As conversational AI, such as chatbots and virtual assistants, increasingly becomes a trusted partner in our digital lives, Momentum is actively exploring the ways we can enhance our workflows. To maximize the value, skillful communication is essential, as AI's capabilities rely on the quality of prompts it receives.
Current language models do favor those good with words, but with practice even simple prompts can be effective. By honing their prompting technique, designers can better understand customer needs, validate solutions, and enhance natural language interfaces.
This post aims to provide accessible yet insightful guidance on effectively leveraging conversational AI tools to empower designers to create more human-centric, useful AI interactions that solve real user needs and support them in their design processes.
Conversational Style and Tone
Direct, linear prompts structured with ideal simplicity and brevity work wonders for optimizing AI comprehension and response quality. When conversing, adopt a respectful, cordial tone of voice, but don’t lapse into being overly informal. Avoid excessive use of confusing slang, unclear abbreviations, or unnecessary emojis that could trip up language processing. Setting the ideal emotional and linguistic tone makes conversational AI more cooperative, useful, and engaging. Keep sentences moderately short and concise and concentrated on one singular topic.
Provide Context with Examples
When starting an exchange, equip the AI with all pertinent details and background to grasp intent and scope. Context setting is everything with prompt engineering; carefully specifying the intended purpose and use of the data that an AI system outputs will lead to improved results. As an example, let's say that you are a product designer wanting to get AI feedback on a new app feature for converting trial users to paid plans. When prompting the AI assistant, first explain that context clearly:
"I am a product designer at Acme Inc. developing our subscription-based mobile app called Stuff Planner that helps people organize their possessions. I am ideating ways to convert more of our 30-day free trial users into paying subscribers when their trial ends. One idea is to add push notification reminders leading up to the trial expiration that highlight premium features unlocked by subscribing. Can you provide feedback on this idea and suggest any alternatives you think may work better based on the app's target demographics of 18-35 year olds?"
This establishes key details like:
- Your role
- The specific product context
- Current business challenge
- Proposed solution
- Relevant user data
With this information supplied upfront, the AI can provide much more useful feedback tailored to the circumstances rather than generic commentary. The more detailed context you provide, the more precise and relevant the AI's response can be.
Ask Focused Questions
Create a productive prompt-response flow by asking questions about specific sections and subsections until the AI’s knowledge has been exhausted. Then, purposefully employ thought-provoking follow-up questions to take conversations to deeper, more nuanced levels of meaningful discourse.
Request that the AI further elaborate and expand upon responses, explain its reasoning, or provide supporting evidence. Pose hypothetical “what if” scenarios that appropriately examine the boundaries of the AI’s capabilities. Allow for, but judiciously steer, any meandering tangents back to the topic at hand when needed. The objective is to graduate beyond basic superficial question and answer exchanges, cultivating genuine back-and-forth dialogue, thinking critically, and organically developing mutual understanding.
Mind AI Limitations
Stay cognizant of the inherent software-based limitations of even the most advanced conversational AI. Refrain from anthropomorphizing with projected human traits, emotions or senses of humor. Maintain realistic expectations aligned with demonstrated capabilities, taking care to independently verify any questionable responses rather than accepting them at face value. Recognize that errors or conversational breakdowns are not intentional defiance, but stem from technical constraints.
Also be vigilant to sidestep other common missteps like excessively rambling, disjointed prompts, peppering in far too many questions simultaneously, or employing cryptic, overly technical language. Exercise patience if the AI companion seems confused or stuck, taking time to thoughtfully reframe the request rather than growing frustrated. A bit of metacognition and mindfulness goes a long way.
In addition, it is important to keep the AI honest to engender trust and veracity. Be specific and avoid hypotheticals to keep the conversation focused on verifiable details. For instance, rather than saying "What if we lowered prices?" ask "According to our 2021 sales data report, how would decreasing our standard price by 15% impact revenue?" You can also request the AI to cite any sources it references in its responses. Adding "Please list the sources you used in your explanation" encourages transparency and allows you to verify the information provided.
Applications for Designers
Effective prompting enables designers to leverage AI for:
- Grasping user needs by having an AI identify additional pain points or use cases from initial research. This comprehensively defines the problem space.
- Getting thoughtful feedback on proposed content strategies. An AI writer prompted on audience demographics/goals can suggest more effective structures.
- Ideating solutions by getting an outside AI perspective on design challenges that yields novel ideas.
- Prioritizing features more effectively when an AI analyzes proposed roadmaps and recommends high-value features based on criteria.
- Tapping AI's potential for enriching dialogue through purposeful prompting.
In summary:
Conversational AI is already becoming integral for designers. As systems like Figma's Genius demonstrate, AI can augment human creativity and productivity. By automating tedious tasks and providing insights, AI will transform workflows. Thoughtfully leveraged, it will become an indispensable creative partner.
However, the effectiveness of conversational AI is only as good as the inputs with which we provide it; prompting will be a skill of the future (indeed, it is already a skill). Yet there are a number of things that we can do to improve not only our results but also our general enjoyment when interacting with our AI companions:
- Structure prompts clearly with a focused, respectful tone optimizing comprehension while avoiding ambiguity that could confuse AI.
- Supply detailed background and examples to establish intent so AI can provide maximally useful, tailored responses vs generic commentary.
- Ask directed questions methodically gathering details on each facet before employing thought-provoking follow-ups to elevate discourse beyond superficial exchanges.
- Independently verify questionable responses, recalibrate expectations, and reframe stuck prompts with patience to account for AI's software-based limitations.
