TerraZipAll work

SELECTED WORK / 01

A trusted recommendation layer for conversational AI

TerraZip designed and built an integration layer that helps conversational products surface relevant, clearly labeled partner recommendations inside answer flows, with SDK integration, product controls, and end-to-end attribution.

ENGAGEMENTStrategy · Design · Engineering
TYPEConversational AI infrastructure
STATUS Live product
Visit live product
Conversaic recommendation controls and measurement workspace

01 / CONTEXT

A product problem worth making clear.

Conversational products create high-intent decision moments that do not fit conventional interruption patterns. A partner recommendation can be useful when it matches the user’s task, but it can quickly damage trust when commercial intent is hidden or relevance is weak.

Conversaic needed a product system that could make partner recommendations understandable to product teams, keep every user-facing placement explicit, and connect qualified interactions to downstream performance.

02 / APPROACH

Turn the operating model into the experience.

We organized the product around a clear integration journey: apply for access, review placement and policy requirements, integrate the SDK, and monitor a controlled pilot before expanding.

The experience treats no placement as a valid outcome. Sponsored recommendation cards and sponsor-aware follow-up prompts are clearly labeled and intended to appear only when an eligible offer strongly matches the conversational context.

03 / DELIVERED

From proposition to live system.

01

Product onboarding

A structured application and approval path covering product fit, audience, geography, technical readiness, acceptable recommendation patterns, blocked categories, and restrictions.

02

Integration path

A lightweight SDK-oriented implementation flow with product documentation, placement guidance, and a concrete route from approval to a live pilot.

03

Recommendation controls

Clearly labeled recommendation patterns, product controls, eligibility rules, and a no-placement threshold designed to protect answer quality.

04

Measurement workspace

A unified view of recommendation activity, qualified clicks, attributed outcomes, and the performance signals needed to evaluate a pilot.

04 / OUTCOME

The result is a live, end-to-end foundation for partner recommendation pilots: a clear proposition, an integration path, implementation guidance, controlled answer-flow placements, and measurable attribution while keeping partner recommendations clearly identified.

Explore the product