liveBook · Manning

Conversation design

As explained in Effective Conversational AI

Conversational design is the practice of structuring dialogue, interaction steps, and responses to support user goals.

When a user asks for a representative after a bot has already found the needed information, the problem may be the conversation flow rather than the underlying service. Conversation design structures dialogue so users reach useful outcomes with less effort. section 8.1.2 section 8.1.3

A practical mental model is to remove unnecessary decisions, anticipate the user’s likely next need, and communicate what remains in a longer task. Reducing user effort can require more complex dialogue design, backend integration, natural language understanding, personalization, or automation, so improvements should be weighed against cost and measured value. section 8.1.3 section 11.3.4

Claim information: decision-heavy versus context-aware

A burdensome flow says, “I found your claim. Would you like to have this information read to you or faxed?” The caller asks for an agent. A simpler flow immediately reads the claim number, approval, payment, and date, then offers a fax as an optional follow-up. The improved exchange removes a decision before the user has received the obvious information. section 8.1.2

Use these checks when shaping or reviewing a conversational flow:

  • List every step and identify where users abandon, loop, request human help, or face avoidable choices. section 8.1.3
  • Prioritize improvements that benefit the majority of users, using metrics and an approximate 80/20 rule rather than designing for every possibility. section 8.1.3
  • Set expectations before a long task, then show progress with concise status cues, numbered questions, or progress indicators suited to the medium. section 11.3.4
Listing 8.3 Caller asks for agent even though claim search was successful
VIRTUAL ASSISTANT: I found your claim. Would you like to have this information read to you or faxed? 
CALLER: Speak to a representative.

Use the failed claim-search exchange to spot how an unnecessary choice can trigger an agent request. section 8.1.2

Listing 8.4 Claim details are provided immediately if search was successful
VIRTUAL ASSISTANT: I found your claim, number 10012345. This claim was  approved and a payment for $100 was issued on October 3rd. Would you also like to receive a fax with this information?
CALLER: (may reply yes, no, ask for agent, or hang up)

Compare the revised claim exchange, which delivers the likely-needed information before offering a less common fax option. section 8.1.2

How Effective Conversational AI covers it

How deep it goes
Explains flow simplification, trade-offs, prioritization, and progress signaling, but stops short of implementation detail.
Examples
Uses claim-search dialogues, agent requests, immediate claim details, fax options, and progress cues in multistep tasks.
What it assumes
Assumes familiarity with users, business objectives, and process flows while emphasizing user understanding and research.
What you can do after
Evaluate and improve conversational flows by reducing needless choices, prioritizing high-value changes, and signaling task progress.
How it connects
Connects user research, transactional conversations, natural language understanding, personalization, automation, and business metrics.

Also explained in

Compare how 2 books explain conversation design →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Effective Conversational AI. The section links point to the parts of the book it draws on.