Hero AI Agent

Simple, Powerful Training

When you are ready to train your AI Agent, you can either upload your own documents directly or simply provide the address of a public website for the system to crawl. The training pipeline then goes to work: every file or web page is ingested and carefully broken down into smaller, semantically meaningful sections (often called “chunks”) so that even long manuals or detailed knowledge base entries can be recalled accurately.

During this process, unnecessary formatting, repeated passages, and irrelevant boilerplate are stripped away, leaving only the high-value content. The cleaned and structured information is then indexed in a specialised vector database, which allows the AI Agent to search with concepts and context rather than just keywords. This combination of intelligent chunking, noise reduction, and vector-based indexing ensures that when a customer asks a question, the system can quickly retrieve the most relevant information with high recall and precision, grounding every response in your own trusted knowledge base.

Accurate Answers with RAG

When a customer interacts with your AI Agent, it doesn’t just rely on generic large language model knowledge. Instead, it uses Retrieval-Augmented Generation (RAG) to actively pull information from the custom vector database created during training. This means the Agent searches through your indexed knowledge base, finds the most relevant chunks of information, and incorporates those into its response in real time. By grounding its replies in your own documents, policies, and data, the Agent avoids “hallucinating” or providing outdated information.

Every answer is tied back to specific, verifiable content sources, which can be cited or displayed for full transparency. This not only increases trust and accuracy but also ensures that conversations remain compliant, auditable, and aligned with your organisation’s official knowledge. With RAG, your AI Agent delivers responses that are both intelligent and accountable—always current, always backed by evidence.

Personalisation via Webhooks & Authentication

Your AI Agent can go far beyond static, pre-trained knowledge by tapping into your live business systems in real time. Using multiple authentication options, the Agent can securely identify who it is speaking with—whether that’s by recognising the Caller ID on an inbound phone call, prompting the customer to enter an account number and PIN or passphrase, or validating identity through a web chat session. Once the customer is authenticated, the Agent can trigger webhooks that connect directly to your external APIs or databases. This allows it to retrieve dynamic, up-to-the-minute information such as account balances, service entitlements, order statuses, billing history, or CRM records.

The data is then seamlessly incorporated into the conversation, enabling the Agent to tailor its responses to that individual customer’s needs instead of providing generic answers. Because the process happens in real time, customers feel like they are talking to an informed representative of your organisation, rather than an isolated bot. The result is a highly personalised experience that combines the efficiency of automation with the accuracy and relevance of live data integration.

Embed on Your Site

Bringing the power of your AI Agent to the web is as simple as adding a single line of code. By inserting a lightweight script tag into any page on your website, you can instantly deploy a fully featured chat and voice widget that connects directly to the same trained backend as your phone-based AI Agent. This means customers experience consistent answers and seamless support whether they call your business or engage through your site. The widget supports natural text-based conversations, as well as voice interactions for a more human-like experience, complete with speech recognition and text-to-speech playback. If at any point the AI determines that a customer’s needs go beyond automated assistance, the system can automatically escalate the session to a human representative. This handover is powered by a built-in WebRTC client, enabling live voice calls right from the customer’s browser without requiring any extra software or plugins. With customisable styling and flexible configuration options, the embedded widget blends perfectly into your website, providing a professional, always-available channel for customer engagement.

Click on the green AI widget in the bottom right corner now to talk to the Hero AI Agent!

<script src='https://portal.hero.co.nz/api/embed.js?...'></script>

AI‑Suggested Articles

Your AI Agent doesn’t just answer questions — it can also learn what knowledge might be missing. By continuously analysing transcripts from phone calls and chat sessions, the system identifies patterns, recurring questions, and gaps where customers are seeking answers that are not yet well covered in your documentation. From these insights, it automatically generates suggestions for new knowledge base entries, complete with draft titles and content recommendations.

This means your knowledge base naturally grows and evolves alongside real customer interactions, without requiring a manual audit of call recordings or transcripts. Each suggested article can be reviewed and approved by your team before publication, ensuring quality control and alignment with your brand voice. Over time, this feature helps close information gaps, reduces repeated inquiries to your staff, and makes your AI Agent progressively smarter and more capable. In short, your customers benefit from increasingly relevant answers, while you gain a self-improving knowledge system that requires minimal ongoing effort.

AI Receptionist (Included)

The AI Receptionist is a simplified version of the full AI Agent, designed specifically to handle call routing with speed and accuracy. Instead of requiring complex training, you simply define your organisation’s structure—such as departments, teams, or individuals—and provide a short description of their roles and responsibilities. When a caller phones in, the AI Receptionist listens to their request in natural language, interprets their intent, and seamlessly transfers them to the right destination without the need for menus or button presses. If the request is vague or incomplete, the AI can politely ask clarifying questions until it has enough information to make the correct decision.

With voice recognition accuracy of over 99%, the system ensures callers spend less time waiting and more time speaking with the right person. This reduces frustration, minimises hang-ups, and saves your staff from having to manually redirect calls. Since the AI Receptionist is included free with all AI Agent plans, every customer benefits from a smart, efficient, and professional front line for inbound calls.