Frequently Asked Questions

How do I add Decagon to my product?

You can add Decagon to your website or mobile app with just a few lines of code! The full technical docs are here.

How can I customize answers based on the user?

You can easily pass in metadata to the AI about who the user is, which it can use throughout the conversation to personalize its responses. This often includes a user email, user type, etc. This will allow the AI to answer based on whether the user is logged in or not, the user type, or any other data it can fetch on the user from internal APIs.

How does a conversation get escalated to a human agent?

If escalation is enabled, the AI will escalate to an agent when it cannot sufficiently answer the question or it simply does not know the answer. Decagon can also be configured to immediately escalate any sensitive topics.

We integrate directly with your ticketing system (Zendesk, Salesforce, Slack, etc) to create a ticket or facilitate live chat. Your agents can stay in the tools they are familiar with, and users don't need to switch context.

Do you integrate with Zendesk? Slack? Salesforce?

We integrate with any ticketing or chat system that you use internally so you can join the chat whenever needed. Real-time chat, alerts, etc all work out of the box.

How does the AI get trained on our knowledge?

We are able to connect to and sync with any data source you use, from help centers, to knowledge bases, to internal databases. The relevant information is digested by our AI in a way that can be used to answer future questions. Through integrations with your ticketing system, we're also able to teach the AI from large volumes of historical tickets. At the same time, we're able to connect to any endpoints or APIs to fetch real-time information to use in the answer. This information is all fetched and utilized to generate the best possible answer in just a few seconds.

Can your AI connect to X? What about APIs?

Wherever your data or knowledge exists today, we can integrate with it. We're committed to taking on any extra engineering lift as needed.

How do you handle sensitive user data or PII?

We take security and data privacy very seriously. We are SOC 2 Type 2 certified, GDPR Compliant, and don't store PII other than for the conversations themselves. Even when processing historical tickets, they are processed and PII is removed before the AI ingests them.

How do we improve the AI over time?

Our admin portal is a major part of our offering. Through the portal you're able to instantly give feedback and make updates to the AI's knowledge. Moreover, the admin portal gives you the insights and tools to know where improvements can be made. As such, the AI is always learning from new conversations and messages from your agents.

What metrics do you track to monitor the AI's responses?

All industry-standard metrics around conversations, customer satisfaction, deflection, resolution, etc are tracked in real-time in the admin portal. In addition, we offer a whole suite of AI-powered metrics and insights that we're previously not possible without Large Language Models (see below).

What is the admin portal?

Our admin portal is a major part of our offering. It's a web app where you and your team and log in at any point to monitor metrics, analyze conversations, and view all the messages sent by users and the AI. It's a central control console for the AI itself and a place for your team to interact directly with it.

What are your AI-powered insights?

Valuable data is hidden inside your large volumes of data. It's impossible for your team to read every message. Instead, or AI is able to parse all the text, identify major themes, track trends over time, and surface interesting insights about what users are asking about or struggling with. You get a top-down view of potential bugs, feature requests, and product improvements. You are even able to ask the AI open-ended questions about your conversations directly.

Can you help our human agents be more efficient?

Yes! We have separate agent-assist products that plug directly into your ticketing platform (e.g. Zendesk) so that your agents don't need to switch context but can rely our the same models we train for your end users. The AI can do the heavy lifting of surfacing relevant information and drafting suggested answers so your agents can save countless hours and focus on high-level tasks.

What if my team doesn't have many engineering resources?

Not a problem! We're committed to working with you and taking on any engineering lift needed to customize the optimal solution. Many of our customers have chosen us because of our willingness to work with your team and our relentlessness in building the best product for you.

Last updated