**Type:** Insight **Growth Stage**: <font color="#00b050">evergreen</font> **Last Tended**: 2024-12-16 **Topics**: #AI #conversationalAI #humanbehaviour --- >[!directors commentary] >*this article was inspired upon reflection of a generative AI assistants exercise during a team training day - we walked through a formulaic, step-by-step guide to refine a client brief and prepare a response. There were some follow up questions "if we follow this every time, we will get a winning answer right out of it?" (or something similar) and "we usually like deep answers, how deep was this programmed?" - and I thought in the moment - we are missing the point a bit here. To prove the point I put an unstructured 1 shot prompt as an instruction and question in against the document and refined the output through collaboration - and I think got a better output. I'm not one for these prescriptive prompt templates - I just prefer to work together: human and machine.* **Abstract** “I’m not a robot” —or at least, I’m more than that. Yet so many of us treat AI tools as if they’re lifeless command-line interfaces. We craft rigid prompts, obsess over optimised templates, and reduce these systems to formulaic input-output machines. What if there’s a better way? Instead of trying to control AI with perfection, what if we approached it like a human collaborator? By embracing conversational interactions, we might just unlock its true potential. **The Problem with Formulaic Approaches** The dominant approach to AI interaction feels, well, robotic. The emphasis on “perfect prompts” is great for precision-driven tasks but can leave little room for creativity or collaboration. It’s like working with a brilliant colleague but only ever sending them rigid instructions via email. As we continue at great pace up the adoption curve, I can't help but notice that following templated chains of prompts might be leading us to assume that the output will be closer to, or even 'perfect' and that it can be arrived at by coldly following the series of inputs without engaging in the process along the way. The risk here is that while there might be the promised productivity to be gained - we haven't learned anything new along the way. **Beyond Templates: Why Conversation Matters** Think about the best conversations you’ve had—they’re dynamic, collaborative, and full of context. That’s the kind of interaction AI thrives on. When we treat these tools like calculators instead of collaborators, we waste their ability to engage in nuanced, human-like dialogue. AI might not have emotions, but it’s incredibly good at working within the context you provide. Instead of rigid commands, imagine saying: _“Here’s the problem I’m trying to solve. What do you think? Can we explore some ideas together?”_ This small shift can transform AI from a passive tool into an active participant in your creative process. **The Elements of Effective AI Dialogue** To engage AI conversationally, consider these key elements: 1. **Start a Dialogue**: Frame tasks conversationally. Instead of “Write an email about X,” try “Here’s what I need: a quick, engaging email about X. Can you help me draft it?” 2. **Context Sharing**: Provide background information, goals, and constraints as you would to a colleague. 3. **Keep It Collaborative**: Use phrases like “What do you think?”, “Does this make sense?”, or “Can you suggest something different?” 4. **Iterative Refinement**: Treat responses as starting points for discussion rather than final answers. 5. **Follow-Up Questions**: Push for clarification, explore alternatives, and challenge assumptions. 6. **Feedback**: Inform the AI about what works or doesn’t to guide better outputs. 7. **Mix Methods**: Combine structured prompts with dialogue when appropriate. AI can handle both! **Looking Ahead: Let’s Rethink This Relationship** AI doesn’t need to be treated like a machine that only responds to perfect inputs. In fact, it performs better when we let it engage in the give-and-take of human-like dialogue. This isn’t about making AI human—it’s about recognising how human communication styles can make our interactions with AI richer, more creative, and ultimately more effective. **The Future is Collaborative:** The way we interact with AI has profound implications for how we work, think, and create. By treating these systems as collaborators rather than tools, we open the door to richer, more meaningful interactions. This isn’t about abandoning templates or rejecting optimisation—it’s about recognising the value of dialogue in fostering innovation and engagement. **Conclusion: Let’s Talk Like Humans** The next time you sit down with an AI tool, think about this: what would happen if you treated it less like a robot and more like a collaborator? Share your thinking process, ask open-ended questions, and let the dialogue evolve naturally. You might be surprised at the ideas that emerge. This is just the beginning of what’s possible when we stop talking _to_ machines and start talking _with_ them. The conversation is just beginning. Share your experiences with conversational AI using **#AIInnovation**, **#ConversationalAI**, **#HumanCentredAI**, **#AICollaboration**, **#Talklikehumans** ## Open Questions & Implications - General system prompts and model prompts in concert with conversation should take care of most practical needs - There is a place for strict prompt process and templates if you need more predictability and productivity - I'd like to conduct a few experiments to see how outputs vary between unstructured and structured inputs - I expect as models and tools advance that the more complicated prompt templates will become obsolete --- *This is a living document in my Digital Garden. It grows and evolves with my thinking and represents my personal thoughts and opinions, and is not part of my work at IBM. However, it is part of my desire to contribute a broader conversation on how we 'get things done' - exploring the impact of tools and techniques aligned to my mission to help individuals and organisations create the settings for sustained growth.* ## Growth Log - 2024-11-21: Initial seed planted - 2024-11-24: Major revision - 2024-11-24: Published on https://www.linkedin.com/pulse/im-robot-lets-talk-like-humans-chris-moreton-q922e/?trackingId=4PnCXGGdRPi627rRf7hNOw%3D%3D