5 Things Your AI Agent Can Do That You Haven't Tried

TL;DR: Most agents are built for basic Q&A and never pushed further. File uploads, guided conversation flows via conversation starters, collaboration links, widget themes, and knowledge base tricks can transform a simple FAQ agent into something genuinely sophisticated — without any additional technical skill.

You built your agent. It answers questions. People like it.

Most builders stop at version one. But Alysium agents support file uploads during chat (up to 30MB per file, 20 files per conversation), collaboration links for team access without accounts, domain restriction for private deployments, and multiple distinct agents from the same knowledge base. Here are five use cases worth adding this week. That's understandable — building a working agent is already an accomplishment. But Alysium has a set of capabilities that most users discover late, if at all. Here are five worth trying this week.

1. Let Visitors Upload Files to the Conversation

You can configure your agent to accept file uploads from visitors during a chat session. This changes the interaction from one-directional (user asks, agent answers) to genuinely collaborative.

What this unlocks: a coaching agent where clients upload their homework before a session and the agent reviews it against your framework. A document review agent where someone uploads a draft and gets feedback based on your criteria. A legal intake agent where a prospective client shares a document and gets preliminary orientation.

To enable it: in the widget appearance section, toggle on the File upload feature under the Features panel. Once on, visitors will see a file attachment option in the chat window.

This single toggle turns a read-only knowledge base into a two-way exchange. It's one of the most underused features in Alysium.

The file upload capability changes the agent from a query-answering tool to a collaborative work session. A user can upload their draft business plan and ask "does this conflict with the program requirements you described?" or upload a screenshot of an error message and ask for troubleshooting help. This use case is particularly powerful for coaching agents (clients uploading their homework) and tutoring agents (students uploading their work for feedback). Enable it in the widget Features panel — it's one toggle.

2. Use Conversation Starters as a Guided Onboarding Flow

Most people add conversation starters as a shortcut — a few suggested questions so visitors know what to ask. That's fine. But the smarter use is as an explicit guided flow.

Think about the first interaction a new client has with your practice. You probably walk them through a sequence: what brings you here today, what have you tried already, what outcome are you hoping for. You can encode that sequence as conversation starters that feel like a welcome, not a menu.

Example for a coaching agent: "I'm new — walk me through how this works." / "Tell me about the 3-step framework." / "I'm stuck on [specific challenge] — where do I start?" / "What's the first thing I should read?"

Each starter pulls the visitor into a meaningful first interaction instead of leaving them staring at a blank input field. With up to 5 starters and optional emoji, the welcome screen becomes a genuine entry point — not just a list of prompts.

3. Share Without Embedding Using Collaboration Links

Not every agent needs to live on a website. For agents you share with clients, students, or colleagues — rather than embedding publicly — Alysium generates a direct collaboration link that requires no Alysium account from the recipient.

Anyone with the link opens a real session, their conversation history persists while they use it, and you can see all their conversations in your analytics dashboard. It works perfectly for:

• Client-facing agents you share via email or a client portal
• Student agents shared inside a course platform (Canvas, Teachable, etc.)
• Team tools shared with colleagues who don't have their own accounts
• Beta testing before a public launch

The collaboration link is the fastest distribution path for agents that don't need public discovery. Most builders don't use it because they default to thinking about embedding — but for private or semi-private agents, it's the better choice.

Collaboration links work because they bypass the website requirement entirely. You can test your agent before you have a website. You can share it in an email campaign. You can post it in a community forum or newsletter. You can use it as a landing page destination from a social media bio link. For creators who primarily live in email and social rather than on a website, the direct link may be the primary deployment method rather than a fallback option.

4. Theme Your Widget to Match Your Brand

There are 36 built-in widget themes — from Light and Dark to Synthwave, Cyberpunk, Cupcake, and Retro. You can also set a custom hex accent color or add your own CSS for complete control.

Why this matters more than it sounds: a widget that looks like a generic chatbot gets treated like one. A widget that matches your brand's design language gets treated as part of your product. Visitors unconsciously trust it more.

The live preview panel in the widget designer shows exactly what your visitors will see before you publish anything. If you've been running the default theme since you launched, spend 10 minutes going through the 36 options — there's almost certainly one that fits your aesthetic better than the default.

For advanced users: the custom CSS field is a full stylesheet override. If you want pixel-perfect control over typography, borders, sizing, or animation, you have it.

5. Build Knowledge Base Depth With Targeted Documents

Here's the trick most builders miss: the agents that perform best don't just have more content — they have more targeted content.

A general "about our services" document will answer general questions adequately. But if you add a dedicated document for your top 3 most-asked questions, each one written specifically to answer that question in detail, those questions get noticeably better answers.

Think of it as training the agent on your hardest problems. Write a "common objections" document. Write a "detailed pricing explanation" document. Write a "step-by-step process for first-time clients" document. Each one becomes a retrieval target for exactly the questions it was written to answer — producing better results than hoping the general content surfaces the right detail.

This is especially valuable for agents that handle high-stakes questions: pricing conversations, policy clarifications, technical support. The more specific the document, the more precise the answer.

Seen something in Alysium you haven't explored yet? Log in and dig in — or if you're starting fresh, your first agent is free.

For the full guide to making your agent as useful as possible, see How to Make Your AI Agent Actually Useful (Not Just Cool). For conversation starter strategy, see The Beginner's Guide to Conversation Starters.

The most effective knowledge base isn't a comprehensive document — it's a collection of targeted ones. An agent with 15 focused 2-page documents (one per FAQ, one per feature, one per process) outperforms an agent with one 30-page comprehensive guide, because retrieval precision improves when chunks are topically focused. Building a knowledge base of targeted documents also makes maintenance easier: updating one FAQ doesn't require editing a monolithic document.