Agentic Mode transforms your Aiden product finder into an AI-driven advisor that holds a real back-and-forth conversation with shoppers. Unlike Scripted Mode and Conversational Mode — where the customer clicks through a fixed sequence of questions — Agentic Mode lets the shopper ask follow-up questions, request alternative recommendations, and much more. You stay in full control of the advice logic and product data; the agent takes care of the conversation.
💡 Agentic Mode is not available by default. Contact [email protected] to explore your options.
Why Agentic Mode?
Shopping online is increasingly driven by conversational interfaces — from ChatGPT to service chatbots. Shoppers expect answers and advice. But most AI chatbots are unpredictable: they don't know your product catalog, they ignore your commercial logic, and they often give unreliable advice.
Agentic Mode takes a different approach.
Your product knowledge — the questions, answers, matching rules, and product data — stays exactly as you configured it in Aiden. The agent uses that foundation to have a natural conversation, score products against the shopper's answers, and explain why a product fits. You get the best of both worlds: a genuinely conversational experience and consistently reliable advice.
How does Agentic Mode work?
Want to see it in action? Try it in the Aiden demo store.
Agentic Mode is not a generic chatbot. The agent works from your existing advice logic, and uses your question flow, answer options, matching rules and product data as the foundation for the conversation.
Here's what happens during a session:
The shopper opens the advisor and the agent introduces itself with a start message.
The agent uses your question flow as its foundation: it starts with your first question, then guides the shopper naturally from there — asking one question at a time, confirming answers, and handling relevant follow-ups.
As the shopper answers, the agent scores products in the background using your matching rules.
When enough information is gathered, the agent presents its top recommendations — and can explain why each product fits based on the shopper's answers.
The shopper can go off-script and ask follow-up questions (e.g. "Is this one good for daily use?") and the agent responds using the product data.
How does the agent score products?
Every answer option in your advice model has a fit label — perfect, good, neutral, or bad — defined by your matching rules. The agent looks up those labels for each product and adds the corresponding scores to produce a total fit score. Products are ranked by that score, and the agent recommends the top results.
For off-script questions — things a shopper asks outside the conversation flow, like "Is this one suitable for sensitive skin?" or "How does this compare to that one?" — the agent takes a different approach, using your product data and matching rules as a combined source of truth. It won’t speculate or go beyond what your data supports.
The agent uses your published advice logic by default. In Preview, it uses the unpublished (draft) version — so you can test changes before going live.
Good to know before you start
Agentic Mode builds on your existing advice logic and product data, so it's important that foundation is solid.
Unlike Scripted Mode and Conversational Mode, the agent doesn't follow a fixed question order. It uses your conversation script and matching rules as a framework, but decides for itself when and how to ask each question based on the flow of the shopper conversation. This means the order you set in Conversation doesn't constrain the agent — it uses the topics and answer options to gather what it needs, in whatever sequence feels natural. What does matter is that your questions cover the right topics and that each question and its answer options are clear and self-contained. The AI co-pilot can help you with this!
Start with Aiden Agentic Mode
Ready to launch your first agentic advisor? Follow these steps.
1 — App settings
Go to Personality to configure your advisor.
Avatar and start message
Avatar image: Upload an icon or image to give your advisor a face.
Name: Keep it simple — a first name works well.
Role description: The text shown under the name. Note: under the EU AI Act, you must indicate this is an AI.
Tone of voice
Customize how the agent communicates to match your brand:
Switch to Use custom tone of voice instructions.
Write instructions describing your preferred style — one aspect per instruction (e.g. formality, sentence length, personality). English or the app language both work.
Optionally paste up to 3 example texts that capture your brand voice (in the app language).
Save your settings.
Test in Preview → Agentic Mode and refine if needed.
Publish your app.
💡 Note: Tone of voice instructions affect communication style only, not the content of recommendations or the advice logic.
2 — Look & Feel
In your existing Look & Feel templates, you'll find a new Chat tab. Use it to adjust the appearance of the advisor — fonts, bubble colors, and more.
Once you've configured your settings, test thoroughly in Preview → Agentic. Have several conversations, including edge cases, before going live.
3 — Embed
Embedding Agentic Mode works just like Scripted or Conversational Mode: go to your finder → Embed → Yes, I've set up the script. Select Agentic Mode, choose Sidebar app, and copy the embed code.
Example:
<!-- Do not use this example code, copy the code from the Embed page in Aiden -->
<button
data-advisor-id="00000000-0000-0000-0000-000000000000"
data-aiden-mode="agent-sidebar"
style="
border: none;
background: #F5C402;
border-radius: 0.25rem;
padding: 0.75rem 1.25rem;
color: #000000;
font-size: inherit;
font-weight: bold;
cursor: pointer;
">
Start Product Advisor
</button>
Already launched a product finder in Scripted Mode, and want to switch to Agentic Mode? Even easier. You only need to replace the text "sidebar" after data-aiden-mode= with "agent-sidebar".
Scripted Mode:
Agentic Mode:
Done? You're ready to launch your first advisor in Agentic Mode. 🎉