From: aidotengineer
Current trends in AI integration often involve simply adding chatbots to products, which may not genuinely assist users [00:00:02]. Arthur, a product designer at Evil Martians, challenges this approach, advocating for a different perspective on integrating AI into products [00:00:19].
Challenging the Status Quo
Arthur’s work on Tigon, an AI issue tracker, involved taking the complete opposite approach to conventional AI integration, which proved successful [00:00:34]. The core idea for effective AI has been around for 25 years, reminiscent of Clippy [00:00:48]. While Clippy had terrible execution and timing, the underlying concept was correct, and current technology allows for its proper implementation [00:00:57].
Instead of making AI assistants more reactive, the focus should be on proactive AI that anticipates user needs [00:01:09].
Proactive AI in Practice (Tigon Examples)
Tigon demonstrates AI that doesn’t wait for user input but rather observes real-time user actions, understands context, and intervenes at the opportune moment with relevant, contextual questions [00:01:22].
Suggestion Mode
In “suggestion mode,” the AI tracks what the user is writing in real-time, understands the context, and proactively offers specific, non-generic questions to move work forward, all without needing a chat window [00:01:41]. For example, if a user reports a legality issue, the AI immediately asks specific questions related to the issue [00:01:25].
Action Mode
“Action mode” is demonstrated when a user writes an issue that could be split into sub-issues [00:01:58]. Unlike typical reactive AI, Tigon’s AI recognizes the complexity and suggests a better way to organize the work [00:02:03]. It identifies sub-issues, leveraging previous data and general understanding of optimal organization [00:02:12]. This represents AI that truly understands the domain [00:02:29]. Beyond organization, this AI can consider timelines and resources, acting like a constantly attentive project manager [00:02:38].
Question + Action Mode
This combined mode facilitates interaction within the natural workflow, eliminating context switching, extra windows, or chat interfaces [00:02:51]. It asks relevant questions and assists in managing issues more effectively [00:02:59]. Users maintain control and can easily revert any changes with a single click [00:03:08].
Benefits of Proactive AI
This form of AI seamlessly guides the user, eliminating time wasted on trying to articulate needs [00:03:20]. It supports users in creating better work without disrupting their flow [00:03:28]. This pattern, while built for issue tracking, holds significant potential across various professional tools [00:03:33].
Principles for Developing Proactive AI Systems
To foster proactive AI systems in products, three simple rules should be followed [00:03:41]:
Rules for Proactive AI
- AI should supplement user agency, not replace it [00:03:47].
- AI should offer recommendations, never force them [00:03:53].
- AI should be a part of the natural workflow, not interrupt it [00:03:56].
Applications in Other Tools
This approach can be applied broadly, leading to design process improvements with AI:
- Code editors can proactively watch for common pitfalls and suggest improvements, especially valuable for developers learning new languages or frameworks [00:04:02].
- Design tools could make suggestions for accessible design as the user works, removing the need for post-design checks [00:04:12].
- Communication tools could prepare relevant context before meetings or find documents mentioned during calls [00:04:24].
In all these scenarios, the user remains in control, benefiting from an integrated advisor [00:04:31].
How to Integrate Proactive AI into Products
To begin implementing this approach in your own products, consider these steps:
- Identify Friction Points: Look for areas where users pause their work to ask for help; these are opportunities for proactive assistance [00:04:41].
- Recognize User Behavior Patterns: Identify common scenarios where users need help or frequently ask questions, providing clues for automation [00:04:51].
- Prioritize Context: Focus on where users get stuck, as this is where AI can provide the most assistance [00:05:00].
Conclusion
AI interface design is still in its nascent stages, meaning there isn’t a complete playbook of best practices yet [00:05:13]. Simply replicating chat interfaces is not the solution [00:05:22]. It is crucial to experiment, challenge the status quo, and propose unexpected UI solutions in product development [00:05:26].