Two tools, one goal
You have two ways to answer your visitors automatically: Yaplet, the AI agent, and chatbot workflows. From the outside they look the same — something replies to your visitor without you having to. Inside they work completely differently, and picking the wrong one is the most common reason a first setup disappoints. Both belong to a brand, and each of them is a tab of the brand: AI agent and Workflows.
Yaplet — the AI agent
Each brand has exactly one Yaplet, at Brand → AI agent. There is no list of AI chatbots to pick from any more. Yaplet answers from everything the brand knows — its knowledge bases, documentation sets, uploaded files, website pages we read for you, question-and-answer entries and the product catalog — all managed on one screen, Brand → Knowledge. When a visitor writes, Yaplet searches that material and answers in its own words.
The same material also answers phone calls, writes the suggested replies your team sees in the inbox, and feeds AI social posting. It is one search over one pool, with nothing to connect up.
- Powered by a large language model (LLM)
- Answers open questions it has never seen before
- Replies in the visitor's language automatically — 100+ of them
- Searches content written in up to three different languages at once, each source in its own language
- Links to the knowledge base article or documentation page it used
- Hands the conversation to a person when you have told it to
One naming note, because you will see it all over the dashboard: Yaplet is called an AI agent, and the word chatbot is now kept for the scripted workflows below.
Chatbot workflows — the scripted builder
A workflow is a script you draw in a drag-and-drop builder. It belongs to one brand and lives at Brand → Workflows. The bot follows exactly the path you designed: show a message, offer buttons, ask a question, branch on the answer, collect data, hand over to a person. Every step is deterministic — the same input always produces the same result.
The list has a Runs on column showing where each script can actually run: All channels, or the ones among Chat, Social, Email and Calls that it can run on. A workflow can run in the chat widget, in Facebook and Instagram messages, in email conversations and on phone calls — but some steps rule some channels out. A Start a form step rules out Facebook, Instagram, email and phone calls. On email, a question, buttons, the AI agent or a collect-data step other than the visitor's own email address rules the channel out as well. On phone calls, a Form, Link, Buttons, Reply-time, Request-agent, Collect-data, Ask-question or Stop step does. Hover the column and it tells you which.
- Scripted — fully predictable and auditable; the AI joins only where you add a step that uses it
- Best for structured processes: lead capture, triage, onboarding
- Supports branching logic and data collection
- Can hand over to Yaplet mid-script with the Send to the AI agent step
- Can start when Yaplet recognizes a request you describe, with the User says trigger
If you run more than one brand, you reuse a script with Copy to another brand in the workflow's row menu on that same list — there is no attaching one workflow to a second widget.
At a glance
| Yaplet | Chatbot workflows | |
|---|---|---|
| Where it lives | Brand → AI agent | Brand → Workflows |
| Handles open questions | Yes — it searches what the brand knows | Only the paths you scripted |
| Requires a script | No | Yes — you build every step |
| Predictable output | No — the AI writes each reply | Yes — deterministic |
| Collects structured data | Through API tools on Brand → Knowledge | Built in — the collect-data step |
| Languages | Replies in 100+, searches up to three at once | The language you wrote it in; a message with Let the AI say it on is said in the visitor's language |
| Where it can run | Website chat, Facebook Messenger, Instagram and email (each with its own Answers on switch), and phone calls through phone agents. The same brand knowledge also writes inbox suggestions | Chat, Facebook, Instagram, email and phone calls — where its steps allow |
| Best for | General support questions | Triage, lead capture, onboarding |
When to pick Yaplet
Choose Yaplet when:
- Your visitors ask open-ended, unpredictable questions
- You already have help articles, documentation or files worth answering from
- You want the AI to handle most chats without scripting every answer
- Visitors write in different languages
- You want a link back to the source article in the reply
When to pick chatbot workflows
Choose workflows when:
- You need a fixed path — collect an email, show pricing, route to the right team
- Outcomes must be predictable and auditable
- You are building an onboarding flow or a qualification sequence
- You want a different script running on different pages of your site
Using both at the same time
You do not have to choose one or the other. A common setup: a workflow fires when the widget opens, collects the visitor's name and topic, then hands over to Yaplet for the open questions — or to a person if the case is complex. The Send to the AI agent step inside any workflow hands over to that brand's Yaplet, and when the brand has no Yaplet, or Yaplet does not answer on that channel (its Answers on switches), the step hands the conversation to your team instead. A workflow whose trigger is User says works the other way round: Yaplet starts it when a visitor asks for what you described.
Next step
Ready to start? Set up Yaplet for the first time, or build a scripted bot with Build your first chatbot workflow.