Conversation basics
Conversational user interface (CUI) or conversation design is the art of making communication with non-human technology natural and helpful.
Good conversation design:
- Makes customers feel heard and requires less effort to interact with
- Doesn’t feel robotic
- Responds to contextual cues and considers the entire conversation, not just one response
- Strives to get the customer to their goal as quickly as possible
- Anticipates customers’ unasked needs
- Lightens customers’ cognitive load by leading with a point of view
Conversational maxims for AI
In the 1960s, language philosopher Paul Grice introduced a cooperative principle to explain how people communicate effectively. He identified four key categories, known as Gricean maxims:
- Quantity: Be informative
- Quality: Be truthful
- Relation: Be relevant
- Manner: Be clear
To be successful, generative AI conversational experiences must follow customer expectations and conversational norms. Conversations that don’t follow Gricean maxims are annoying and frustrating.
Some more conversational maxims that AI experiences should follow:
- Goal-oriented. User goals and needs should be explored via user research as part of the design process.
- Quick and clear. Be concise and clear to save users time and effort. Unlike human conversation, software users expect instant, accurate answers and have little patience for delays—customer research confirms this.
- Turn-based. Functional conversations should avoid long monologues on the part of the system and make it clear whose turn it is at every moment.
- Polite. Design interactions that respect a user’s time and and where they might be emotionally. Don't talk down to customers, or act like something should have been obvious to them. Aim to make them feel seen and heard.
- Error-tolerant. Texting or even spoken language is notoriously imprecise and error-prone. Generative AI experiences must be able to forgive customer errors in typing, spelling, and grammar. They must also understand made-up words like business names or non-English customer names.
AI principles
Foster trust with transparency
As designers, we hold a critical responsibility in shaping a customer’s relationship with AI. AI systems can often feel mysterious, leaving customers confused about how responses were generated or why a decision was made. This lack of visibility can lead to confusion, frustration, and ultimately, a breakdown of trust.
At Intuit, we solve for this by making information accessible without increasing cognitive load. Trust is rooted in clarity. Always be clear about when the customer is talking to a bot, and avoid overselling what AI can do for customers. Setting the right expectations will impress customers more than overpromising and underdelivering on what AI can actually do.
See also: Disclaimers, Status, Errors, Mentioning AI
Focus on the customer benefit, not the technology
Our primary goal is to communicate the tangible benefits and seamless functionality of our products to our customers. Overemphasizing the technology itself can distract from the customer’s goals.
Focus on the customer and the jobs AI is getting done for them. The design patterns should make it clear that AI has taken the action. The content should focus on the work, not what we're doing in the background to make it happen.
See also: Mentioning AI
Aim to reduce cognitive load
AI experiences should reduce cognitive load, not add to it. Avoid making customers read long text blocks, answer too many questions, or make complex decisions. AI should provide a clear point of view and a suggested course of action based on its analysis.
Something to note: AI isn’t a salesperson, so it shouldn’t push products. When presenting a solution, a product or subscription can be options, but those should also be paired with an alternate path that doesn’t cost the customer extra money.
Be brief and clear
Text-based AI experiences are usually in space-constrained areas with limited formatting. In a chat experience, long responses force the customer to scroll, making it harder to read.
Voice-based bots or avatars put the burden on customers to remember what was said. But people can only remember so much at a time.
For any AI experience, be sure to prompt and evaluate for short, organized responses. A customer can always ask the bot for more info if needed. For more on this, see our guidance on prompting and evaluation (Intuit employees only).