What Makes a Bot Complicated?
A bot gets more complicated with each thing it has to know, connect to, or do. A bot that answers general questions is quick to build. A bot that looks up live data, knows who it's talking to, and changes records needs much more care. Complexity jumps at three points: when the bot touches live data, when it can change things, and when it has to know who the user is.
This page walks through the levels, from simplest to hardest, and what each one adds.
What Are the Levels of Bot Complexity?
| Level | What the bot does | The hard part |
|---|---|---|
| 1. General knowledge | Answers from the AI model's own knowledge. | It knows nothing about your business. |
| 2. Your content | Answers from your website or documents. | Finding the right passage. |
| 3. Many sources | Answers from several sources at once. | Searching across all of them. |
| 4. Conflicting sources | Picks the right answer when sources disagree. | Deciding which source wins. |
| 5. Private data | Answers from information only you have. | Keeping it private. |
| 6. Automatic updates | Pulls in new content on its own. | Every source works differently. |
| 7. Stale content | Removes information that's out of date. | Knowing what's stale. |
| 8. Learning | Improves from Q&A pairs your team reviews. | Keeping bad answers out. |
| 9. Compliance | Follows your required wording, disclaimers and off-limits topics. | Covering everything it must not say. |
| 10. Lead capture | Collects contact details for your CRM. | Consent and a reliable handoff. |
| 11. Notifications | Emails your team alerts and summaries. | Sending the right things to the right people. |
| 12. Files and images | Accepts files and images from customers. | Reading them safely. |
| 13. Channels | Works on web, phone, email and Slack. | Each channel behaves differently. |
| 14. Languages | Works in other languages, menus and buttons included. | Translating everything, not just the answers. |
| 15. Rich responses | Shows live charts and interactive widgets. | Building and updating each widget. |
| 16. Memory | Remembers people between conversations. | Privacy and how long to keep data. |
| 17. Live data | Looks up information in your systems. | Failures, timeouts and out-of-date data. |
| 18. Missing APIs | Needs a connection that doesn't exist yet. | Building the connection first. |
| 19. Open to AI assistants | Lets Claude or ChatGPT connect to it over MCP. | Limiting what other software can see. |
| 20. Actions | Books, updates or changes records. | Permissions, confirmations and an audit trail. |
| 21. Human handoff | Passes the conversation to a person. | Routing and staff availability. |
| 22. Identity | Confirms who the user is. | Showing each person only their own data. |
| 23. Outreach | Starts conversations with customers. | Opt-ins and timing. |
| 24. Multi-step workflows | Completes a process step by step. | Recovering when a step fails. |
| 25. Change management | Tests every change before it goes live. | Separate test and live versions, and rolling back. |
We add to this list as bots take on new jobs.
Where Does Complexity Jump?
Most levels add a little work. Three add a lot:
- Live data (level 17). Answers now depend on another system being up and correct.
- Actions (level 20). Mistakes now reach outside the conversation.
- Identity (level 22). A mistake could show one person someone else's information.
Do I Need Every Level?
No. Most businesses start with a bot that answers from their own content, follows their rules and captures leads, levels 2 to 10. Live data, sign-in and actions come later, once there's a clear need for them. Starting simple gets a bot working for you sooner, and real conversations show which next step is worth it.
How Does Authava Handle Each Level?
- Answers and accuracy. Every Authava bot answers from your own content, with controls designed to reduce made-up answers. See How Authava Keeps Bots Accurate.
- Keeping content current. Your website can be re-crawled and documents can sync from where your team keeps them. See Our Information Changes All the Time.
- Knowledge that builds up. Q&A pairs your team reviews and edits in Studio. See Q&A Pairs and Tuning in Studio.
- Compliance. Your wording, disclaimers and off-limits topics, with conversation history you can review. See Financial Services.
- Lead capture. Contact details collected during the conversation and sent to your CRM. See Chat Leads Never Make It Into Our CRM.
- Notifications. Email alerts and conversation summaries for your team. See Review Conversations and Get Notified in Studio.
- Files and rich responses. Customers can attach files and images, and the bot can show interactive charts and widgets, such as a booking calendar or order tracker. See A Partial Feature List.
- Channels. The Chatbot for your website, the Phone Bot for calls, the Email Bot for email, and a Slack version of your bot.
- Live data and actions. The MCP Gateway lets the bot call your systems mid-conversation. When a connection doesn't exist yet, our team can scope and build it. See The Integration We Need Doesn't Exist Yet. The Scheduling and Quote modules book appointments and give prices from your own rules.
- Human handoff. Live Chat brings in your team on the website, and the Phone Bot transfers calls.
- Identity. The Portal Bot works inside your signed-in website or app, so users don't log in again. See How Do Bots Know Who They're Talking To?
- Outreach. The Outreach Bot sends a short guided conversation by link.
- Running it. Our managed services team sets up, tunes and maintains your bot, and the weekly Growth Report shows what people are asking. See How is my data safe with Authava? for security.
Frequently Asked Questions
What is the simplest kind of AI bot?
A bot that answers from what the AI model already knows, with no content of your own and no connections to other systems. It's quick to set up, but it knows nothing about your business.
Why does connecting a bot to live data make it harder?
The bot's answers then depend on another system being up, fast and correct. It has to handle errors, slow responses and out-of-date data without giving a wrong answer.
Why is knowing who the user is such a big step?
Once a bot shows personal information, a mistake could show one person someone else's data. The bot has to confirm who it's talking to and show each person only their own information.
Do I need every level of complexity?
No. Most businesses start with a bot that answers from their own content and captures leads, and add live data, sign-in and actions when there's a clear need.