Agentic vs. Conversational Design

‍

Let me start with something familiar.

You’ve probably:

• Used ChatGPT or Notion AI
• Played with an AI tool that suggests what to do next
• Or talked to a bot that tried to sound human

Those experiences are shaped by two powerful forces:

👉 Agentic Design

👉 Conversational Design

They often work together. But they’re not the same.

‍

💬  What is Conversational Design?

It’s not just about screens. It’s about designing interactions that feel like a real conversation. Conversational design is the crafting of dialogues between a person and a machine in a way that feels helpful, intuitive, and human.

‍

Think:

💬 Words that sound like they came from a person, not a script

🧠  Interfaces that can handle messy input (typos, slang, half-thoughts)

🤝 Flows that feel more like a helpful nudge than a set of rigid steps

‍

Good conversational design makes tech feel less like a tool — and more like it gets you.

‍

‍

🧠 Core principles of conversational UI:

Human-centered – Reflects how people naturally speak or type, including errors, slang, or ambiguity.

Goal-oriented – Helps users complete tasks (e.g., “book a flight,” “check balance,” “schedule a meeting”) via conversation.

Context-aware – Remembers what was said earlier in the flow.

Tone-appropriate – Uses the right voice (e.g., friendly, professional, witty) based on brand and audience.

Fail-safe – Handles mistakes or unclear input gracefully (“I didn’t get that — do you mean…”).

‍

💡 Where you see it:

• Chatbots (support, onboarding, shopping assistance)
• Voice interfaces (Alexa, Siri, Google Assistant)
• AI copilots or agents (like GitHub Copilot, Notion AI)
• In-product nudges (microtext prompts, AI suggestions)
• Custom GPT-based assistants in apps and services

‍

🧰 What does a conversational designer do?

• Designs flows and branching dialogues (like a script or flowchart)
• Writes microcopy that feels clear and human
• Builds intents/entities for NLP platforms (if technical)
• Works with PMs, AI teams, and engineers to tune tone, timing, and fallback logic
• Tests with users to see what people say vs. what they mean
• Tests with users to see what people say vs. what they mean

‍

🧪 Example:Instead of:

System: “Please input your 10-digit account number.”

Conversational UI says:Bot: “What’s your account number? It’s usually 10 digits — I’ll wait.”

‍

🎯 Why does it matter?

Because AI is becoming the interface.

‍As products integrate AI (ChatGPT, Notion AI, Adobe Firefly), designers must shift from static screens to flows, prompts, and intent.

We’re no longer just designing tools — we’re designing collaborators.

‍Users now expect products to understand context, respond naturally, and take initiative. That’s where Conversational Design and Agentic Design come in — helping us build smarter, more human-centered systems that act, adapt, and assist.

‍

We’re not just designing tools anymore.We’re designing systems that act, respond, and collaborate.

Great — let’s unpack Agentic Design, which is becoming a crucial concept in the age of AI-powered products.

‍

🤖 Agentic Design?

Agentic Design is about building products that do more than just wait for instructions — they help. Instead of acting like passive tools, agentic systems are designed to:

• Understand what the user needs
• Offer helpful suggestions
• Take meaningful actions
• Act like a teammate, not just a toolI

It’s a shift in mindset: from designing static interfaces to creating responsive, collaborative systems.

‍

🎓 The word “agentic” comes from psychology — it describes the ability to make choices and act with intention.

In product design, we apply this to digital agents: chatbots, copilots, AI assistants, etc., that can help the user in an active, goal-oriented way.

✨ Key traits of agentic products:

1. Proactive

It offers timely help without waiting to be asked.

“You usually post on Tuesdays — want me to draft something?”

2. Context-aware

It remembers past actions and adapts.

Like picking up where you left off or adjusting suggestions based on behavior.

3. Goal-oriented

It focuses on what the user is trying to achieve, not just what they click.

“Do you want this formatted for social or print?”

4. Supportive, not controlling

It guides, nudges, or asks — but never forces.

You’re still in charge, always.

5. Naturally interactive

It works through clear, intuitive interactions — not just buttons and menus.

Feels more like a helpful teammate than a machine.

6. Teaches and empowers

An agentic product doesn’t just help — it helps you grow.

It explains actions, offers choices, and builds user confidence over time.

‍

‍

🧰 Examples of Agentic Design in products:

Product / Tool

ChatGPT (Custom GPTs)

Agentic Feature Example

Takes a role (coach, editor, assistant) and acts with intent.

‍

Product / Tool

Notion AI

Agentic Feature Example

Offers to auto-complete docs based on past structure

‍

Product / Tool

GitHub Copilot

Agentic Feature Example

Writes code while you type, anticipates your next step

‍

Product / Tool

Reclaim.ai

Agentic Feature Example

An intelligent calendar that auto-schedules tasks and meetings around your priorities — and adapts as plans change.

‍

Product / Tool

Superhuman

Agentic Feature Example

Auto-drafts emails, reminds you to follow up, prioritizes inbox

‍

Product / Tool

Figma AI (soon)

Agentic Feature Example

May soon help wireframe or design components from context

‍

🎯 Why Agentic Design matters now:

As we shift from tool-based interaction (e.g. clicking buttons) to AI-powered interaction, users:

• Expect more intelligent behavior
• Want to offload repetitive tasks
• Crave tools that understand intent, not just input

Designers now have to think:

“How can this product act on the user’s behalf — while staying trustworthy and transparent?”

‍

🧪 How to design agentically:

Start with the user’s goal: What are they trying to get done — and how can the product help?

Give the system a clear role: Is it assisting, suggesting, organizing, coaching?

Set healthy boundaries: What decisions should the product make, and what should it leave to the user?

Plan for mistakes: Design clear ways to recover if something goes off track — feedback, corrections, or resets.

Let the user lead: Always provide options to review, confirm, undo, or take back control.

‍

‍

What now?

Now, let’s unpack the similarities and differences between conversational and agentic design.

‍

🧠 Agentic Design vs. Conversational Design:

They’re different but deeply connected — think of conversational design as one of the key interfaces through which agentic design comes to life.

‍

‍

‍

Concept

Agentic Design

Definition

Designing products as intelligent, proactive agents that collaborate with users to achieve goals.

Role in Design Stack

Behavior + Intelligence Layer

‍

Concept

Conversational Design

Definition

Crafting natural, human-like interactions between users and systems (text/voice).

Role in Design Stack

Interaction + Interface Layer

‍

💬 How They Work Together

Agentic Design = the brain 🧠

Conversational Design = the mouth and ears 👄👂🏻

An AI product with agentic behavior (e.g. proactively helping you write, plan, or fix something) will often use conversational UI as its primary way to ask, explain, or confirm with you.

Example:

🧠 Agentic system detects you’re working on a social post and notices a missed CTA.

💬 Conversational layer says:

“Would you like me to add a ‘Learn More’ link to this post? It might boost engagement.”

Here, conversational design enables the agentic behavior to feel human and usable.

Another example:

In Notion AI: Agentic design powers the AI to suggest a meeting summary, tag people, or reorganize content.

Conversational design governs how the AI asks, confirms, or presents suggestions in a natural tone:

“Want me to rephrase this paragraph more formally?”

‍

⚠️ Important Distinction:

• You can have conversational UI without agentic behavior — like a basic chatbot that just answers FAQs.
• You can have agentic design without chat UI — like a recommendation engine or smart automation.
• But when AI becomes a proactive collaborator, the conversational layer helps build trust, clarity, and control.

Thanks for sticking with me through these slides!If you have any questions or need clarification, or any feedback for future posts, feel free to reach out

Best regards!

‍

‍