lets users refine the control.

The decisions

Three moves carried the story.

Surface intent before toggles

Categories such as credit card and loan were presented as recognizable call purposes rather than implementation labels.

Make silence reversible

Muted contacts and category-level states needed to feel manageable so users could trust a strong action.

Close the loop with proof

Confirmation, undo windows, and muted-call receipts made the system state visible after an AI decision.

What I would test next