Talka.ai

The context

Designing behavior-aware interfaces for AI-powered customer insights. Led the product’s UX from 0 to 1 — transforming raw interaction data into actionable insights and helping Talka.ai earn a spot in CB Insights’ Top 100 AI Companies (2023).

Talka.ai is an AI platform that decodes customer behavior by analyzing video and voice interactions. The goal: help product and marketing managers uncover pain points, reduce churn, and make smarter decisions — using data that usually stays hidden.

As the lead product designer, I shaped the UX direction from scratch, translating technical AI outputs into clear, human-centered interfaces that made complex behavioral data accessible and useful.

Talka.ai card dashboard organised by acquisition channel and market segment, with filters and a conversation sidebar. — **A stable structure for exploration**
A matrix organises pain points across channels and market segments.

Talka.ai Charts tab with horizontal stacked bars comparing pain-point categories across acquisition channels. — **Change the view, keep the question**
The chart view reuses the dashboard’s categories and filters.

Talka.ai video view with a recording above a transcript, analysis tabs below and call information and filters on the right. — **Read the conversation in context**
The recording, transcript and call details share one workspace.

Talka.ai clustering workspace with expandable customer-problem groups beside a video player and transcript. — **Move between themes and examples**
Expandable groups turn a large collection of comments into themes and subthemes.

Highlights

Challenge

How do you make invisible signals (tone, pace, non-verbal cues) actionable in a fast-moving B2B product?

Strategy

Pair behavioral research with rapid iteration to define a UX that balances complexity with clarity.

Execution

Defined the experience architecture for the first product version
Collaborated closely with AI engineers to translate insights into meaningful UI patterns
Embedded lean research into weekly design cycles to guide and validate decisions

Impact

2023

CB Insights AI 100 recognition, validating the platform's strategic relevance in the global AI startup landscape

100+

Non-verbal cues translated into actionable interface patterns, enabling decision-makers to decode customer behavior with speed and clarity

1-week

Iteration cycles aligned AI research, UX logic, and user validation, accelerating cross-functional delivery and insight-driven product decisions

1

Integration with HubSpot, enabling automatic export of Talka’s conversation insights, like win probability and behavioral flags, into sales workflows

Talka.ai analytics page with category totals, small bar charts and comparison panels for segments and channels. — **An overview before deeper comparison**
Category totals establish the overview before the page introduces smaller charts and comparisons.

Talka.ai comparison panels above a colour-coded bubble plot of clustered pain points. — **Two ways to examine the pattern**
The comparison panels summarise differences between groups, while the bubble plot provides another view of how pain points are distributed.

What I Learned

Designing for AI at this level taught me how to translate subtle behavioral signals into actionable UI design. I built confidence in creating interfaces that foster trust with unfamiliar technology, and reinforced my ability to collaborate deeply with data and engineering teams to ground product decisions in research.