Conversational Design and Interaction Flow Optimisation

Written By Alesya Podlesnaya

Context

A fintech company was building a new AI-driven assistant inside their platform, but conversations felt robotic, inconsistent, and disconnected from the brand tone. Teams struggled to design coherent interaction flows, manage edge cases, and maintain consistency across multiple conversation scenarios. They needed a scalable way to design, evaluate, and refine conversational UX.

Challenge

  • Fragmented conversation flows created dead-ends and looping states

  • Tone of voice varied across designers and product teams

  • No centralised library of reusable prompts, states, or messages

  • High friction in testing and validating conversational logic

  • Manual flow mapping was slow and difficult to maintain

  • Users reported frustration in complex tasks such as onboarding or troubleshooting

How freska.ai Helped

Freska AI enabled the team to design, benchmark, and optimise conversational experiences using a combination of generative AI, flow intelligence, and automated UX evaluation.

1. Automated Flow Mapping

Freska AI ingested conversation transcripts, existing scripts, and planned use cases to automatically:

  • reconstruct the full conversation flow graph

  • identify states, transitions, edge cases, and dead-ends

  • visualise breakpoints where users became confused or dropped out

This created instant clarity for the design and product teams.

2. Tone & Style Normalisation

The system analysed messages for tone, clarity, and emotional consistency.
Freska AI generated improvements by:

  • aligning tone of voice with brand guidelines

  • simplifying language

  • making responses more empathetic, proactive, and context-aware

  • ensuring consistency across all user intents

3. Conversational UX Evaluation

Freska AI ran heuristics across each step of the flow:

  • Are instructions clear?

  • Are questions structured?

  • Does the user know what to do next?

  • Are there unnecessary loops or cognitive load spikes?

It scored flows and prioritised issues automatically.

4. Generation of Optimised Flows & Content

The platform proposed improved conversational structures:

  • clearer onboarding interactions

  • robust edge-case handling

  • adaptive follow-up questions

  • fallback states

  • improved responses across intents

Designers received updated scripts and conversation trees directly in Figma or their preferred workflow.

5. Developer-Ready Implementation Guidance

Freska AI prepared implementation-ready logic:

  • structured JSON outputs

  • improved intent handling

  • updated system messages

  • recommendations for tool use in LLM orchestration

Impact

  • Conversation success rate increased by 40% for key flows

  • Reduced user confusion through more intuitive branching

  • Consistent brand tone across all AI interactions

  • Faster iteration cycles — what previously took weeks now took hours

  • A maintainable conversational architecture that scales with new features

Result

Freska.ai transformed disjointed conversational scripts into a unified, intelligent conversational experience—delivering clarity, consistency, and a human-like flow aligned with business goals.
The product became more intuitive, helpful, and aligned with user expectations for AI-driven support.

Alesya Podlesnaya
Previous

UX Audit and Web Accessibility Optimisation